CN104121985A - Selective decimation and analysis of oversampled data - Google Patents

Selective decimation and analysis of oversampled data Download PDF

Info

Publication number
CN104121985A
CN104121985A CN201410173043.XA CN201410173043A CN104121985A CN 104121985 A CN104121985 A CN 104121985A CN 201410173043 A CN201410173043 A CN 201410173043A CN 104121985 A CN104121985 A CN 104121985A
Authority
CN
China
Prior art keywords
value
data
sampling interval
sampling
interval data
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN201410173043.XA
Other languages
Chinese (zh)
Other versions
CN104121985B (en
Inventor
R·E·加维三世
J·A·弗尔巴
S·V·鲍尔斯三世
R·D·斯凯伊里克
H·霍尔特曼斯波特
M·D·梅德利
K·斯蒂尔
D·A·曼恩
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Emerson Electric US Holding Corporation Chile Ltd
Original Assignee
Emerson Electric US Holding Corporation Chile Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority claimed from US14/252,943 external-priority patent/US9778080B2/en
Application filed by Emerson Electric US Holding Corporation Chile Ltd filed Critical Emerson Electric US Holding Corporation Chile Ltd
Publication of CN104121985A publication Critical patent/CN104121985A/en
Application granted granted Critical
Publication of CN104121985B publication Critical patent/CN104121985B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Abstract

Useful and meaningful machine characteristic information may be derived through analysis of oversampled digital data collected using dynamic signal analyzers, such as vibration analyzers. Such data have generally been discarded in prior art systems. In addition to peak values and decimated values, other oversampled values are used that are associated with characteristics of the machine being monitored and the sensors and circuits that gather the data. This provides more useful information than has previously been derived from oversampled data within a sampling interval.

Description

The selectivity of over-sampling data extracts and analyzes
Technical field
The present invention relates to machine performance and fault analysis field.More particularly, the present invention relates to from responding to the analysis of the over-sampling data of the one or more dynamic transducers that contact with machine.
Related application
The application requires to submit on April 29th, 2013, and application number is 61/816,974, the right of priority of the unsettled temporary patent application of the associating of by name " selectivity of over-sampling data extracts and analyzes ", and its full content is incorporated to herein by reference.
Background technology
The modern mechanical analyzer of for example vibration analyzer is the dynamic numerical data of over-sampling on than the sampling rate of the large manyfold of the maximum frequency of data acquisition (FMAX) conventionally.The data of over-sampling are typically converted to the frequency of expectation by filtering extraction or peak filtering.In these methods one or other method are normally used for during sampling interval, by the over-sampling data reduction collecting to single scalar value.By utilizing filtering extraction, described scalar value is generally corresponding to mechanical vibration information.By utilizing peak filtering, described scalar value generally accords with mechanical stress ripple information.The difference of peak filtering and filtering extraction is: filtering extraction is abandoned the data of over-sampling arbitrarily, and peak filtering is optionally abandoned the data of over-sampling.
During sampling interval, over-sampling and filtering extraction mechanical oscillation signal, to obtain the scalar amplitude of the mechanical vibration that sensed, are proposed by Canada first in US5633811.Peak filtering is (also referred to as " PeakVue tM", the trade mark of computing system incorporated company) and the mechanical vibration data of over-sampling represent the scalar PeakVue of stress wave information to obtain tMvalue is described by Robinson first in US5895857.PeakVue tMbe to extract with the difference of extracting and a little at random abandon over-sampling data, and PeakVue tMoptionally abandon over-sampling data, and PeakVue tMto carry out the signal (rectified signal) of having corrected is upper.Leigh (US7,493,230) has instructed a kind of mode of digital decimation, and the mode of described digital decimation is used " a kind of averager is to determine arithmetic mean or the root-mean-square value (RMS) of correcting rear sample ".
Envelope technique is different from filtering extraction and peak filtering.The example of envelope technique comprise root mean square (RMS), demodulation, short-time RMS (STRMS), spectrum emittance (Spectral Emission Energ, the trade mark of SEETM-SKF group), electric spark energy (Spike Energy, be called again gSE, conventionally quoted by Entek IRD) and shock pulse supervision (Shock PulseMonitoring, SPM, is quoted by SPM instrument conventionally).The difference of these enveloping methods and peak filtering and filtering extraction is that described enveloping method has minimum advection or decay inherently, causes envelope not comprise the actual margin of tested value.
For the known technology of trend analysis and the compression of trend data piece, for example use data online or that roaming condition surveillance equipment collects, generally use the maximal value of each, the mean value of each and the minimum value of each.For example, in secular trend, each data point can represent the minimum of 64 record values, maximum and average value.(referring to reference manual AMS tM: healthy for CSI4500 machinery tMthe machinery health of monitor tMkeeper's on-line system application software, numbering #97460.7, Ai Mosheng process management (2007), 3-53 page).
The system and method that is incorporated in full prior art herein in this by reference comprises by Canada (US5,633,811), Robinson (US5,895,857 and US7,424,403), Piety (US5,965,819 and US5,943,634), Baldwin (US2012/0041695), Leigh (US7,493,220) and Leigh (US8,219,361) described.Various embodiment of the present invention is different from above-mentioned all prior aries.
Table 1 is below shown schematically in the various application that wherein digital vibration signal is post-treated and extracts (being labeled as each row of " aftertreatment " and " extraction ").Notice that this chart also represents simulating signal, for example, from a simulating signal of piezoelectric accelerometer, it typically (saw " digital signal " row) and is sent to simulation process step (seeing " pre--to process " row) before analog to digital conversion.Then digital signal is post-treated and extracts continually.Then be extraction step (being skipped if extracted, is post-processing step), for example, use AMS machinery healthy tMmanagement software, analyzes digital vibration signal information, and for example by using machinery healthy tMthe vibration analyzer of administrator software, understands digital vibration signal information.
Table 1: for understanding the process of analog sensor signal information
In table 1 general introduction step modal be by use analog acceleration meter (I) engage analog vibration data acquisition unit (II) or analog vibration analyzer (III) carry out.To from data acquisition unit from the analysis of the digit data stream of vibration analyzer or completing of further analyzing can be by carrying out by programmable computer analyser.
For example, simulation piezoelectric accelerometer can be installed on machine with collection machinery vibration and change mechanical vibration into simulating signal.Simulating signal is typically transmitted in cable, and analog voltage signal is ratio, for example a mV/g.Described cable is also connected to vibration analyzer, for example CSI tMthe hand-held analyzer of type 2140 or CSI tMtype 6500 on-line analysis devices.For example CSI tMthe hand-held analyzer of type 2140 is usually used in analyzing and helps operator to understand vibration signal information.For example CSI tMit is healthy that type 6500 on-line analysis devices are often connected to for example machinery of utilization tMthe personal computer of the vibration analysis software programming of administrator software.On-line analysis device makes operator can analyze and understand vibration signal information with the combination of the personal computer that utilizes vibration analysis software programming.
For example CSI tMthe vibration transmitter (V) of type 4100 and be for example well-designed with the simulation transducer of CSI type 9420 vibration transmitters couplings and be programmed to carry out the deciphering of whole analyses and analysis result.For similar these independently, automanual equipment is understood result and is not needed mankind analyst's existence, has typically replaced situation to monitor that the mankind of analytical information understand at the FPGA (Field Programmable Gate Array) firmware of center processing unit.
The digital transducer of for example digital accelerometer (VI) typically comprises that embedded analog acceleration meter or MEMS sensor or other situation monitor transducer.Before the aftertreatment of the pre-service of simulating signal, analog to digital conversion, digital signal and extracting typically occurs in, by wired or wireless medium, digital waveform or other digit data stream is sent to the receiving equipment of for example computer analyser (IV) or programmable digital vibration analyzer (VII).
Summary of the invention
The general scheme that makes various embodiment described here be different from prior art comprises special digital signal aftertreatment, or the optionally application of (for example, not being arbitrarily) extraction technique, or both.The described post-processing step of listing in table 1 and extraction step can be by completing more than a kind of approach and on more than one position.If need to obtain case of machines information, various embodiment described here can combined statement 1 in row A to two row in G or multirow more, it starts with analog dynamic signal, and situation state or the last result of other significant deciphering formal output to understand.
Embodiments of the invention provide a kind of system, analyze by the over-sampling numerical data to being conventionally dropped in prior art systems the useful and significant information that obtains.Described various embodiment is applicable to by using the over-sampling data that for example the dynamic signal analysis device of vibration analyzer, motor current signal analyzer and motor flux distribution device collects, and it can be used as handheld device, in-service monitoring device, protection system and transmitter and implements.
Except being used in the peak value and decimation value of prior art systems, the preferred embodiment of the invention is used other over-sampling value, and these other over-sampling values are associated with the machinery being just monitored, the collection sensor of data and the characteristic of circuit.Than previous acquired information from over-sampling data in sampling interval, this provides more useful information.
In certain embodiments, machine or equipment situation information, transducer or sensor performance information and electronics or circuit performance information in the sampling interval of each Dynamic Signal, are extracted from over-sampling numerical data.Sampling rate interval is 1/F sR, sampling interval is 1/F mAX, and the number of data point is F in sampling interval sR/ F mAX.For example, work as F sR=100kHz and F mAXwhen=2kHz, in a sampling interval, there are 50 data points.
Some embodiments described here for example comprise programmable logic device, to automatically understand the possibility of the cause and effect data (basic reason) in over-sampling data set, this is by least one realization in following technology: (i) relatively intermediate value and mean value, (ii) minimum value after comparison pattern value and rectification, (iii) maximal value after comparison pattern value and rectification, (iv) poor between standard of comparison deviate and maximal value and minimum value or peak-to-peak value, and (v) calculate measure of skewness or other Statistical Shape factor.
Some embodiment comprise that programmable logic device is with difference cause and effect data and Gaussian data, and statistic evidence based at one or a series of over-sampling data sets, specify possible the situation of selecting from following situation list: impact, sensor fault, fault, machine operation, noise, stable case, random occurrence, system event and environmental parameter variation.Notice that described environmental parameter may be operation characteristic, basic material, temperature or the variation near the cross-talk of machine.
In certain embodiments, programmable logic device operates to process the digital data sets of over-sampling to digital accelerometer data, from comprise the set of intermediate value, mean value, RMS and pattern, obtain the waveform of mid-range value simultaneously, and from comprising maximal value, minimum value, is up to the waveform that obtains maximum magnitude value in the peaked set after minimum peak-to-peak value and rectification.
Some embodiment comprise digital accelerometer data are carried out to FPGA (Field Programmable Gate Array) operation, to process the digital data sets of over-sampling, simultaneously from comprising intermediate value, variance, measure of skewness, obtain the waveform of statistics mid-range value in the set of kurtosis and other statistical value.
In certain embodiments, programmable logic device produce intermediate value waveform simultaneously and correct after maximal value waveform, and deduct intermediate value waveform energy is concentrated on to the peak event occurring in each sampling interval in maximal value waveform from described rectification.This waveform difference can be further by processing by FFT or auto-correlation, to identify characteristic frequency and the amplitude fault of for example impulse fault.
Some embodiment comprise programmable logic device, by the minimum and the maximum value data that use statistical model and collect in the sampling interval of over-sampling, analyze over-sampling data, understand fault sensor.Can in continuous sampling interval, compare these statistical informations, to detect possible sensor fault, thereby avoid the machine error being caused by fault sensor.
In certain embodiments, in the time being identified for the normal state vibration data value of normal state vibrational waveform, by by the over-sampling data that produce due to stress wave and remaining over-sampling data separating, then show described remainder, programmable logic device has improved normal state vibration survey.
An embodiment provides a kind of computerized method, for specifying Gauss's attribute or non-Gauss's attribute by the sampling interval data set that contacts collected over-sampling dynamic measuring data with machine or treatment progress induction.This method preferably comprises following steps:
(a) determine the intermediate value from described sampling interval data set;
(b) determine the mean value from described sampling interval data set;
(c) determine the difference between intermediate value and described mean value described in described sampling interval data centralization;
(d) described difference and the threshold limit comparison that will in step (c), determine, to determine whether described sampling interval data set comprises the possibility of Gauss's normal state data or non-Gauss's normal state data, the difference that wherein exceedes described threshold limit has been pointed out the possibility of non-Gauss's normal state data;
(e) specify Gauss attribute or the non-Gauss attribute relevant to described sampling interval data set; And
(f) extract described sampling interval data set, to obtain at least one scalar value in a series of scalar value that comprise kinetic measurement waveform; And
(g) storage Gauss attribute or the non-Gauss attribute relevant to described data set or waveform.
An embodiment provides a kind of computerized method, and for extracting numerical data, described numerical data is from by responding to the simulating signal that the analog sensor that contacts generates and obtain with machine or treatment progress.This method preferably comprises following steps:
(a) described simulating signal is converted to the digit data stream of over-sampling;
(b) digit data stream of low-pass filtering over-sampling is to obtain machine or process situation information;
(c) digit data stream of over-sampling is divided into sampling interval data set;
(d) analyze at least a portion of described sampling interval data set, to determine the statistical attribute of the data set that is selected from group, described group by intermediate value, mode value, standard deviation value, maximal value, value range, minimum value and by by the value in described group and mean value or reference value comparison and definite fiducial value form;
(e) data set of sampling interval in proper order of analyzing in extraction step (d), to produce the scalar value corresponding with each sampling interval data set;
(f) generate the waveform that is included in the described scalar value producing in step (e); And
(g) preserve the data set statistical attribute relevant to described waveform.
Various embodiment provides computerized method, for the treatment of the dynamic measuring data of over-sampling, the data set that the dynamic measuring data of described over-sampling comprises the multiple over-samplings that collect by one or more sensors that append to machine or process, wherein the data set of each over-sampling is corresponding to specific sampling interval.In the first embodiment, described method comprises following steps:
(a) determine multiple standard deviation values, each standard deviation value obtains from corresponding one of multiple over-sampling data centralizations;
(b) determine multiple maximal values, each maximal value obtains from corresponding one of multiple over-sampling data centralizations;
(c) determine multiple minimum value, each minimum value obtains from corresponding one of multiple over-sampling data centralizations;
(d) determine multiple the first differences, each the first difference is determined by the difference between maximal value and the minimum value of definite corresponding over-sampling data set;
(e) determine multiple the second differences, each the second difference is determined by the difference between standard deviation value and first difference of definite corresponding over-sampling data set;
(f) one or more the second differences and the threshold value comparison that will in step (e), determine, with the possibility that determines whether that described dynamic measuring data comprises cause and effect data or Gauss's normal state data, the second difference table that is wherein greater than described threshold value understands the possibility of cause and effect data; And
(g) under comparison step (f) has shown situation that described dynamic measuring data comprises cause and effect data, appointment may cause the situation of described cause and effect data, and the free impact of wherein said situation, sensor fault, fault, mechanically actuated, noise, stable case, random occurrence, system event and environmental parameter change in the group forming to be selected.
In a second embodiment, comprise following steps for the treatment of the Computerized method of the dynamic measuring data of over-sampling:
(a) obtain the intermediate range waveform that comprises multiple mid-range value, each mid-range value in multiple mid-range value in wherein said intermediate range waveform is selected from a corresponding over-sampling data centralization of multiple over-sampling data centralizations, and wherein said multiple mid-range value comprise multiple intermediate values, multiple mean value, multiple RMS value or multiple mode value;
(b) obtain the maximum magnitude waveform that comprises multiple maximum magnitude values, each maximum magnitude value in multiple maximum magnitude values in wherein said maximum magnitude waveform is selected from corresponding over-sampling data centralization of multiple over-sampling data centralizations, and wherein said multiple maximum magnitude values comprise the maximal value after multiple bare maximums, multiple rectification, multiple minimum value or are multiplely up to minimum peak-to-peak value;
(c) obtain the statistical straggling waveform that comprises multiple statistical straggling values, each statistical straggling value in multiple statistical straggling values in wherein said statistical straggling waveform is selected from a corresponding over-sampling data centralization of multiple over-sampling data centralizations, and wherein said multiple statistical straggling values comprise multiple variance yields, multiple measure of skewness value or multiple kurtosis value;
(d) obtain the maximum waveform after the peaked rectification comprising after multiple rectifications, maximal value after each rectification in maximal value after multiple rectifications in maximum waveform after wherein said rectification is selected from a corresponding over-sampling data centralization of multiple over-sampling data centralizations, and
(e) obtain in the following manner the waveform merging:
The value that comprises a waveform in the waveform obtaining in step (a) to (d) is added with the corresponding value that comprises another waveform in the middle waveform obtaining of step (a) to (d), or
In the corresponding value of another waveform in the described waveform obtaining, deduct the value that comprises a waveform in the described waveform obtaining in step (a) to (d) from comprise step (a) to (d),
The waveform table wherein merging is shown in the peak value in over-sampling data set.
In the 3rd embodiment, comprise following steps for the treatment of the Computerized method of the dynamic measuring data of over-sampling:
(a) determine multiple mode value, the value or the value scope that the most frequently repeat of each mode value based on appearing in a corresponding over-sampling data set of multiple over-sampling data centralizations;
(b) determine multiple minimum value, each minimum value obtains from corresponding one of multiple over-sampling data centralizations;
(c) determine multiple maximal values, each maximal value obtains from corresponding one of the data centralization of multiple over-samplings;
(d) determine multiple MODE-MIN differences, each MODE-MIN difference is determined by the difference between mode value and the minimum value of definite corresponding over-sampling data set;
(e) determine multiple MAX-MODE differences, each MAX-MODE difference is determined by the difference between maximal value and the mode value of definite corresponding over-sampling data set;
(f) if be less than default threshold value in the above MODE-MIN difference of data set of carrying out continuous over-sampling, determine that in one or more sensors, at least one is fault; And
(g) if be less than default threshold value in the above MAX-MODE difference of data set of continuous over-sampling, determine in one or more sensors that at least one is in saturated conditions.
In the 4th embodiment, comprise following steps for the treatment of the Computerized method of over-sampling dynamic measuring data:
(a) determine multiple maximal values, each maximal value obtains from corresponding one of multiple over-sampling data centralizations;
(b) determine one or more kurtosis momentum values based on described multiple maximal values;
(c) determine form factor by deduct integer 3 from least one kurtosis momentum value;
(d), in the time that described form factor equals zero, determine that over-sampling dynamic measuring data is normal distribution;
(e) in the time that described form factor is greater than 0, determine that over-sampling dynamic measuring data is that spike distributes, and
(f), in the time that described form factor is less than 0, the dynamic measurement data of determining over-sampling is flat distribution.
In the 5th embodiment, comprise following steps for the treatment of the Computerized method of over-sampling dynamic measuring data:
(a) determine multiple mode value, the value or the value scope that the most frequently repeat of each mode value based on appearing in corresponding one of described over-sampling data centralization;
(b) determine multiple intermediate values, each intermediate value obtains from corresponding one of described over-sampling data centralization;
(c) determine multiple MODE-MED differences, each MODE-MED difference is determined by the difference between mode value and the intermediate value of definite corresponding over-sampling data set; And
(d), if the absolute value of one or more MODE-MED differences is less than predetermined threshold value on continuous over-sampling data set, determine the stable measurement situation that exists.
In the 6th embodiment, comprise following steps for the treatment of the Computerized method of over-sampling dynamic measuring data:
(a) determine multiple intermediate values, each intermediate value obtains from corresponding one of over-sampling data centralization;
(b) determine multiple maximal values, each maximal value obtains from corresponding one of multiple over-sampling data centralizations; And
(c) determine multiple crest factors, intermediate value and the peaked difference of the over-sampling data set of each crest factor based on corresponding are determined.
In the 7th embodiment, comprise following steps for the treatment of the Computerized method of over-sampling dynamic measuring data:
(a) determine multiple maximal values, each maximal value obtains from corresponding one of over-sampling data centralization;
(b) determine multiple minimum value, each minimum value obtains from corresponding one of over-sampling data centralization;
(c) determine multiple MAX-MIN differences, each MAX-MIN difference is determined by the difference between maximal value and the minimum value of definite corresponding over-sampling data set;
(v) determine multiple standard deviation values, each standard deviation value obtains from corresponding of over-sampling data centralization;
(e) determine multiple SDV differences, each SDV difference is determined by the difference between standard deviation value and the MAX-MIN difference of definite corresponding over-sampling data set;
(f) one or more described SDV difference and the threshold value comparison that will in step (e), determine, with the possibility that determines whether that described dynamic measuring data comprises cause and effect data or Gauss's normal state data, the difference that is wherein greater than described threshold value characterizes the possibility of cause and effect data.
In the 8th embodiment, comprise following steps for the treatment of the Computerized method of over-sampling dynamic measuring data:
(a) determine multiple maximal values, each maximal value obtains from corresponding one of described over-sampling data centralization;
(b) 3 or more value before the maximal value of definite immediately one or more over-sampling data sets;
(c) 3 or more value after the maximal value of definite immediately one or more over-sampling data sets;
(d), for one or more over-sampling data sets, based on described maximal value, immediately 3 before described maximal value or more value and 3 or more value after described maximal value immediately, determine peak value form factor characteristic; And
(e) the peak value form factor characteristic based on definite in step (d), determines the possible causal event relevant to described maximal value.
In the 9th embodiment, comprise following steps for the treatment of the Computerized method of over-sampling dynamic measuring data:
(a) determine multiple maximal values, each maximal value obtains from corresponding one of described over-sampling data centralization;
(b) determine multiple standard deviation values, each standard deviation value obtains from corresponding one of described over-sampling data centralization;
(c) determine multiple parameter cause and effect ratio characteristics, each parameter cause and effect ratio characteristic obtains from corresponding one of described over-sampling data centralization;
(d) determine multiple peak value form factor characteristics, each peak value form factor characteristic obtains from corresponding one of described over-sampling data centralization; And
(e) based on described maximal value, standard deviation value, parameter cause and effect ratio characteristic and peak value form factor characteristic, determine whether to exist one or more following situations:
The cracked situation existing due to the fatigue of roller bearing parts, and the broken teeth situation existing due to the fatigue failure of gear assembly.
In the tenth embodiment, comprise following steps for the treatment of the Computerized method of over-sampling dynamic measuring data:
(a) determine multiple minimum value, each minimum value obtains from corresponding one of described over-sampling data centralization;
(b) determine multiple intermediate values, each intermediate value obtains from corresponding one of described over-sampling data centralization;
(c) determine multiple mode value, each mode value obtains from corresponding one of described over-sampling data centralization;
(d) determine multiple mean value, each mean value obtains from corresponding one of described over-sampling data centralization;
(e) determine multiple standard deviation values, each standard deviation value obtains from corresponding one of described over-sampling data centralization;
(f) determine multiple peak value form factor characteristics, each peak value form factor characteristic obtains from corresponding one of described over-sampling data centralization; And
(g), based on described minimum value, intermediate value, mode value, mean value, standard deviation value and peak value form factor characteristic, determine whether to exist and slide friction condition due to what lack of lubrication caused.
In the 11 embodiment, comprise following steps for the treatment of the described Computerized method of the dynamic measuring data of over-sampling:
(a) determine multiple intermediate values, each intermediate value obtains from corresponding one of described over-sampling data centralization;
(b) determine multiple mode value, each mode value obtains from corresponding one of described over-sampling data centralization;
(c) determine multiple mean value, each mean value obtains from corresponding one of described over-sampling data centralization;
(d) determine multiple standard deviation values, each standard deviation value obtains from corresponding one of described over-sampling data centralization;
(e) determine multiple peak value form factor characteristics, each peak value form factor characteristic obtains from corresponding one of described over-sampling data centralization; And
(f), based on described intermediate value, mode value, mean value, standard deviation value and peak value form factor characteristic, determine whether to exist due to the suitable lubricated even running situation causing.
In the 12 embodiment, comprise following steps for the treatment of the Computerized method of over-sampling dynamic measuring data:
(a) determine multiple intermediate values, each intermediate value obtains from corresponding one of described over-sampling data centralization;
(b) determine multiple mean value, each mean value obtains from corresponding one of described over-sampling data centralization;
(c) determine multiple standard deviation values, each standard deviation value obtains from corresponding one of described over-sampling data centralization;
D) determine multiple peak value form factor characteristics, each peak value form factor characteristic obtains from corresponding one of described over-sampling data centralization; And
(e) based on described intermediate value, mean value, standard deviation value and peak value form factor characteristic, determine whether to exist misaligned situations.
In the 13 embodiment, comprise following steps for the treatment of the Computerized method of the dynamic measuring data of over-sampling:
(a) determine multiple maximal values, each maximal value obtains from corresponding one of the data centralization of described over-sampling;
(b) determine multiple standard deviation values, each standard deviation value obtains from corresponding one of the data centralization of described over-sampling;
(c) determine multiple parameter cause and effect ratio characteristics, each parameter cause and effect ratio characteristic obtains from corresponding one of the data centralization of described over-sampling;
(d) determine multiple peak value form factor characteristics, each peak value form factor characteristic obtains from corresponding one of the data centralization of described over-sampling; And
(e), based on described maximal value, standard deviation value, parameter cause and effect ratio characteristic and peak value form factor characteristic, determine in pipeline placement process whether because sleeve pipe resonance exists time top layer fatigue crack.
In the 14 embodiment, comprise following steps for the treatment of the Computerized method of the dynamic measuring data of over-sampling:
(a) determine multiple maximal values, each maximal value obtains from corresponding one of the data centralization of described over-sampling;
(b) determine multiple mean value, each mean value obtains from corresponding one of the data centralization of described over-sampling;
(c) determine multiple standard deviation values, each standard deviation value obtains from corresponding one of the data centralization of described over-sampling;
(d) determine multiple peak value form factor characteristics, each peak value form factor characteristic obtains from corresponding one of the data centralization of described over-sampling; And
(e), based on described maximal value, mean value, standard deviation value and peak value form factor characteristic, determine whether due to the excessive slip-stick that occurs of static friction coefficient loading on interface.
In the 15 embodiment, comprise following steps for the treatment of the Computerized method of the dynamic measuring data of over-sampling:
(a) determine at least one in the following:
Multiple minimum value, each minimum value obtains from corresponding one of the data centralization of described over-sampling;
Multiple intermediate values, each intermediate value obtains from corresponding one of the data centralization of described over-sampling;
Multiple mode value, each mode value obtains from corresponding one of the data centralization of described over-sampling;
Multiple standard deviation values, each standard deviation value obtains from corresponding one of the data centralization of described over-sampling; And
Multiple peak value form factor characteristics, each peak value form factor characteristic obtains from corresponding one of the data centralization of described over-sampling; And
(b) at least in part based on definite value in step (a), determine whether to exist one or more following situations:
Occur near the shelf depreciation of high voltage electric equipment; And
Occur near the leakage situation of the generation fluid turbulent of pressurized leak.
In the 16 embodiment, comprise following steps for the treatment of the Computerized method of the dynamic measuring data of over-sampling:
(a) determine at least one in the following:
Multiple intermediate values, each intermediate value obtains from corresponding one of the data centralization of described over-sampling;
Multiple mode value, each mode value obtains from corresponding one of the data centralization of described over-sampling;
Multiple standard deviation values, each standard deviation value obtains from corresponding one of the data centralization of described over-sampling; And
Multiple peak value form factor characteristics, each peak value form factor characteristic obtains from corresponding one of the data centralization of described over-sampling; And
(b), based on the definite value of step (a), determine whether three-phase power line exists intermittent fault situation.
In the 17 embodiment, comprise following steps for the treatment of the Computerized method of the dynamic measuring data of over-sampling:
(a), for the data set of one or more over-samplings, generate the scalar value of the dynamic measuring data in multiple expression data sets;
(b), based on described multiple scalar value, determine the characteristic value of feature, quality or the characteristic of the dynamic measuring data in one or more designation data collection; And
(c) preserve multiple scalar value and the one or more characteristic value relevant to the identifier of data set in computer memory.
In the 18 embodiment, comprise following steps for the treatment of the described Computerized method of the dynamic measuring data of over-sampling:
(a) determine multiple the first statistics scalar value, each the first statistics scalar value obtains from corresponding one of the data centralization of multiple over-samplings;
(b) based on one or more described the first statistics scalar value, determine that machine or treatment progress are in the first state instead of in the second state;
(c) determine multiple the second statistics scalar value, each the second statistics scalar value obtains from corresponding one of the data centralization of multiple over-samplings; And
(d) based on one or more described the second statistics scalar value, determine that machine or treatment progress are in the second state instead of in the first state.
An embodiment provides a kind of Computerized method, for alleviating the aliasing effect of over-sampling dynamic measuring data frequency transformation, the data set that described over-sampling dynamic measuring data comprises the multiple over-samplings that collect by one or more sensors that append in machine or treatment progress, wherein the data set of each over-sampling is corresponding to specific sampling interval.The method preferably comprises following steps:
(a), for the data set of one or more over-samplings, according to from minimum to senior general's data value sequence, distribute with the cumulative data forming after sequence; And
(b) for the data set of one or more over-samplings, determine intermediate value be the absolute intermediate value that distributes of cumulative data after two or more immediately sort on or under the mean value of data value.
An embodiment provides a kind of Computerized method of avoiding aliasing effect in the time processing the dynamic measuring data of the over-sampling collecting by one or more sensors that append to machine or treatment progress.In this this method, at sampling rate F sabove the dynamic measuring data of over-sampling is sampled.The execution that relates to the Nonlinear Processing of the dynamic measuring data that extracts over-sampling will cause aliasing effect.The method preferably comprises following steps:
(a) by inserting N-1 0 between data sample adjacent in the dynamic measuring data of over-sampling, on integer up-sampling rate N, the dynamic measuring data of described over-sampling is carried out to up-sampling, thereby generate up-sampling data;
(b) by using cutoff frequency F s/ 2 low-pass filter carries out low-pass filtering to described up-sampling data, removes any spectral image producing in step (a), and wherein L is more than or equal to 1 integer, thereby generates not containing F sup-sampling data after the low-pass filtering of more than/2 spectral images.
(c) if N<L and L>1, show to exist part resampling rate, by retaining each L sample and abandoning L-1 sample between each L sample, up-sampling data after described low-pass filtering are carried out to down-sampling, do not contain up-sampling frequency F thereby generate sdown-sampled data after the low-pass filtering of × spectral image on (N/L);
(d) the up-sampling data after described low-pass filtering are carried out to the Nonlinear Processing extracting, there is thereby generate the F of being aliasing in sthe data of more than/2 distortion components;
(e) be F by using cutoff frequency s/ 2 low-pass filter, carries out filtering to the data that generate in step (d), gets rid of F thereby generate sthe data of more than/2 alias component; And
(f) by retaining each N sample and abandoning N-1 sample between each N sample, the data that generate in step (e) are carried out to down-sampling, thereby generated the non-linear aftertreatment data that wherein aliasing effect is alleviated.
An embodiment provides the sample frequency to fix, and within the time cycle of one section of expansion, gathers the Computerized method of the dynamic measuring data of over-sampling.The data set that described over-sampling dynamic measuring data comprises the multiple over-samplings that collect by one or more sensors that append to machine or process, wherein the data set of each over-sampling is corresponding to specific sampling interval.This method preferably comprises following steps:
(a), during the period 1 within the expansion time cycle, use the first sampling interval 1/F sR1gather described dynamic measuring data, the sample of the first number of the each over-sampling data centralization collecting in being created in during the period 1; And
(b), during the second round within the expansion time cycle, use than the first sampling interval 1/ fSR1the second sampling interval 1/ that duration is long fSR2gather described dynamic measuring data, be created in the sample of the second number of the each over-sampling data centralization collecting during second round; Wherein the second number of sample is more than described the first number.
Embodiment provides acquisition and processing by the Computerized method of multiple vibration datas that append to the over-sampling collecting for the sensor of the physical construction of material processed.Described physical construction operationally sends to vibration transducer by the energy of vibration from material.The vibration data of described over-sampling comprises multiple over-sampling data sets, and wherein the data set of each over-sampling is corresponding to specific sampling interval.This method preferably comprises following steps:
(a) receive the vibrational energy of the first vibration transducer of described multiple vibration transducers, wherein said vibrational energy generates by event, and described event occurs in described material when just processed and within being sent to the first vibration transducer by described physical construction;
(b), based on described vibrational energy, described the first vibration transducer generates the first vibration signal;
(c) the first over-sampling vibration data of the data set that described in over-sampling, the first vibration signal comprises multiple the first over-samplings with generation;
(d) for the data centralization of multiple the first over-samplings each, determine one or more from by maximal value, minimum value, mean value, intermediate value, mode value, standard deviation value, be up to the first scalar value of selecting the group that minimum zone value, kurtosis value, measure of skewness value and wavelength value form;
(e), based on one or more described the first scalar value, determine one or more the first characteristic values that provide event type to indicate;
(f) generate very first time stamp value, the vibrational energy that event generates described in described very first time stamp value representation is in the first received time of vibration transducer place;
(g) receive the vibrational energy of the second vibration transducer in multiple described vibration transducers, wherein said vibrational energy is sent to the second vibration transducer by described physical construction;
(h), based on described vibrational energy, described the second vibration transducer generates the second vibration signal;
(i) the second vibration signal described in over-sampling, to generate the vibration data of the second over-sampling that comprises multiple the second over-sampling data sets;
(j) for the data centralization of multiple the second over-samplings each, determine one or more from by maximal value, minimum value, mean value, intermediate value, mode value, standard deviation value, be up to the second scalar value of selecting the group that minimum zone value, kurtosis value, measure of skewness value and wavelength value form;
(k), based on one or more described the second scalar value, determine one or more the second characteristic values that provide event type to indicate;
(l) generate the second timestamp value, the vibrational energy that described the second timestamp value representation generates by described event is in described the second received time of vibration transducer place; And
(m) by by one or more the First Eigenvalues and one or more Second Eigenvalue comparison, identical with the event type characterizing by one or more Second Eigenvalues to determine the event type characterizing by one or more the First Eigenvalues.
An embodiment provides a kind of Computerized method of acquisition and processing mechanical vibration data, is used in mechanical protection system to realize the object of automatic triggering machine shutdown.Described mechanical vibration data are gathered by one or more vibration transducers that append to machine.The method preferably comprises following steps:
(a) with mechanical vibration data described in the sampling rate over-sampling far above Nyquist frequency;
(b) numerical data of processing over-sampling;
(c) generate the digital data sets of a series of over-samplings;
(d) for the digital data sets of one or more described over-samplings, generate the scalar value or the attribute that represent the mechanical vibration data in data set, described data set is the group based on selectivity decimation value at least partly, and described selectivity decimation value comprises intermediate value, maximal value, minimum value, standard deviation value and peak value form factor value; And
(e) scalar value and the attribute based on definite in step (d) at least in part, obtains for gathering the machine of described mechanical vibration data or the characteristic of method.
An embodiment provides a kind of by using one or more current sensors, the Computerized method of acquisition and processing electric electromechanics flow data.The method preferably comprises following steps:
(a) by using one or more current sensors, measure analog motor current signal message;
(b), to be at least the sampling rate of 10 times of line frequency, simulating motor current signal information is converted to the dynamo-electric flow data of digitalized electric of over-sampling;
(c) generate the over-sampling data set of the dynamo-electric flow datas of a series of digitalized electrics from described over-sampling, the data set of each over-sampling is corresponding to a sampling interval;
(d) data set that extracts described over-sampling is to obtain the scalar value after extracting;
(e) optionally extract the data set of described over-sampling, with the data set characteristic based on selecting, obtain corresponding attribute from the group being formed by intermediate value, kurtosis value, maximal value, minimum value, standard deviation and peak value form factor; And
(f) the scalar value after extraction definite in step (d), in step (e), definite attribute and the characteristic of electric electromechanics flow data connect.
An embodiment relates to the device of the vibration data of a kind of acquisition and processing machine or treatment progress.Described device comprises the vibration transducer that appends to machine, and it is F that described machine produces a maximum frequency mAXanalog vibration signal, wherein F mAXbe greater than the temporal frequency that occurs in the event in machine or process.Described device is also included in and is at least F mAXthe analog to digital converter of over-sampling analog vibration signal in the sample frequency of 7 times, to generate multiple sampled data set, each data set is corresponding to specific sampling interval.Described device also comprises the abstraction module that contains multiple parallel field programmable gate array (FPGAs).The one FPGA receives over-sampling data set and determines the first scalar value from each over-sampling data centralization.Described the first scalar can be maximal value, minimum value, intermediate value, mode value, mean value, standard deviation value, parameter cause and effect ratio, ruuning situation value or peak value form factor value.The 2nd FPGA receives over-sampling data set and determines second scalar value different from the first scalar value from each over-sampling data centralization.Described the second scalar can be maximal value, minimum value, intermediate value, mode value mean value, standard deviation value, parameter cause and effect ratio, ruuning situation value or peak value form factor value.
Brief description of the drawings
With reference to detailed description with the accompanying drawing, further advantage of the present invention becomes apparent, and wherein in order to clearly show details, element is not proportionally drawn, and the same reference numbers that wherein runs through multiple accompanying drawings characterizes identical element, wherein:
Fig. 1 show according to the embodiment of the present invention for gathering and the device of analytic engine data;
Fig. 2 show according to the embodiment of the present invention for gathering and the method for analytic engine data;
Fig. 3 A shows two tail normal state or the parameter distribution of data set;
Fig. 3 B shows the zero-based skewness cumulative distribution of data set;
Fig. 3 C shows the skewness cumulative distribution starting from high-value of data set;
Fig. 3 D shows has the skewness cumulative distribution that the data set of discrete data starts from high-value;
Fig. 4 shows the curve map of time synchronized trend data;
Fig. 5 shows the example of the vibration data display window being generated by the embodiment of the present invention;
Fig. 6 shows the MAX time domain waveform and the frequency spectrum data that are generated by the embodiment of the present invention;
Fig. 7 shows the MED time domain waveform and the frequency spectrum data that are generated by the embodiment of the present invention;
Fig. 8 shows the MODE time domain waveform and the frequency spectrum data that are generated by the embodiment of the present invention;
Fig. 9 shows the MIN time domain waveform and the frequency spectrum data that are generated by the embodiment of the present invention;
Figure 10 shows the AVE time domain waveform and the frequency spectrum data that are generated by the embodiment of the present invention;
Figure 11 shows the SDV time domain waveform and the frequency spectrum data that are generated by the embodiment of the present invention;
Figure 12 shows the measure of skewness data that generated by the embodiment of the present invention;
Figure 13 shows the kurtosis data that generated by the embodiment of the present invention;
Figure 14 shows the related coefficient waveform being generated by the embodiment of the present invention;
Figure 15 shows the parallel processing plan of implementing in extraction processor according to the embodiment of the present invention;
Figure 16 shows the example of the vibration data display window being generated by the embodiment of the present invention;
Figure 17 shows the demonstration of the single sample waveform being generated by the embodiment of the present invention and the corresponding frequency spectrum data obtaining on progressive position along sampling interval;
Figure 18 shows and extracts the example of the spectrum mode that can be used for fault diagnosis according to the embodiment of the present invention by selectivity;
Figure 19 and 20 shows the example that is positioned at lip-deep sensor array;
Figure 21 shows the imaging transmitter according to the embodiment of the present invention;
Figure 22 shows according to the overlapping region of multiple imaging transmitters visual field of the embodiment of the present invention;
Figure 23 shows according to the machinery in the imaging transmitter field of view of the embodiment of the present invention;
Figure 24 shows according to the electric component in the imaging transmitter field of view of the embodiment of the present invention;
Figure 25 shows according to pipe arrangement and valve in the imaging transmitter field of view of the embodiment of the present invention;
Figure 26 shows according to electric power transfer and distribution member in the imaging transmitter field of view of the embodiment of the present invention;
Figure 27 shows according to the reference point in the imaging transmitter field of view of the embodiment of the present invention.
Embodiment
Below some abbreviations that use in the description of the embodiment of the present invention:
The measured value of AVE-average type, for example mean value;
The absolute peak-peak of MAX-, or represent the mean value of peaked two or three actual maximal values or bare maximum;
The series of characteristics of the difference between MAX-MIN – ordinary representation maximal value and minimum value;
MED-intermediate value is also the 50th percentile in cumulative distribution;
MIN-bare minimum, or represent two or three actual minimum of minimum value or the mean value of bare minimum;
The measured value of MODE-mode value type, as modal value in sampling interval entire scope, or the close limit of modal value;
The ruuning situation of OPC-sensor or circuit;
Over-sampling data (Oversampled data)-be greater than the maximum frequency (F of data acquisition mAX) carry out the dynamic digital data that over-sampling obtains in the sampling rate of manyfold;
PeakVue tM-a kind of during sampling interval the characteristic relevant to selected peak value retention value typically, and typically comprise that described peak value keeps being used for determining from tested value in full-wave rectification (rectification) step before step or sampling interval other technology of peak value;
Hundredths (Percentile)-position relevant with cumulative distribution or probability density distribution, value is wherein arranged from minimum (0%) to the highest (100%), and the 1st, 3rd, the 5th, the 10th, the 50th (for example MED), the 90th, 95th, the value of the 97th, 99th hundredths may extract and have great importance selectivity;
PSF-peak shape is because of sub-feature;
PVC-parameter cause and effect ratio characteristic, wherein parameter is subject to showing the control that good two tail distribution statisticses are learned, and cause and effect " reason " do not expected, and common single tail distributes, also the statistics control of normal and measure of skewness or kurtosis related characteristics;
RMS-root mean square characteristic;
The measurement of SDV-statistical straggling type, for example standard deviation;
Sk-measure of skewness characteristic;
Smax-is as the maximum stroke by the defined dynamics track of ISO7919;
SopMax-is as by the defined maximum vibration displacement of ISO7919;
SppMax-is as by the defined maximum vibration shift value of ISO7919;
Normal and the parameter of SPC-, the statistical Process Control that Gauss normal distribution is relevant;
Enforcement of the present invention has promoted the development of prior art, before abandoning over-sampling data, again obtains the useful information of observational measurement, sensor and circuit by analyzing over-sampling data.The important fresh information that embodiments of the invention provide is the root data different from the vibration data of normal state.Root or " cause and effect " data set are different from normal state or Gaussian data collection statistically, and for example it uses difference between mean value and intermediate value relatively or the form factor analysis of over-sampling data set.One of cause and effect data is exemplified as by using PeakVue tMdetect and impact the high frequency stress wave producing, this is because event instantaneous generation in sampling interval conventionally.For example being vibrated by the uneven normal state producing of machine, is more likely Gaussian data, because it scatters in each axle rotation in multiple sampling interval.On the other hand, the frictional vibration causing due to lubrication circumstances deficiency may record the high PeakVue that reads when finishing that starts from sampling interval tM.
The remarkable advantage that the embodiment of the present invention provides comprises centre (median) value of each over-sampling data set or catching of " middle (middle) " value.During being illustrated in sampling interval, there is normal state vibration in intermediate value.Compare easily intermediate value and mean value, to determine cause and effect-Gaussian data.In addition, process over-sampling data set simultaneously and be reasonably, the maximal value producing intermediate value after rectification, after correcting and the scalar value of not correcting of being understood, abandoning or retain in order further to show, to analyze and understanding.
In each embodiment, by for example carried out the Bandwidth Compression Technique that then low-pass filtering extracts before frequency inverted, take measures to alleviate the aliasing effect of observing in frequency conversion process.At selectivity extraction technique, for example intermediate value of over-sampling data has in the situation of the potentiality that find aliasing information, increases in order to suppress aliasing, can comprise equalization step.For example, for intermediate value, not from ordering cumulative distribution, to take out single intermediate value, but 3 values of the centre that is positioned at ordering cumulative distribution are averaging together.
Selectivity extraction process based on statistical indicator is the non-linear process that a meeting causes distortion.When described distortion comprises higher than nyquist frequency (F s/ 2), when frequency component, these component aliasings turn back to 0 to F son the frequency spectrum of/2Hz, thereby pollute the purity of frequency spectrum.The frequency that important aliasing returns is called as " folding frequency ".Keep 0 to F sa kind of novel method of the purity of frequency spectrum of/2Hz is before non-linear process, to carry out pre-treatment step to extend described folding frequency, and turns back to from 0 to F at aliasing sbefore the scope of/2HZ, for nonlinear distortion component is set up the spectrum space that more can insert.Similarly technology is used in field of audio processing the nonlinear distortion that digitizing simulation produces by high-gain guitar amplifier, and as described among U.S. No.5789689 (' 689 patent), its full content is by reference to be incorporated to this paper at this.The pre-service of ' 689 patent is defined as " super sampling (ubersampling) " technology, in " super sampling " technology, the data of catching are carried out resampling with higher speed, described speed is defined by rational ratio of integers (N/L), wherein N is the up-sampling rate of integer, and L is the down-sampling rate of integer.In order to carry out super sampling, introduce factor N, first described data are carried out to up-sampling by insert N-1 individual 0 between the sample of every pair of input.If definition down-sampling speed L is greater than 1, utilizing cutoff frequency is (F s/ 2) * N) wave filter of/L carries out low-pass filtering to described up-sampling data, and retain each L sampling and abandon L-1 sampling between each L sampling.In frequency domain, the effect of this operation is to be created on the frequency spectrum with initialize signal in initial bandwidth with same frequency spectrum component, but in multiple initial sampling rate, has the image of initial spectrum.Do not damaged by these images in order to ensure initialize signal, utilize at F sthe low-pass filter of/2 cut-offs is removed and is all greater than F s/ 2 frequency.Filtered result frequency spectrum with from 0 to F sthe initial spectrum of/2Hz is identical, but does not comprise from F at present s/ 2 to F sthe spectrum component (dead band) of/2* (N/L) scope.This dead band is served as and will be extracted the effect of frequency spectrum storage box of generated described high frequency nonlinear component by selectivity.New collapsible frequency is than the high N/L of initial spectrum at present.The order of severity producing according to distortion, can improve super sampling speed, turns back to from 0 to F guaranteeing at aliasing sbefore/2 scope, there are enough spectrum space to hold the distortion component that may reduce the purity of frequency spectrum.After completing Nonlinear Processing, at F son/2, use low-pass filter to carry out filtering to super sampling data, to carry out a contrary process, described super sampling data-switching is turned back to initial bandwidth, then carry out down-sampling with the ratio reciprocal that equals described super sampling speed.Shown in an example, suppose the ratio of N=2 and L=1, it produces effective super sampling speed is 2.If described sampling rate is 50kHz, described initial folding frequency is F s/ 2=25kHz.Through-rate is 2 and after the up-sampling of the low-pass filtering of 25KHZ, result obtains identical with initial frequency spectrum, but has from the frequency spectrum of the additional frequency spectrum space of 25kHz to 50kHz extension and the new folding frequency of 50kHz.If described Nonlinear Processing generates the component that is greater than 25kHz initial fold frequency, this can be regarded as the alias component in initial spectrum.But, using the described super sampling method that super sampling ratio is 2, the highest frequency component (aliasing limit (AL)) before aliasing occurs is 75kHz at present.Aliasing limit can be passed through AL=F s((N/L) – 0.5) calculates.
Embodiments of the invention provide programmable logic device, for by using as the dynamic signal analysis device of vibration analyzer, and the numerical data of the over-sampling that previously abandoned of collection, then analyze described data and obtain useful, significant information.The present embodiment is applicable to multiple dynamic signal analysis devices; described dynamic signal analysis device includes but not limited to the flux distribution device of vibration analyzer, motor current signal analyzer and motor, and it can be hand-heldly establishing, realize in in-service monitoring device, protection system, transmitter and the system that is associated with wherein one or more.
Preferred embodiment is extracted situation information, transducer or survey sensor performance information and electronics or the circuit performance information of machine or equipment in each Dynamic Signal sampling interval from over-sampling numerical data.Over-sampling interval can be expressed as 1/F sR, dynamic sampling interval can be expressed as 1/F mAX, the number of the data point in dynamic sampling interval can be expressed as F sR/ F mAX, wherein F sRthe sampling rate of over-sampling, F mAXdynamic sampling speed.For example, work as F sR=100kHz and F mAXwhen=2kHz, in every dynamic sampling interval, there are 50 over-sampling data points.
Except peak value and decimation value, the preferred embodiments of the present invention are used that extract and other over-sampling value relevant with the characteristic of equipment, sensor and circuit, so that more useful information to be provided before obtaining information in sampling interval from over-sampling data.
Some embodiments of the present invention have been improved trend data analysis.A kind of method of improving trend analysis is in time waveform, utilizes Extracting Information optionally to imagine and analyzes the feature of selectivity extraction.Consequent information can utilize programmable logic device or mankind's logical thinking or both to understand, to seek and to identify for example pattern of fault mode trend.This sometimes contribute to find may with machine performance, procedure parameter and other vibration intercorrelation or have a related pattern.Fig. 4 has shown the example of a trend map.Although this is to offer the graph curve that the mankind understand, the trend data of time synchronized and the analysis of relevant information and deciphering can realize with programmable logic device.
During the selectivity that is sometimes used in cumulative distribution or probability density distribution extracts.All scalar value that produce during sampling interval are sorted to the highest from minimum, effectively to represent cumulative distribution or the probability density distribution of sampling interval data set.In certain embodiments, one or more attributes of a relation can be assigned to the position of order that the tested value that for example reflects in sampling interval is associated or the scalar value of the sequence number of sequential.Each trifle has below been discussed the measured value of may the each sampling interval in over-sampling waveform carrying out.In the example of many these measured values is also included within.
Fig. 1 shows a preferred embodiment of the multiple channel mechanical vibration measurement device 80 of acquisition and processing over-sampling numerical data.In the present embodiment, AFE (analog front end) comprises 8 input pickup 82a-82d and 84a-84d.Although the invention is not restricted to the sensor of any particular type, described sensor 82a-82d is that preferred accelerometer and described sensor 84a-84d are preferred voltage sensors.After each sensor 82a-82d and 84a-84d, analog signal link comprises differential amplifier 85a-85h, 3 frequency dividing circuit 86a-86h, and differential amplifier is to 88a1-88h1 and 88a2-88h2 and low-pass filter 90a-90h.These 8 analog sensor passages be provided for 24 Sigma-delta ADC's that 8 sampling rates are controlled by fixed clock 94 (ADC ' input end of 92a-92h s).
8-4 cross point switches 96 provides any one in 8 passages of ADC92a-92h output terminal has been switched to any one in 4 digital processing passage 98a-98d of digital signal processor 98, in a preferred embodiment, described digital signal processor 98 is FPGA.In Fig. 1, FPGA passage 98a is described in detail.In this preferred embodiment, the parts of passage 98b, 98c and 98d are identical with the parts of passage 98a.The passage 98a of described FPGA comprises filter module 102, described filter module 102 can comprise Hi-pass filter, low-pass filter or bandpass filter, and the passage 98a of described FPGA also comprises first integrator module 106, second integral device module 110, data block indicator module 114, data manager module 116, data pick-up device module 118 and FIFO120.The output of FIFO120 is provided to processor 100.
In a preferred embodiment, wave filter 102 is Hi-pass filter, and it has removed the DC component in signal at its input end.Switch 104 needs to provide the bypass to Hi-pass filter 102 in measured application in DC offset signal.
First integrator 106 degree of will speed up signal integrations, to convert thereof into rate signal.In a preferred embodiment, first integrator 106 is unlimited input response (IIR) integrators.In alternative embodiment, first integrator 106 can be carried out other angular quadrature scheme that uses other integral algorithm.Switch 108, not needing in the application of First-order Integral, provides the bypass to first integrator 106.
Second integral device 110 is by rate signal integration, to convert thereof into position signalling.In a preferred embodiment, second integral device 110 is structurally with in function to be equal to the IIR integrator of first integrator 106.In alternative embodiment, second integral device 108 can be carried out other angular quadrature scheme that uses other integral algorithm.Switch 112 is not needing, in the application of second-order integration, to provide the bypass to second integral device 110.For example, in the time only needing first integrator 106 to convert speed to from acceleration, can bypass described in second integral device 110.When desired output be acceleration time, can two integrators of bypass 106 and 110.In the time that desired output is displacement, use two integrators 106 and 110.
In certain embodiments, one or two integrator 106 and 110 at least one FPGA passage 98a-98d can carry out dual-integration to vibration signal at its input end.For example, first integrator 106 can receive acceleration signal and carry out dual-integration to provide displacement signal at its output terminal.In the present embodiment, second integral device 110 is by switch 112 bypasses, so that data block indicator module 114 is from first integrator 106 received bit shifting signals.In an alternative embodiment, first integrator 106, by switch 108 bypasses, makes second integral device 110 receive acceleration signal, and makes second integral device 110 carry out dual-integration to provide displacement signal at its output terminal.In another embodiment, at least one in FPGA passage 98a-98d only comprises single integrator, and described integrator receives acceleration signal and carries out dual-integration to provide displacement signal at its output terminal end.
Data block indicator module 114 has been specified the over-sampling data block of single sampling interval, and this can obtain more detailed description hereinafter.
Data manager module 116 is managed sampling interval data set, and this can obtain more detailed description hereinafter.
The characteristic value of data pick-up device module 118 specified data collection, for example MAX,, MED, MIN, AVE, SDV, PvC, OPC and PSF.In certain embodiments, data pick-up device module 118, in single process or by the step of execution sequence and other parallel step, as shown in figure 15, by signal being divided into multiple parallel processing/circuit paths, or by compiling multiple values of for example MAX value and med value, extract over-sampling data.The data pick-up device module 118 of utilizing field programmable gate array (FPGAs) to realize is good at the parallel or sequential processes multiple values from over-sampling extracting data especially.
FIFO120 allows FPGA98 to generate real-time vibration data, allows processor 100 to access asynchronously described data simultaneously.
Processor 100 receives vibration signal data from each 4FPGA passage 98A-98D, and carry out one or more vibration analysis functions, for example time-domain waveform analysis, average analysis, cross aisle analysis, FFT spectrum analysis, phase analysis, autocorrelation analysis and data distributional analysis.Described processor 100 can also be processed user interface and Presentation Function.In alternative embodiment, the some or all of functions of being carried out by processor 100 can be carried out by FPGA98.
In a preferred embodiment of system shown in Figure 1, ADC92a-92h is unusual 24 sigma-delta-converters of high-quality.The ADC of latest generation has the dynamic range that is greater than 120dB and the signal to noise ratio (S/N ratio) that is greater than 110dB.By large dynamic range, whole input voltage range can obtain sufficiently high resolution, to eliminate the needs to gain amplifier and AC coupling amplifier.The great dynamic range of ADC92a-92h has been resolved the little AC signal being superimposed upon in large DC skew, and sensor output signal can be directly coupled to ADC, and DC component can be removed by the real-time digital filtering in FPGA98.
Fig. 2 shows mechanical vibration measurement mechanism as shown in Figure 1 of utilization and gathers and analyze a preferred embodiment of the method 200 of over-sampling mechanical vibration data.First, collection machinery vibration data (step 202) in multiple sampling interval.For each sampling interval, all scalar value are sorted to high-amplitude from lowest amplitude, thereby effectively represent cumulative distribution or the probability density distribution (step 204) of sampling interval data set.In certain embodiments, in sampling interval data set, the scalar value of data can be before sequence, and the full-wave rectification taking absolute value, as conventionally completed by PeakVue and other peak detection technology.In certain embodiments, positive negative value is sorted to the highest from minimum.In certain embodiments, one or more attributes of a relation can be assigned to different scalar value, for example, reflect the position of order that the tested value in sampling interval is associated or the scalar value of the sequence number of sequential.For each sampling interval, determine, store and process the value of several scalars, just as described in detail below.
Fig. 5 shows an example of the display window of the vibration data being produced by embodiments of the invention.In Fig. 5, waveform untreated, over-sampling is presented at the top of window.In its lower section, show along the frequency spectrum corresponding with it by the extraction waveform of getting the mean value of these values and producing in each time interval.The selector switch of the average waveform extracting and frequency spectrum below allows user in extraction process, to select the various measured values that used.After making a choice, show the extraction waveform of selected measurement in each time interval and corresponding frequency spectrum.
Determine that peak value keeps the measured value of type, for example maximal value (MAX), it represents an absolute maximum peak amplitude or the mean value (step 206 in Fig. 2) in 2 or 3 absolute peak-peaks of sampling interval data centralization.MAX value can be for the further PeakVue of waveform, frequency spectrum or other conversion tMprocess.The MAX time domain waveform being generated by one embodiment of the present of invention and the example of frequency spectrum data are presented in Fig. 6.
MAX value can be for the further PeakVue of waveform, frequency spectrum or other conversion tMprocess.The peak value of the over-sampling data that receive keeps measuring very high frequency(VHF) sampling rate data typically, for example, and >>20kHz.Many times, these high frequency measurement have reflected the characteristic of stress wave information, for example produced by the impact of roller bearing defect under the load of roller-seat ring or by the compression or the shearing wave information that produce in the impact of engagement load lower tooth defect.On the other hand, under the relatively low situation of sampling rate, for example, in the frequency range of the mechanical resonant that can find just measured structure, peak value keeps or traditional PeakVue tMmeasured value also can disclose the many information about the mechanical motion of described structure, and is not only stress wave information.Programmable logic device or the mankind's deciphering can be for understanding these differences and inferring the information of deducing.Embodiments of the invention are the application in conjunction with MAX of kurtosis momentum, strengthen or have refuted the cause and effect aspect that shows basic reason.This is one of many logic device examples of the present invention, it can be taught operator or be realized its application by employing programmable logic device, and be shown to operator, or for triggering automatic function or the impact of for example warning indicator or controlling the escape function that machine moves.
Determine the measured value of intermediate value type, for example intermediate value (MED), it represents absolute single intermediate value or average in 2 or 3 absolute intermediate values of described sampling interval data centralization.Fig. 7 shows the MED time domain waveform that produced by embodiments of the invention and an example of frequency spectrum data shows.Described med value can be used for normal state vibration processing, is similar to the RMS value in waveform/frequency spectrum or other conversion.The processing of med value is the important advantage of some embodiments of the present invention.Although RMS and mean value can be used for supplemental characteristic, described data are not parameterized under situation monitors sometimes.On the contrary, it can be the skewed distribution with embedded root, relates to herein and is called " cause and effect data set ".When root forced measurement to is extreme or other--when typical high value, but be sometimes Di Zhi – its described average or mean value are had to important impact.But this event has less impact or not impact to intermediate value.Impact that the distribution of described intermediate value is caused by the Outliers value of the end in arbitrary distribution is very little even not to be affected.Embodiments of the invention utilization is selected from implication and the stability of the intermediate value of the data set of the over-sampling with cause and effect data, so that even do not affect the relatively impact of intermediate value and mean value is very little in the change of an end of distribution or the extremum of other end.
The measured value of deterministic model type, for example mode value (MODE), it is illustrated in the value or the value scope that the most frequently repeat of sampling interval data centralization.The example that Fig. 8 shows the MODE time domain waveform that produced by embodiments of the invention and frequency spectrum data shows (because the total generation time spacing value of some embodiment of described MODE algorithm can not obtain frequency spectrum data in some cases).Because described MODE value is the measurement of frequent cycle, the sign of that it has been or bad measurement, it can be used as detecting or confirming the ruuning situation of sensor.For example, in the time that described MODE value relatively approaches MIN value or MAX value, this has indicated fault or track outer sensor.In the time that MODE value approximates med value, this has indicated stable measurement situation.Good measurement typically has the MODE value that approaches med value or AVE value, according to its distribute be whether normal state or cause and effect.It is the definite trigger switch of root of possible influence factor that cause and effect distributes.For example, the minimum sensor of 0-approach sensor or the fault electronic circuit that low extreme MODE value can be indicated defective sensor, be selected sensor, inappropriate installation improperly.Those skilled in the art can be logically from understand MODE, MED, AVE and MIN value relatively obtain many cause-effect relationshiies.Various embodiment of the present invention adopts these experiences or theoretical deciphering.
Determine the low value type of measuring, for example minimum value (MIN), it represents a definitely minimum tested value or at the definitely mean value of minimum measured values of 2 of sampling interval data centralization or 3.Fig. 9 shows the MIN time domain waveform that produced by embodiments of the invention and an example of frequency spectrum data shows.MIN value can be as the adaptability of verificating sensor and circuit operation.The noise floor that MIN value also can be measured as the assessment limitation of signal to noise ratio.For example when mode value approaches relevant MIN value during away from med value, described MIN value is also the mark that has the sensor of latent defect.
Determine the measurement of average type, for example mean value (AVE), it represents the average of sampling interval data centralization data.Described AVE value can be used for normal state vibration processing, is similar to waveform, the RMS value in frequency spectrum or other conversion.Figure 10 shows the AVE time domain waveform that produced by embodiments of the invention and an example of frequency spectrum data shows.In the time analyzing over-sampling data and follow the over-sampling data of Gauss normal distribution structure, the information that gives to transmit by the measured value of AVE value is more put letter.Difference in a sampling interval between AVE value and MAX value can be the form of the over-sampling data instruction crest factor in over-sampling data set.Between described med value and AVE value, poor (or with similar in essence calculating) of essence disclosed the obvious cause and effect deviation in normal state vibrotechnique.
Determine the measured value of statistical straggling type, for example standard deviation value (SDV), its representative is worth at the Σ of sampling interval data centralization (sigma).Figure 11 shows the SDV time domain waveform that produced by embodiments of the invention and an example of frequency spectrum data shows.Described SDV value can be for being disclosed in the cause and effect data of sampling interval data set inside and other impact of for example friction or physical abrasion or sliding contact.Discuss in conjunction with other value at this, SDV value can be indicated the use of inappropriate sensor, and for example those bandwidth are not enough or other misuse.With the comparison of MAX-MIN scope, SDV value can be used for determining at data set separately or measures the possibility of cause and effect data in colony.In certain embodiments, from one or more sampling interval execute figure Dent (Student) T distribute or Brigit Fischer (Schmidt) (Fisher) distribute data analysis can be used for determine put letter (confidence) compartment analysis and further understand with the probability by tested value is passed in over-sampling data colony the statistical information being associated.
As shown in Figures 12 and 13, SDV value again can be for calculating measure of skewness and kurtosis value.Measure of skewness and kurtosis are three rank and the quadravalence moments of signal.Also can determine the variance of second order and six rank moments.Measure of skewness and kurtosis can be used for disclosing the distributed wave shape that caused by impact, division and the acoustic emission subtle change than static event.
With reference to figure 2, preferred embodiment has determined that parameter cause and effect is than (PVC) information, it can be one or more scalar value, and described scalar value instruction for assessing the biased shape factor or other probability density form factor (step 208) of measured data in sampling interval.PVC can be used for disclosing affects the root evidence of sampling interval data set, for example, impact, division and acoustic emission or other.The distribution of supplemental characteristic also can be called as Gauss normal distribution data.The distribution of root or cause and effect data is often positive or be inclined to negatively distributed data collection.There are several technology to can be used for characterizing deflection or the form factor of probability density.For example, the simple difference between AVE value and med value may be the measurement of measure of skewness.Other example that PVC measure of skewness is calculated comprises D'Agostino-Pearson test and the Pearson's coefficient of measure of skewness, is expressed as SK=3 × (AVE-MED)/SDV.Another kind of well-known technology is kurtosis.Various embodiment can be used for distinguishing positive skewness, negative skewness and other style characteristic.Using kurtosis value, made D'Agostino (D ' Agostino) the K-square test of Special Contributions by Pearson and Ascome & Glynn, is an example of the technology for checking distribution.
Some embodiment have calculated scalar value, and it represents the mathematics comparison of MIN value or SDV value or two or more above-mentioned values, with the ruuning situation of qualitative sensor or circuit (OPC) (step 210).OPC value can be for disclosing sensor or the circuit of possible non-functional or interruption.
In certain embodiments, by for example comparison of absolute value, value of symbol and Δ (delta) value, catch immediately following 3 scalar value before or after MAX value, and calculate the peaked peak shape factor (PSF) (step 212).PSF value can be used for, by empirical test or the physical theory of the possible cause associated with MAX scalar peak value, characterizing relevant inherent characteristic or quality.
Related coefficient is another scalar value, from the autocovariance of data set of over-sampling data set, instruction data set, initialize data collection or generation, obtains.As term as used herein, " instruction data set " is the data set of the result of empirical process, " initialize data collection " is (to be for example stored in system storage, a reference data set that storing state is good) or data set in external data base, " generation data set " is by system geological information and (the tooth engagement of typical fault pattern information, ball rotation, outer ring, the frequency of inner ring) data set that creates.Related coefficient can be calculated according to following formula:
R ( i , j ) = C ( i , j ) C ( i , i ) C ( j , j )
Wherein, C (i, j) is covariance matrix, and i is the vector (each segmentation normalization) of input data segment, and j is the vector of a sinusoidal reference signal.Figure 14 shows the example of related coefficient waveform being produced by embodiments of the invention and shows.
In a preferred embodiment, the scalar value of data can be before sequence in sampling interval data set, and the full-wave rectification (rectification) taking absolute value, as conventionally by PeakVue tMcomplete with other peak detection technology.But, in certain embodiments, by positive negative value according to from minimum to the highest sequence.
People can adopt and produce the desirable characteristics that for example slope or speed change as the mathematical operation of n order derivative, if this information that is the data of being correlated with or pass on for the waveshape from scalar provides better program or the mankind's deciphering.
Use filtering to be absorbed in specific frequency band to eliminate high base band component, data set can be compared with stochastic distribution signal, and adjust the frequency response of the installation site that has different decay and resonance.Before the statistical study step that filtering can be summarized in this manual, during this time, or carry out afterwards.For example, be actually before filtering data stream and calculated the statistical attribute that adopts data stream, after filtering, calculate other statistical attribute.
Preferably, store some or all of scalar value MAX, MED, MIN, AVE, SDV, PvC, OPC and PSF for further processing (step 214).Can comprise wave form analysis, spectrum analysis, cepstral analysis and other transform analysis to the further processing of one or more MAX, MED, MIN, AVE, SDV, PVC, OPC and PSF, as mentioned below.Cepstral analysis is the contrary FFT of useful power spectrum, and the information about rate of change on different spectral band is provided.One or more analytical technologies can be used for A-B-A-B relatively or A-B-C comparison.For example, such analysis can be at " A data set ", and " B data set " and " C data set " is upper to be carried out.This has activated the comparison of different conditions or situation, for example, by good state and state slightly almost or compare for bad state even at all.Carry out that such analysis is quoted or benchmark situation and current operation situation.This relatively can be by understanding by FPGA (Field Programmable Gate Array), or it can be for example by checking that form or graph data PowerPoint be understood intuitively by operator.
time-domain waveform analysis(step 216)-in existing system, use the scalar value or the PeakVue that extract tMselected peak value retention value is carried out time-domain waveform analysis.In the preferred embodiment of the invention, MAX value (represents PeakVue tM), med value (indicate or vibrate without the normal state of cause and effect data), AVE value (representing to suppose to ignore the normal state vibration of cause and effect data), SDV value (monitoring noise and other variance) is all processed in time domain waveform and other analytical technology as herein described.
average analysis(step 218)-average analysis generally relaxes or has removed randomness, thereby makes to repeat to become obvious.The mean value of sampled data comprises the each row in sample is averaging, and wherein sample is the data of many row, and a line is the data value from single sampling interval.In a preferred embodiment, user can select the value of some row in sample, for example 400 lines or 12800 lines or conventionally other numeral between these two, and this depends on the resolution of expectation." sample " is the number of data type measured value, for example, and 400 lines or 12800 lines or selected any row.User selects some mean values conventionally, as 2 or 20 or some other numeral, the sample number being averaged is set.
cross aisle is analyzedthe analysis of (step 220)-cross aisle adopts the synchronous comparison of the common signal from two points mechanically.The reference signal that this analysis is stored by more current sampled signal with at learning phase or during mechanical kilter is carried out.This technology has disclosed the relevant information of phase place, and contributes to distinguish and locate fault.
fast fourier transform (FFT) spectrum analysis(step 222)-traditionally, use the scalar value or the PeakVue that extract tMselected peak value retention value is carried out FFT spectrum analysis.In the preferred embodiment of the invention, MAX value (represents PeakVue tM), med value (indicate or vibrate without the normal state of cause and effect data), AVE value (representing to suppose to ignore the normal state vibration of cause and effect data), SDV value (monitoring noise and other variance) is all processed in FFT spectrum analysis and in other analytical technology as herein described.
phase analysis(step 224) ,-phase analysis adopts velocity gauge information or cross aisle analysis or other technology to identify the relevant pattern of phase place in time domain data.
autocorrelation analysis(step 226)-in a preferred embodiment, autocorrelation analysis comprises: (1) is by described waveform (signal) section of being divided into, (2) on each section, carry out given peak value algorithm, and (3) relatively homogeneity of result between section.This from manage to calculate described average, intermediate value or even STD be separation and be different.If described signal is highly periodic, described section should be closely similar.In fact be not that periodically described section remains similar if described signal is almost constant, but autocorrelation value should be diverse.If event is instantaneous, between section, will there is significant difference so, wherein may in all sections, only have one to there is similar value, a section has the value being different in essence.In order to maximize the number of the value comparing, " overlapping " can be applicable to be similar to FFT and processes.This is consistent with the conforming idea of inspection signal.
data distributional analysis(step 228)-some embodiment have merged the FPGA (Field Programmable Gate Array) of cumulative distribution or probability density distribution or other statistical distribution and have understood, other statistical distribution represents the assimilation data set of sampling interval data set or multiple sampling interval or sampling interval data set array, for example, falling between or exceeding these colonies of the sampling interval data set of these line numbers in 400 lines or 3200 line sampling interval data sets or sampling interval.
Fig. 3 A-3D (obtaining from ASTMD7720) provides the example of cumulative distribution characteristic.Fig. 3 A shows two tail parameter distribution, is called again Gauss normal distribution.The distribution of this type is " statistically showing good ", this means that it defers to statistics and control the lower expectation of processing.In this case, people expect that statistical Process Control (SPC) standard deviation value carrys out the described discrete feature of the concentrated discrete data of data of description.Fig. 3 B shows single tail distribution character of the measured value based on zero.Height mean value or the average than intermediate value or intermediate value typically shown in such distribution.Fig. 3 C shows the skewed distribution based on degree reference, wherein has upper tail or the upper limit, and DATA REASONING looks and is perhaps artificially or is naturally limited herein.For example, this kind of sensor responds the use (being the sensor of bandwidth deficiency) that can indicate inappropriate sensor in application-specific.Fig. 3 D shows and 3 discrete distributions that continuous distribution is different substantially above.Discrete distribution is often considered as integer data, numerical data or some other step function.
Some embodiments of the present invention have the cumulative distribution of at least a portion or the feature of probability density distribution, and the arrangement value of for example sampling interval data set or array distributes.Except MED, MAX and MIN value, other Useful Information can extract from such distribution, for example:
-for example, at the data of low side, 0-1% or 1-5%;
-in high-end data, for example 99-100% or 97-99%;
-at linearity and/or the log slope of low side data profile;
Linearity and/or the log slope of-data profile by quartile;
-linearity and/or the log slope of data profile in the region of described centre;
-at linearity and/or the log slope of high-end data profile; And
-the flex point that distributes at linearity and/or log, the slope of described flex point occurrence positions and described flex point occurrence positions.
process the example of over-sampling data
Table 2 has below been illustrated possible step and the example sequentially in order to obtain physical state relevant information.Be to be understood that to change or alternative order and various step is suitable, and more sequential steps can be skipped and maybe can comprise additional step.For example, if need integration, can perform step C or E or both.Again for example, easily expect simulating signal to be converted to digital signal, the described numerical data of wireless transmission, to another position, is then returned as simulating signal for further processing in the second place by digital data conversion, for example step C.In another example, if need signal rectification, can complete at analog domain or in numeric field.Therefore, the technician in described field can select many variations and rearrange.
Table 2: for obtain the possible step of situation information and the example of order from the dynamic transducer that should contact with mechanical sense.
Steps A is sensor response by physical state transformations.In the healthy application of machinery, the example of physical state for example comprises the situation of breaking that (1) is caused by roller bearing component fatigue, (2) the broken teeth situation being caused by gear fatigue failure, (3) the sliding friction situation being caused by lack of lubrication, (4) by the suitable lubricated even running situation causing, and (5) increase by heat the misaligned situations causing during mechanically actuated.In the healthy application of on-mechanical assets, the example of physical conditions for example comprises the subsurface fatigue that (6) are caused by pipeline middle sleeve resonance, (7) there is slip-stick when the static friction coefficient on loading interface is excessive, (8) near shelf depreciation high voltage electric equipment, (9) near the easy small opening being under pressure, produce the leak condition of fluid turbulent, and the intermittent fault situation of (10) three-phase power line.
In order to monitor physical state, steps A is typically being included in and will in the machinery being monitored or structure, placing sensor, for example sensor of accelerometer, displacement probe, calibrate AE sensor, ultrasonic sensor, current clamp, flux coil, voltage table or other type.
Step B typically comprises sensor response is converted to simulating signal.The example of the simulating signal of some types is listed in table 2.In the embodiment in figure 1, step B carries out by sensor 82a-82d and 84a-84d.
Step C typically comprises by filter preprocessing simulating signal or conditioning signal to delete unnecessary information or better by signal content and noise isolation.In the embodiment in figure 1, step C passes through amplifier 85a-85h, 3 frequency dividing circuit 86a-86h, and amplifier 88a1-88h2 and low-pass filter 90a-90h carry out.
Step D typically comprises signal from analog is converted to numeral.Modern analog to digital converter conventionally utilize ten heavy or more multiple come over-sampling data.64 (64) inferior over-samplings are conventions.When how many over-samplings in the application of understanding embodiments of the invention are sufficient, affiliated technical field personnel have applied statistical theory and practical experience.In certain situation, for example calculate med value, it needs the data of the minority over-sampling in data set.On the other hand, calculate MODE value and need more data.In the embodiment in figure 1, step D carries out by analog to digital converter 92a-92h.
Step e comprises over-sampling digital data conversion is become to processed numerical data.This step can comprise carries out high-pass filtering, low-pass filtering, integration, dual-integration or other digital processing to over-sampling numerical data.In the embodiment in figure 1, step e is by described Hi-pass filter 102, and first integrator 106 and second integral device 108 are carried out.
Step F has been assigned the over-sampling data block of single sampling interval.Typically, piece is the digital amplitude values that some collect with sample frequency during sampling interval.For example,, if use the sampling rate of 200kHz, the F of 2000Hz maxvalue, so described sampling interval is 1/2000 or 0.0005 second, during sampling interval, the number of amplitude measurements is 0.0005x200,000=100.In this example, a sampling interval comprises 100 measurements.In another one example, if sampling rate is 200kHz, F maxvalue is 10Hz, and sampling interval is 1/10 or 0.1 second so, and during sampling interval, the number of amplitude measurements is 0.1 × 200000=20000.In this example, a sampling interval comprises 20000 measured values.In another example, if sampling rate is 200kHz, F maxvalue is 20000Hz, and sampling interval is 1/20000 or 0.00005 second so, and during sampling interval, the number of amplitude measurement is 0.00005 × 200000=10.In this example, a sampling interval comprises 10 measured values.The set that the described interval data collection of step F is (typically being amplitude) numerical value, described value is measured in the sampling interval of each order, and it is the blocks of values collecting with sample frequency during sampling interval.In the embodiment in figure 1, step F is carried out by data block indicator module 114.
What in step G, manage interval data is sampling interval data set.It is the version after the full-wave rectification that the most frequently occurs of the described data that comprise absolute value.Other administration behaviour can comprise sequence, is organized into cumulative distribution or probability density distribution; Cut apart for example quartile or other division; Or for further other processing of wanting management data in sampling interval data set of analysis and/or processing.In the embodiment in figure 1, step G carries out by data manager module 116.
Step H comprises the data set eigenwert of determining for example MAX, MED, MIN, AVE, SDV, PvC, OPC and PSF.In the embodiment in figure 1, step H carries out by extracting processor module 118.According to the example of mechanical health and non-mechanical equipment physical state, the output of following time interval characteristic is related often, significant and enlightenment.FPGA (Field Programmable Gate Array) can be for relatively, understands and may the indicating of the potential situation of deducing, for example:
(1) utilize MAX, SDV, PvC and PSF to detect the situation of breaking being caused by roller bearing component fatigue;
(2) utilize MAX, SDV, PvC and PSF to detect the broken teeth situation being caused by the fatigue failure of gear assembly;
(3) utilize MIN, MED, MODE, AVE, SDV and PSF to detect the sliding friction situation being caused by lack of lubrication;
(4) utilize MED, AVE, MODE, SDV and PSF to detect by the suitable lubricated ruuning situation stably causing;
(5) utilize MED, AVE, SDV and PSF to detect by increasing by heat the misaligned situations causing during operation;
(6) utilize MAX, SDV, PvC and PSF to detect the subsurface fatigue that pipeline middle sleeve resonance causes;
(7) utilize MAX, AVE, SDV and PSF to detect when to load static friction coefficient on interface excessive and slip-stick occurs;
(8) utilize MIN, MED, MODE, AVE, SDV and PSF to detect near the shelf depreciation of high voltage electric equipment;
(9) utilize MIN, MED, AVE, MODE, SDV and PSF to detect near the leakage situation that produces fluid turbulent the easy small opening being under pressure; And
(10) utilize MED, MODE, SDV and PSF to detect the intermittent fault situation of three-phase power line.
In step I, numeric field data is waveform time domain data sequence typically.Easily expect that embodiments of the invention can usage space reference, in this case, described territory will be spatial domain (Δ-time between the measured value of interval delta-distance instead of order).In the embodiment in figure 1, step I carries out by processor 100.
The information of step J is typically by the analysis of the analysis of waveform or the conversion by for example Fast Fourier Transform (FFT) (FFT) or by autocorrelation analysis, cross aisle is analyzed, and other of phase analysis or Wave data or the data that obtain from Wave data obtains in analyzing.In the embodiment in figure 1, step j carries out by processor 100.
use the selectivity at adaptively sampled interval to extract
Some embodiment of the present invention adopts adaptive sampling interval, wherein in the duration of sampling interval, the number of samples collecting in sampling interval, or or even the position of sampling interval in over-sampling waveform and state change and adapt.The transform analysis of noting for example FFT can become useless, unless all data in FFT are used constant sampling interval collection.
Typically sampling rate is fixing frequency sampling speed, and for example 204800 samplings are per second.According to preferred embodiment, such fixed frequency sampling rate can keep constant, and sampling interval can be adjusted the number of sampling in a sampling interval that adapts in fact increase or reduce simultaneously.The longer sampling interval with fixed sample rate has increased and has contributed to the optionally statistic mass of the sampled measurement of extraction step.The self-adaptation with fixed sample rate lengthens the better compression that the result of sampling interval has comprised better statistical confidence and raw data.The shorter sampling interval with fixed sample rate has reduced and has contributed to the optionally statistic mass of the sampled measurement of extraction step.The result with the self-adaptation shortening sampling interval of fixed sample rate has comprised statistical confidence less in sampling interval and the larger bandwidth of metrical information.
An interchangeable embodiment comprises the situation based on occurring in signal, the use of different sampling interval in identical data set.For example, extraction can be by during a part or many parts for waveform, uses relatively longer sampling interval and occurs, and meanwhile, described extraction can be during another part of waveform or many parts, in relatively short sampling interval and occur.
Some unpredictable consequence that uses the experiment of suitable sampling interval is that the sampling interval position in over-sampling waveform has material impact to producing the data that extract.Figure 16 shows the screen of grabbing for the test procedure of this investigation.In Figure 16, untreated, over-sampling waveform is presented at the top of window.As a reference, below over-sampling waveform, show by the frequency spectrum corresponding with it by the extraction waveform that each interval is averaged to generation.In the bottom of window, show by the frequency spectrum corresponding with it by obtain the extraction waveform that single sampling forms from each interval.User's input control below single sample graph allows described user to select the sampling for extracting in each interval.The image providing in Figure 17 has been demonstrated the use of this control to study the effect of sampling interval position in single extraction.The image providing in Figure 17 shows the image of obtaining in different positions along described sampling interval.An attracting result of this research is that to have the characteristic of how many extraction waveforms and spectral change be because sampling location in interval changes.
The ultimate principle of self-adaptation adjustment sampling interval comprises need to obtain data compression, need to change the radio-frequency component that statistical confidence maybe needs to adjust measured data.Any these need in the response that changes the mechanical response showing or external trigger (being process variable), to be identified.For example, during normal operation simultaneously parameter information generally in normal range, can extract planning according to the preferred embodiment selectivity of programming, with the relatively large data block of example interval collection from long.Then, in order to be adapted to the state of one or more changes, selectivity that can be identical extracts planning, with the data block relatively more frequently gathering and selectivity extraction is relatively little.Alternatively, can adopt adaptive optionally extraction by adjusting sampling rate, during the lasting sampling interval fixing, gather more or less data.
In addition,, in given over-sampling waveform, wish to change sampling interval in response to the variation characteristic of over-sampling waveform itself.For example, if bursting out of energy is apparent (may owing to impact or some other similar pulse events), the more closely-spaced contiguous time domain that can be used for bursting out, and large interval is used for other place.Further embodiment is provided as in response to situation of change, revises the sampling rate of over-sampling waveform itself.
The further application of the sampling interval changing or the sampling rate of variation comprises selects sampling interval adaptively, selects adaptively F mAX, select adaptively the mean value of sampling interval data, and select adaptively selectivity extraction technique.Select preferably based on signal characteristic at every turn, for example, can be detected and trigger by the result that uses the over-sampling data analysis of programmable logic device operation to study, or detect and trigger by end user's logic of class.The combination of adaptive sampling interval and adaptive sampling rate can be used to call.
By lengthening interval or increasing sampling rate, can be in each sampling interval, sampling more data.The data that increase in this gatherer process expose and make selectivity extraction technique distinguish in time or pick out relatively rare event, if use statistics or other extraction technique optionally, amplitude or with related other metrical information of present event be recognizable, to distinguish in time one or more such characteristic of event.Selectively, by shortening interval or reducing sampling rate, can be in each sampling interval, sampling data still less.The data that reduce in this gatherer process expose and make selectivity extraction technique in specific time domain, record higher frequency measurement composition.
use the selectivity of auto-correlation, frequency transformation or cepstral analysis to extract
Traditionally, in the first step, extract the over-sampling data in each sampling interval, treatment step is subsequently the waveform by analyze acquisition with auto-correlation, FFT or other frequency transformation or cepstral analysis.Some embodiment of the present invention identifies the significant information being included in over-sampling data by these and other analytical technology to the data in sampling interval.For example, the circulation pattern in sampling interval data set can be by adopting programmable logic device to detect.Wavelength or in sampling interval duration between continuous appearance mutually of this pattern, can disclose the periodicity instruction with the contrary cycle content of content non-periodic.
The extraction of the analysis and selection of the data of over-sampling is typically for generation of the scalar value of one or more its sampling interval of expression.In the time identifying the characteristic cycle or that other is relevant, can further in described one or more scalar value, increase attribute or characteristic mark.
For example, one optionally extraction process can from each sampling interval, produce first, second, and third scalar value.For example, can be extracted in single selectivity extraction process, to produce mean value, minimum value and maximal value from all data at single interval.Except these 3 scalar value, the characteristic of for example periodic feature or qualitative characteristics can be triggered, or cause and effect data characteristics is triggered, or Gauss normal distribution characteristic can be triggered.These all calculating can realize at the single Data processing from sampling interval.Configurable described 3 scalar value (for example mean value maximal value and minimum value) and attribute or property feature (for example indicating possible cause and effect data set or the information state in cycle).Configuration data is typically by understanding to collect by preferred digital data communication agreement with by the information of FPGA (Field Programmable Gate Array), wherein preferred digital data communication agreement is applicable to actual data storage, data transmission, data receiver, data processing and data analysis.
the additional application of over-sampling data
Because machine machine upkeep system of future generation becomes more advanced and integrated high performance electron device, their front end also must possess capability and produce the data of over-sampling.The sampling rate of these circuit is measured the required high 10-20 of sample rate doubly than first conventionally.Described processing, by Measurement bandwidth notebook data being extracted and filtering is expected with acquisition, is removed high fdrequency component.The reason of doing is like this in exploitation and maintenance efficiency-this has designed a high performance front end in essence more advisably, and with it obtain high and low frequency data.In other words, with the current state of this area, low-frequency data can obtain from use the high frequency stream of same hardware, and there is no and extra cost-use current standardized digital signal processor can relatively simply obtain the additional treatments that needs filtering and extraction.
A spinoff of this method is that original data stream unfiltered and that do not sample comprises also can be for PeakVue tMthe over-sampling data of analyzing.In history, these full sampling rate measuring route only produce encapsulating type measured value, and it remains main output, but now by using PeakVue tM, identical raw data can be simultaneously for peakology.In addition, other the representational scalar obtaining in peak value retention value or MAX value and never extraction stream can be used in sampled signal processing path and carry out peakology, and doing is like this RMS, peak value and the peak-to-peak value in standard bands analysis.
Acceleration input is used in traditionally in protection system and comprises RMS, peak value (Peak), peak-to-peak value (Peak-to-Peak) and S to produce mAX" shell vibration " measure, (note, " protection " refers in order automatically to trigger the on-line vibration analysis such as the object of the mechanical shutdown of turbine), also can be used to produce PeakVue tMwaveform, and if need, as the frequency spectrum of further analyzing.
Speed input is used in traditionally in protection system and comprises RMS to produce, peak value, and " the shell vibration " of peak-to-peak value and SMAX measured, and also can be used to produce PeakVue tMwaveform, and if need, as the frequency spectrum of further analyzing.The numerical differentiation of over-sampling speed data produces the acceleration information of over-sampling, and then described acceleration information is used to produce PeakVue tMwaveform, and if need, as the frequency spectrum of further analyzing.
Displacement input is used in traditionally in protection system and comprises RMS, peak value, peak-to-peak value and S to produce mAXradial vibration displacement measurement or axial thrust displacement measurement, also can be used to produce PeakVue tMwaveform, and if need, as the frequency spectrum of further analyzing.The digital double differential of over-sampling displacement data produces the acceleration information of over-sampling, and then described acceleration information is used to produce PeakVue tMwaveform, and if need, as the frequency spectrum of further analyzing.
in sampling interval, characterize one or more events
In certain embodiments, the event in sampling interval is described in time domain.For example, except a peak value or multiple peak value of catching within an interval, may be identified at least one the given number percent in amplitude, for example 80% of maximum peak height, multiple peaks, and further identification have how many characteristic items occur in the spacing between interval or event-in table 4 " measurement interval ".Such event can further identify conventionally from mechanical fault, and for example impact or friction or sensor fault maybe may be identified as from other physical source.The significant amplitude that occurs in the peak value at whole interval can identify possible result or lack of lubrication, fatigue crack by experience logic, break or the consequence of gear defects.In over-sampling data sampling interval, will further be discussed in this manual for many technology of identifying event signal.
digital data transmission
But another important application of various embodiments of the invention relates to the digital data transmission from sensor to main frame.Than transmitting data with analog form, transmit in digital form data and there are many advantages, especially in some applications.For example, wirelessly transmit data by bluetooth or other wireless signal from sensor to hand-held data acquisition unit or analyzer and there is advantage: do not need cable, the risk of operator grasps collector or analyzer is less, motion more freely, more comfortable, can without retain wireless senser, etc.For another example, by wireless radiofrequency data be sent to node for monitoring the healthy radio conditions of machinery or hub or network there is advantage: can have multiple redundant signals transmission paths, so pipe laying does not save money, install fast, be easier to settle, more easily connect in the place that is difficult to arrive, reduce operator's risk, etc.For another example by directly there being thread path, for example, bus communication protocol in multi-thread signal wire or for example carrier signal by power transmission line or for example bus network, transmit numerical information and there is advantage: digital data signal is conventionally than more robust of the simulating signal of long lead, more reliable, because digital signal does not allow to be subject to temperature, the impedance of cable, the electric capacity of cable, the impact of electromagnetic interference (EMI), and they tend to not need the digital branch to communication path to calibrate or compensation adjustment.
With reference to table 2 above, complete in step D after preprocessed signal has been the conversion of hits digital data, data are treated to digital form.Bandwidth in step D is quite high, and for the structure of some narrow bandwidth, it may be unfavorable for transmitting full bandwidth data.For example, for from digital sensor to portable collector or the wireless blue tooth signal of analyzer, be particularly conducive in digital sensor, by the numerical data of over-sampling or pretreated numerical data, or the data of sampling interval data set or the characteristic at interval, or before numeric field data is sent to portable collector, by the one or more step process data in step e, F, G, H and I, for further the data in this generation being processed in the later step of process that produces expectation information.
In a similar manner, before it is sent to near of massaging device near of sensor device, be conducive to the data of step D to J in processing list 2.This is applicable to roam data acquisition, wireless surveillance and in-service monitoring.
But, this was conducive to before transmitting numerical data and completing the aftertreatment and extraction of described remainder, carry out a part of aftertreatment, or a part extracts, and may be different from the subsequent processing steps of another equipment of the another location that analog-to-digital described equipment or position occur.
Table 3: for understanding the process of analog sensor signal information
For example, shown in table 3 " TBD " or " undetermined " for selecting to carry out aftertreatment, all or a part of in the extraction of the analog vibration analyzer in step (II) or the digital accelerometer in (VI).This has logically required not carry out relevant aftertreatment and other and has extracted in the device of digital vibration analyzer as (IV) computer for analysis device or (VII).
measure speed and measure interval
Table 4. is measured speed and is measured interval
Table 4 provides the measurement speed that how to use selectivity to extract and the demonstration of measuring interval.The given in this example sampling rate of ADC, for example 200kHz.The the 2nd to the 4th row show the order of magnitude scope of the induction incoming frequency that from slow (every 10 minutes 1 cycles) to (60000 cycles of per minute) soon recognize.These expressions occur in work period or the recurrence interval in measured machine or measured process, and it has the possibility occurring in the work period.For example, the imbalance of machine probably circulates and occurs once at every turn.Another example, when each tooth is connected to another, gear engagement event may occur, and complete gear swing circle once, will occur hunting tooth defect.Many mechanical signals of the similar work period from repeating will have periodic content.Again for example, in corona or electric discharge or the friction of characteristic frequency, typically have relative high frequency, be everlasting in 5 to 100 kilohertz range, the in the situation that of corona, the important periodicity relevant to line frequency can be with described corona or discharge signal.In another example, there is the interactive process of fluidic structures, for example aq slurry process or crushing process or ore crushing technique or shear technique or turbulent flow process perhaps many other manufactures or dynamo-electric technique will have characteristic signals, described signal typically has frequency content, according to the character of processing operation, it it can be the cycle or acyclic.Manyly thisly will to enter the sensor signal of the ADC that meets sampling rate for spreading all over for responding to cycle of induction coverage of incoming frequency.
Measure the F that speed row (the 4th in table 4 and the 5th row) relate to traditionally the sizing device of analyzing for measuring vibrations or motor current signal max.In this case, be similar to F maxmaximum frequency can based target periodicity induction input, or based on and another actual limitation relevant with the something that is not the work period, for example measuring system is limited to or the requirement of limiting to or having the conventional target measurement speed of induction input rate applied widely is set, and selects.No matter root is in what reason, and typically State selective measurements speed is greater than 2 the Nyquist factor, and for example 2.56, be typically applied to derive and measure speed, therefore obtain and measure interval.
Last four row of table 4 relate to measurement interval characteristics, comprise the interval duration, some (for example measure per work period of interval, for example component cycle period or the relatively span of complete information aggregate of covering), the inverse at every measurement of work period interval, and the over-sampling numerical data that more finally (sampling rate/measurement interval duration) collects at each measurement interval.
In the time that the work period of measurement interval Duration Ratio repeated events is grown, for example fuzzy event produces the measured signal of 40kHz in the speed of 1kHz, and then given measurement interim (for example, sampling interval), and signal will circulate.Because friction is not periodically input conventionally, there is not periodic signal in possibility in measurement interval.On the contrary, it may be random, aperiodic input.
Typically, for the measurement interval from single over-sampling obtains the instruction in significant frequency or cycle, the number of cycles at each interval is answered >>1.The interested radio-frequency component of most of fault analyses is inevitable to be obtained from analyze described waveform.For example, use the machinery vibration analysis optionally extracting in the over-sampling data relevant to bearing, to find failure message, but rotate relevant survey frequency in order to distinguish conventionally with ball rotation, idle running and rotating cage, should be to extracting or optionally extracting measured waveform data stream and analyze.
attribute of a relation
Attribute of a relation has passed on the FPGA (Field Programmable Gate Array) of details that may be relevant to measured value to understand.Attribute of a relation conventionally on mathematics, use and, poor, compare, n order derivative or n integration and obtain.Generally designate the scalar that attribute of a relation measures with another one relevant, but the in the situation that of a selection, it can become significant tolerance according to own characteristic.Attribute of a relation can be from other data or out of Memory retrieval, and may there is the association of matter, for example " qualified " or " defective ", "Yes" or "No", " add " or " subtracting " " opening " or " pass ", " low " or " medium " or " height ", " normally " or " hypervelocity ", or " other ".The main target of attribute of a relation is to provide the details that the logic device that can be programmed is further understood.Programmable logic device may possess and in conjunction with significant data set and situation information, the deciphering of attribute of a relation more correctly be applied together with scalar value, thereby has minimized false positive number or false negative result.FPGA (Field Programmable Gate Array) can be used to expansion and is similar to or is at least illustrated in the interior characteristic of paying close attention to of sampling interval part of the over-sampling of waveform.The waveform of this reconstruction may be more neat than original waveform, because uninterested data are weakened, and interested data show with record or graphics mode, even emphasize so that machine or people's deciphering.The data of the waveform tolerable high bandwidth of rebuilding are like this transmitted through narrow bandwidth path, again expand after a while again.
show the example that the sequence of the normal state that comprises cause and effect data and abnormal Gaussian data collection distributes
In table 5a, 5b, 6a, 6b, 7b and 7a below, provide some examples, for the data set of normal state or the Gaussian distribution of reflected measurement value, be typically presented at the n1 row of each table.The example of another data set provides the cause and effect data that include other normal state data.Table 5a, 6a and 7a are endowed " 0 " (zero) value benchmark and export all just measurements.Common situation is, null value can indication fault sensor, but the signal of restriction or slicing can have different impacts and those skilled in the art will understand physical conditions by measured value applied logic according to the relevant information of given sensor and its configuration.Can from these exemplary forms, observe out multiple output can obtain from of a data set stream.Multiple output can comprise for example intermediate value, multiple scalar value of maximal value and scope (MAX-MIN), and it may further include multiple attributes or characteristic mark.All these can obtain from each data centralization collection of a series of data centralizations the waveform of the attribute that is assigned to some or all scalar value in waveform.
In table 5a and 5b, row " n1 " represent the ordering data set of approximate Gaussian normal distribution.Row " n2 " have presented the distribution of identical data set to " n10 ", wherein short duration causal event of one of replacement normal distribution value has occurred for the impact of replaceable " 10 " value representation peak value or other.In table 5a and 5b, highlight unit and can indicate causal influence event.This has shown the size of extremum, " mxm. " or MAX, and adding-subtracting measured value, in the situation of some " minimums " or MIN, be easy to all other tested values and statistical parameter difference with described data set.Also be to it is evident that SDEV, mean value-intermediate value (Mean-Median), mean value-mode value (Mean-Mode), (MAX-MIN)/(Mean-MIN) and MAX-MIN are likely indicating of this ordering causal event exerting an influence or other causal event that produces peak value in data set.
Table 5A, the sequence that normal state and cause and effect data centralization are greater than 0 measured value distributes, and wherein HI high impact value occurs in one-shot measurement, and has replaced a value of other normal distribution.
In table 6a and 6b, row " N1 " represent the ordering data set of approximate Gaussian normal distribution.Row " N2 " to " N10 " have presented similar data set and have distributed, wherein from the high scalar value generation induced response (in row A in table 2) of the additional interruption of physical event.An example of physical event is friction, and the grain-boundary sliding that wherein exceedes a solid surface of the grain boundary (grain boundary) of another solid surface produces sound, vibration, or other useful signal information that can receive by sensor.Other example of high signal message of being interrupted can produce the data set of the similar deflection illustrating to " N10 " at table 6a and 6b " N2 ", the data set of described deflection may comprise corona, electric discharge, boundary lubrication condition, mixed mode lubricating status, fluidic structures reciprocation, abrasive wear, stick wearing and tearing, cutting-vibration, stick-slip, turbulent flow, leak holes, dry contact, become flexible, touch rub, broken, shear, tear, tear to shreds, collision, Quick Oxidation, ftracture, peel off, cut, glue together, the opening or closing, explode and ignite of some connection.Physical theory and experimental evidence can be used for distinguishing and distinguishing time domain and the transform domain analysis between these and other event.Such analysis may be programmed in firmware or software or hardware logic and realize fast automatic deciphering in the mode extracting by selectivity.In table 6a and 6b, highlight unit and can indicate causal influence event.This amplitude that has shown " the highest value " or MAX is diacritic.Also be that the mean value that it is evident that statistics, intermediate value, pattern, SDEV, mean value-intermediate value (Mean-Median), mean value-mode value (Mean-Mode), (MAX-MIN)/(Mean-MIN) and MAX-MIN are likely indicating of this ordering causal event exerting an influence or other causal event that produces peak value in data set.
Table 6a, the sequence that normal state and cause and effect data centralization are greater than 0 measured value distributes, and wherein source produces the induction information of cause and effect type discontinuously.
Data set n1 n2 n3 n4 n5 n6 n7 n8 n9 n10
Mxm. 5 12 8 16 13 15 9 15 10 11
The second height 4 12 7 15 11 14 6 13 10 10
Third high 4 11 7 11 7 14 6 10 7 9
… 3 11 5 8 5 10 4 9 5 8
… 3 8 4 8 3 9 3 7 4 6
… 3 6 3 4 3 7 3 5 3 3
The 3rd is low 2 3 3 3 2 2 2 2 2 2
Second is low 2 2 2 2 2 2 2 2 2 2
Minimum 1 2 1 1 1 1 1 1 1 1
Mean value 3.0 7.4 4.4 7.6 5.2 8.2 4.0 7.1 4.9 5.8
Intermediate value 3.0 8.0 4.0 8.0 3.0 9.0 3.0 7.0 4.0 6.0
Mode value 3.0 12.0 7.0 8.0 3.0 14.0 6.0 2.0 10.0 2.0
SDEV 1.2 4.3 2.5 5.5 4.3 5.6 2.5 5.0 3.4 3.9
Mean value-intermediate value 0.0 -0.6 0.4 -0.4 2.2 -0.8 1.0 0.1 0.9 -0.2
Mean value-mode value 0.0 -4.6 -2.6 -0.4 2.2 -5.8 -2.0 5.1 -5.1 3.8
(MAX-MIN)/(Mean-MIN) 2.0 4.6 3.6 8.4 7.8 6.8 5.0 7.9 5.1 5.2
MAX-MIN 4.0 10.0 7.0 15.0 12.0 14.0 8.0 14.0 9.0 10.0
Data set is understood Gauss Cause and effect Cause and effect Cause and effect Cause and effect Cause and effect Cause and effect Cause and effect Cause and effect Cause and effect
Gauss's attribute 1 0 0 0 0 0 0 0 0 0
Peak value attribute 0 1 1 1 1 1 1 1 1 1
Friction attribute 0 1 1 1 1 1 1 1 1 1
Fault attribute 0 0 0 0 0 0 0 0 0 0
Table 6b, the sequence of the measured value that normal state and cause and effect data centralization add deduct distributes, and wherein source produces the induction information of cause and effect type discontinuously.
Data set n1 n2 n3 n4 n5 n6 n7 n8 n9 n10
Mxm. 2 9 5 12 10 11 3 12 7 8
The second height 1 8 4 5 4 7 1 7 2 6
Third high 1 3 1 1 0 4 0 4 1 3
… 0 0 0 0 0 -1 0 -1 0 0
… 0 -1 0 -1 -1 -1 -1 -1 -1 -1
… 0 -1 -1 -2 -1 -2 -1 -2 -1 -1
The 3rd is low -1 -5 -2 -5 -2 -6 -2 -2 -2 -2
Second is low -1 -8 -2 -8 -2 -11 -3 -6 -4 -5
Minimum -2 -9 -4 -13 -8 -12 -6 -10 -7 -7
Mean value 0.0 -0.4 0.1 -1.2 0.0 -1.2 -1.0 0.1 -0.6 0.1
Intermediate value 0.0 -1.0 0.0 -1.0 -1.0 -1.0 -1.0 -1.0 -1.0 -1.0
Mode value 0.0 -1.0 0.0 None 0.0 -1.0 0.0 -1.0 -1.0 -1.0
SDEV 1.2 6.4 2.9 7.2 4.9 7.7 2.5 6.7 3.9 4.9
Mean value-intermediate value 0.0 0.6 0.1 -0.2 1.0 -0.2 0.0 1.1 0.4 1.1
Mean value-mode value 0.0 0.6 0.1 None 0.0 -0.2 -1.0 1.1 0.4 1.1
(MAX-MIN)/(Mean-MIN) 2.0 9.4 4.9 13.2 10.0 12.2 4.0 11.9 7.6 7.9
MAX-MIN 4.0 18.0 9.0 25.0 18.0 23.0 9.0 22.0 14.0 15.0
Data set is understood Gauss Cause and effect Cause and effect Cause and effect Cause and effect Cause and effect Cause and effect Cause and effect Cause and effect Cause and effect
Gauss's attribute 1 0 0 0 0 0 0 0 0 0
Peak value attribute 0 1 1 1 1 1 1 1 1 1
Friction attribute 0 1 1 1 1 1 0 1 1 1
Fault attribute 0 0 0 0 0 0 0 0 0 0
In table 7a and 7b, row " N1 " represent the ordering data set of approximate Gaussian normal distribution.Row " N2 " have presented identical data set to " N10 ", wherein alternative one or more " 0 " value representation be interrupted or permanent fault sensor or fault or do not move machinery or the machinery of not operational process or the machinery stopping or the process stopping or reverse or the process that reverses maybe may produce from otherwise can produce the zero level output of sensor of the value that is significantly greater than zero or other causal event that " in noise " level is exported.Note that for other example and can find to transmit different products, for example, is not the extremum of null value.In the example of table 7a and 7b, the bias voltage of this low value is measured also and can be had zero or to approach other of physical event of zero induction input former thereby produce owing to producing.The information that highlights unit and may disclose such cause and effect data set in table 7a and 7b.Please note, the mark of " minimum " or MIN, zero or approach null value mark, intermediate value and pattern mark and for any division by 0 or almost nil tested value or calculated value, the mark of division by 0 mistake (for example row " n6 " arrive the SDEV in intermediate value and the row " N10 " in " n10 ").
Table 7a, the sequence that normal state and cause and effect data centralization are greater than 0 measured value distributes, and wherein fault sensor or faulty circuit produce the tested value of atypical zero.
Table 7b, normal state and cause and effect data centralization add and subtract the distribution of sequence of measured value, and wherein fault sensor or faulty circuit produce the tested value of atypical zero.
Data set n1 n2 n3 n4 n5 n6 n7 n8 n9 n10
Mxm. 2 2 2 1 2 1 2 1 2 0
The second height 1 1 1 1 1 0 1 0 0 0
Third high 1 1 0 0 0 0 0 0 0 0
… 0 0 0 0 0 0 0 0 0 0
… 0 0 0 0 0 0 0 0 0 0
… 0 0 0 0 0 0 0 0 0 0
The 3rd is low -1 0 0 0 0 0 0 0 0 0
Second is low -1 -1 -1 0 0 -1 0 0 0 0
Minimum -2 -1 -1 -2 -1 -2 -1 -2 0 0
Mean value 0.0 0.2 0.1 0.0 0.2 -0.2 0.2 -0.1 0.2 0.0
Intermediate value 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
Mode value 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
SDEV 1.2 1.0 0.9 0.9 0.8 0.8 0.8 0.8 0.7 0.0
Mean value-intermediate value 0.0 0.2 0.1 0.0 0.2 -0.2 0.2 -0.1 0.2 0.0
Mean value-mode value 0.0 0.2 0.1 0.0 0.2 -0.2 0.2 -0.1 0.2 0.0
(MAX-MIN)/(Mean-MIN) 2.0 1.8 1.9 1.0 1.8 1.2 1.8 1.1 1.8 0.0
Maximal value-minimum value 4.0 3.0 3.0 3.0 3.0 3.0 3.0 3.0 2.0 0.0
Data set is understood Gauss Gauss Gauss Gauss Gauss Gauss Gauss Gauss Gauss Cause and effect
Gauss's attribute 1 0 0 0 0 0 0 0 0 0
Peak value attribute 0 0 0 0 0 0 0 0 0 0
Friction attribute 0 0 0 0 0 0 0 0 0 0
Fault attribute 0 0 0 0 0 0 0 1 1 1
extract diagnosing machinery vibration fault by selectivity
Figure 18 shows the example that can be used for extracting by selectivity the spectrum mode of the mechanical vibration fault of diagnosis.In the practice of the data acquisition monitoring for machine performance, the scope that frequency spectrum data is typically stored FMAX is about 5kHz, even if data can be at the F that can produce about 80kHz mAXspeed sample.The waveform of sampling is arrived lower frequency waveform by " extraction ", but it may be beneficial to the frequency resolution that provides higher.It is for detecting whether there is high frequency composition that high-frequency detects (HFD), but a kind of technology of the pattern analysis of described higher frequency composition is not provided.Optionally extracting can be by for example, at the storage characteristics (mean value that extracts the data sampling being conventionally dropped in waveform, intermediate value, kurtosis etc.), the diagnosis of the higher frequency data of for example current of electric data and mechanical vibration data is provided, and without preserving/store whole high-frequency data sampling set.(apostrophe (') be used for these calculating of sampling (being dropped in the common extraction waveform) execution between the extraction sampling to normal state to distinguish to the similar computing of the waveform execution to after the extraction of final normal state.) using the data acquisition unit with multiple processors, the FPGA of for example Fig. 1 may be conducive to calculate polytype waveform extraction technique simultaneously.Figure 18 shows the F of the frequency spectrum (using the frequency spectrum of the extraction technique of normal state) that may be present in typical storage mAXon possible diacritic frequency signal.
the material extracting by selectivity diagnosis in processing
Optionally extract over-sampling information and may find potentially the evidence of structure resonance or friction, due to the dry contact in process or boundary lubrication or pressure leakage or solid matter motion, or the impact of the construction package of the interior material of process to for example pipe arrangement.Resonance be for example structure bend mode natural frequency amplitude increase until the energy offset of decay the positive action excitation being subject in the frequency of excitation energy.It is not in this important natural frequency, but is in fact that excitation forces function to put into energy in resonance frequency." feedback " in picture microphone, it should be in fact to extract by selectivity the resonance that detects rising together with the signal designation technology in one or more cycles.
The friction unusual source of wide-band vibration energy often, particularly in the frequency more much higher than natural resonance frequency.Should be very easy to find dry contact friction (friction factor～0.3), boundary lubrication (friction factor～0.1), turbulent flow and fluid leakage.It should be noted that turbulent flow and fluid leakage often occur in situation below: gas or liquid through aperture exceed the velocity of sound, therefore, send the ultrasound wave steady state flow of high frequency.Optionally extract and can be used for distinguishing the periodicity of the signal message in sampling interval, lack periodically, or auto-correlation, and relatively these things between subsequent sampling interval are distinguished friction and by other itself and other high-frequency information source region.
Fluid turbulent in pipe arrangement, the cavitation erosion at the impeller back side and surperficial corrosion are also to use selectivity to extract the input of the energy that can detect by the accelerometer of close contact.Boulder cracker, cement kiln, the sheet material of rotating grinding, the grinding of abrasive machine, the cutting of lathe, milling spindle is all the performance that had or the activity of mechanical vibration or other bad performance.Selectivity extracts and can be used for monitoring these qualities.
Event in detection, location and tracing process can be with realizing as the transducer array of vibration transducer or other dynamic transducer.Embodiments of the invention comprise according to always measurement, analysis, monitoring, adjusting or the control operation for the sensor of mechanical health monitoring.Application comprises the structure that stands resonance or unstability.Further application comprises disintegrating machine, flour mill, comminutor, pipe arrangement, the scale pan, scraper bowl, scoop and the structure by accelerometer monitoring in a preferred embodiment.Alternative embodiment adopts other transducer.The sensor that various embodiment use is to be most possibly selected to stress wave, vibration, strain, sound and/or hyperacoustic response characteristic to measure.
It is a kind of degree of approach of definite event and the technology of time sequencing that the selectivity of over-sampling data extracts.The degree of approach is to decay to establish with relative event signal the time of arrival by comparing event detection relative in sensor array.
Preferred selectivity extraction technique comprises one or more following quantitative scalar value from over-sampling data set and the room and time sequence analysis of quantitative attributes: maximal value, minimum value, mean value, intermediate value, standard deviation, scope, kurtosis or measure of skewness and peak-to-peak value wavelength (for example, radio-frequency component).Optionally extract audio feature extraction or the fingerprint recognition that also can comprise machinery or process, similarly " sound characteristic extraction " or " audio-frequency fingerprint identification " in " MPEG-7 audio frequency and surmount: audio content index and retrieval " by John Wiley & Sons, Ltd. Hyoung-GookKim, Nicolas Moreau and Thomas Sikora are open in 2005.
Space layout in sensor array considers measured distance, and the geometric configuration of area and volume is logical.For example, the spacing between neighboring sensors can be enough little, and an event can be realized by the more than one sensor in sequence.In addition, described spacing is logically arranged, makes signal propagation time typically be longer than or be longer than the sampling interval of a data set, and described signal propagation time is through structure or fluid media (medium), from the position of an event to first sensor, then to the signal propagation time of the second sensor.For example, if signal passes described structure-borne with the form of stress wave with the velocity of sound of place structure, the signal demand of then propagating with this velocity of sound is longer than the sampling interval obtaining to the second sensor from first sensor.By this way, an event will distinguish from the sampled data set of two sensor image data with a while at least in part.
Embodiments of the invention adopt the array that comprises multiple sensors.Where array can be arranged by two or more sensors are several.Typically, this is a two-dimensional array.For example, strategically alignment sensor covers the significant information of two-dimensional surface (, rectangle is annular or cylindrical) or three-dimensional surface (, spherical or conical or other rotational symmetry or truss or other structure) with collection.Described array can be one dimension, for example line or radius.Described array can be fix or movably.Process by described array or other relevant medium may flow into or mobile.Described array can show as line scanner.Can be by one or more such technique construction numerical datas, to construct visual representing, for example pattern matrix or " picture ".Pattern system can be used to describe the time sequencing or the space representation that measurement data are transformed into the information of understanding by the mankind or by FPGA (Field Programmable Gate Array).
Figure 19 shows sensor 5,10,15,20,25,30,35 and 40 for example around the external diameter of conical or cylindrical grinding machine or process pipe arrangement, is placed on the example of lip-deep sensor array.In grinding machine or pipe arrangement, include process material, for example slurry of solid, semisolid, pasty state or solid and liquid or gel or liquid, wherein process material shows as fluidic structures reciprocation or solid contact.Gu-Gu reciprocation can relate to the sliding contact of the form of the combination of dry contact or boundary lubrication contact or corrosion or grinding or crushing or wearing and tearing or failure mechanism.Gu Gu-reciprocation can relate to impact and knock-on.The fluidic structures reciprocation that inside comprises structure or the reciprocation of semisolid/pasty state/gel can relate to laminar flow or turbulent flow or cavity and form or other cavitation erosion or pressure or body force.Some positive actions produce broadband or the broad space array of the energy that inputs to sensor.Other local event, the event A for example describing in Figure 19, B and C occur in one or more positions and sequence.
A primary theme of the embodiment of the present invention be occur in approach and there is time sequencing and other sensor input the difference between different events.For example, if a peak event is to be detected in the first sensor with First Characteristic signal, described First Characteristic signal is detected after a while in the second sensor, then in the 3rd sensor, be detected etc., FPGA (Field Programmable Gate Array) can be used for distinguishing First Characteristic signal and be reported in the arrival order of the signal of the different sensor of sensor array.Sequence itself can be carried out described in inverse event in estimated position and the sequential of room and time for FPGA (Field Programmable Gate Array) provides Useful Information.Therefore vibration or stress wave are used to the position of the impulse source of " triangulation ".For example, event A may be detected at sensor 5, then in sensor 10, is detected, and then sensor 25 and 20, in almost identical time detecting to event A, is then detected in sensor 30, etc.High-frequency signal information typically is along with distance and reduces rapidly, so these signals are weakened, in the time that FPGA (Field Programmable Gate Array) has identified and has predicted next position and sequential, but they also can distinguish with ground unrest, using the signal characteristic as at predictable next sensing station.
Another primary theme of the embodiment of the present invention is unallied or the compensation of background signal information or minimizing.This completes by the signal message that deducts one or more distance sensors collect from the relevant signal information collecting with the sensor in more close in fact source or signal message source.The filtering technique that this utilization is applicable.For example, a compensation sensor that can be identified as for measuring current background signal in the array of sensor.Background signal can comprise noise, from the signal message of other millwork with from the signal message of process operation.In some cases, background signal may be " loud ".In these cases, by using selectivity extraction technique to distinguish especially valuably interest characteristics signal and background signal information, wherein optionally the background signal information of the signal message comparison static analysis of extraction technique to statistical study is more responsive.
Information from array can be understood with programmable logic device and human intelligible, to characterize and to give birth to the process characteristic in container handling.The key difference sexual factor of characterization operational circumstances comprises that the multiple sensors of use are in room and time detection and tracking event.
In Figure 20, represent the decay of signal message around the star-like profile diagram of event A, B and C expansion.For example, be powerful high level signal or low-frequency signals from the signal message of event A.Compared with higher frequency signals information, lower frequency is decayed less conventionally, and propagation distance is farther.In this example, noise information reduction is to before detectable reasonable level therein, from the Information Communication of event A farther distance.Event B represents the signal of intermediate amplitude or intermediate range frequency, so along with the distance of relative shortening, differentiable dropout.In a similar fashion, event C represents high-frequency signal or low amplitude signal.
From the signal message of event A, B and C typically from source radiation and through media transmission.Figure 20 has represented two-dimensional measurement space diagrammatically, as the axisymmetrical surfaces of sensor may be installed around one section of pipe arrangement or grinding machine.It should be noted that be not all source or events be all point source or some event.On the contrary, the signal message source measured by embodiments of the invention can derive from other geometric configuration, the area source of for example line source or similarly body force or pressure.In event, conventionally have ripple information, for example compression stress wave or sound wave or other mechanical energy shift.In these diagram examples, energy wave comprises the event information of propagating from point source.At very first time interval, be expressed as diagrammatically in the space between concentric star profile, gather the over-sampling data for analyzing.
During sampling time interval, for example the duration of the white space between a star graph and the star graph in next outside, gather and process a large amount of measured values.For example, can in a sampling interval, gather 10,100 or 1000 measured values.These data are called as data set, for passing through to use as described herein optionally extraction technique to analyze over-sampling data.Be captured in each sampling interval from the quantitative scalar sum quantitative attributes of two or more sensors in array for comparison and analysis subsequently.
The process of human intelligible and Possible event can be used for building the knowledge base about event machinery possibility signal.Some event is local, and the duration is short.For example, Gu Gu-to impact for relatively short event of typical duration, described event produces the stress wave with the signal that different use accelerometers can detect.Sliding contact is two kinds with turbulent flow may have similar peak amplitude characteristic, but the event of visibly different mean value, intermediate value and standard deviation characteristic.
Use programmable device operative by responding to container the sensed data that the sensor that contacts was collected, the preferred embodiments of the present invention are identified those treatment mechanisms that may occur in container.The technology that described embodiment adopts selectivity for example to extract, with detection event and distinguish multiple induction input sources.The present invention also carrys out the approaching event in space in discovery procedure with measured value array.In addition, the present invention uses from the sampling interval of the adjacent sequential of array and analyzes data, further to locate and or to follow the tracks of/follow event, the signal of described event based on along with distance (mainly) decay and the time of transmitting different sensors signal.Finally, embodiments of the invention provide process the time sequencing periodically interior and motion of matter to represent.Periodically impact by use or other event between time complete.Complete motion by the event location that compares one period.The velocity of sound based on structure or liquid mediums is carried out the process track that locator material moves during the course of locating events and is compared fast in the extreme.The velocity of sound in steel is than the fast order of magnitude of the airborne velocity of sound.In both cases, than typical processing campaign, this is very fast.
If according to planned design and normally operation, some events may be normal state.But in many cases, the event of some types may cause serious spinoff.For example, unstability event or other unexpected saltus step event typically comprise in the time not increasing load, increase the deflection of compressible drive.Not warning before irrecoverable infringement, beam or post may unstabilitys.Plastic yield, the deformation of creep or creep relaxation are typically for the another kind mechanism of load that does not live through elastic response, these impacts can produce catastrophic failure, material sluggish or other may reduce Performance Characteristics.Structural resonance is another characteristic possible with adverse influence, if when particularly it is lasting.
Can use embodiments of the invention to work as External Force Acting while causing unstability, creep, plastic yield and resonance, use about the structure being just monitored with about the knowledge of the possible expression behaviour of described structure, export from two or more inductions of sensor array by supervision, analysis and comparison, detect unstability, creep, plastic yield and resonance.Based on load paths misalignment, based on surface tension or based on typically in the motion of transverse shifting of spacing, can detect close to unstability situation.Active instability status can enough be detected rapidly, is used for automatically unloading load, thereby prevents further destruction.
The hysteresis that creep or plastic yield situation can be observed by tension force and compressive load cycle or be detected by permanent strain.Resonance situation can be detected by the phase place and the deflection that compare between sensor, if feasible, sensor is installed in logic node and anti-node position.These positions can be identified by the pin that is associated with node or junction and the mid point being associated with antinodal points.Pattern analysis or impact test will greatly give a hand.
imaging transmitter.imaging transmitter comprises for example imaging detector of a focal plane arrays (FPA), be suitable for detecting electromagnetic radiation photon, as the combination of ultraviolet ray (UV) or visible ray or infrared (IR) spectral wavelength or one or more different spectral wavelength scopes.Described imaging detector has for example 8 × 8 or 16 × 16 or 80 × 80 or 160 × 120 or 320 × 240 or 640 × 480 pel array conventionally.Each pixel in array is similar to an independent sensor.Whole array signals can test example as the one or more characteristics in fluorescence or phosphorescent characteristics and the more feature of the rubbing characteristics of the heat transfer characteristic of the water properties of the distance of the energy response of the electrical characteristics of the absorption characteristic of the transport property of the radioactive nature of the temperature characterisitic of the chemical characteristic of illumination, color, material, material, material, medium or material, material, material, material, object or area or other spatial character, material, material, material, material.Passively based on just sensed object, the electromagnetic radiation in the surrounding enviroment that medium or material are associated, or the stimulated emission based on electromagnetic radiation or transmission or reflection on one's own initiative, can sense characteristic
The example of the imaging detection of passive type is to use IR detector to observe passively the blackbody radiation that material sends.For example,, by SOFRADIR EC, INC., 373US Hwy46W, Fairfield, the detector that the model that NJ07004USA manufactures is ATOM80 has following specification: micro-radiating curtain of 80 × 80, spectral response is 8-14 μ m, detector NETD<100mK (f/1,27 DEG C), power consumption <0.25W, operating temperature range is from-20 DEG C to+60 DEG C, frame frequency is 30Hz, electric interfaces USB2.0 and 14 bit digital output streams.
An example of active imaging detection is to respond to fluorescence or the phosphorescence of the material that the excitation energy of for example pulse laser is reacted.The fluorescence and phosphorescence feature of material has predictable disintegration constant under the state based on for example temperature of described material conventionally.The transmitter by time pulse excitation energy with the known duration of pulse can be programmed to distinguish active energy source and reason and exercising result from other most of energy of driving pulse by different way, and described transmitter can further be programmed, to distinguish the characteristic response of material to active impulsive energy source.In this example, the first characteristic response can be by the exciting or lack and excite of fluorescence or phosphorescence, and Second Characteristic can be decay or lack decay subsequently.Characteristic response can be by using FPGA (Field Programmable Gate Array) and additional information understand, and described additional information is theory or the experimental evidence of the tested value that is for example used for inferring the temperature of desirable for example material or the concentration of material.
The example of active and passive type imaging detection is approaching near leakage most, and in conjunction with the additional illumination of the illumination of for example additional broadband illumination or additional arrowband illumination or additional high passband or additional low passband illumination, the gas leakage of induction substance, described material absorbing with activating the surrounding enviroment of passive detection common characteristic spectrum can be with.Additional illumination can be the response of expecting with acquisition of stable state or pulse.
Figure 21 shows imaging transmitter 122, and it comprises transmitter part 124, wireless aerial 126, sensor outer housing 128, for airborne sound and unloaded ultrasonic one or more inputs 130 and for one or more inputs 132 of the photon energy of electromagnetic radiation.This description is not intended to limit the logic arrangement of described many assemblies, and those skilled in the art can describe to build transmitter and obtain a sensor or one group of sensor in conjunction with this.
Described imaging transmitter 122 can independently use, or more effectively has the visual field overlapping with the view of other imaging transmitter (FOV).
Figure 22 represented diagrammatically to have the first visual field (FOV1) the first imaging transmitter 122a, there is the second imaging transmitter 122b of the second visual field (FOV2) and there is the overlapped fov FOVs of the 3rd imaging transmitter 122c of the 3rd visual field (FOV3).All overlapping visual fields have surrounded the different impact point from different lens, comprise that mechanical P1, electric component P2, the pipe arrangement with valve P3, line of electric force P4, sky reference by location R1 and earthing position are with reference to R2.Figure 22 also shows a technician who holds for helping the handheld device D1 that is configured to picture transmitter 122a, 122b and 122c.In different embodiment, display D1 can be hand-held or fixing local display, and by using near-field communication (NFC), bluetooth tM, or other wireless communication protocol and imaging transmitter 122a, 122b and 122c carry out radio communication.Alternately, display D1 can with imaging transmitter 122a, 122b and 122c wired connection.
Preferred embodiment for example, is identified by automatic or manual by impact point (P1, P2, P3 and P4) and reference point (as R1 and R2), has realized selection and discrimination process.A kind of technology of doing is like this in order for example to pass through bluetooth tMor other wireless device, send imaging data to handheld device D1 from imaging transmitter 122a, 122b and 122c, and the operator that programmable logic device in the equipment of use D1 is helped is at the scene configured to picture transmitter.In one embodiment, the geometric representation of the focal plane arrays (FPA) in the demonstration of equipment D1 shows described imaging transmitter.In the time checking demonstration, operator can specify one or more targets and one or more reference point.Owing to showing that image is consistent with the position co-ordination of the focal plane arrays (FPA) of pixelation, it is rational using its geometric relationship that represents the space between impact point and reference point.As long as the FOVs of imaging transmitter does not move, this mutual relationship be should give reservation.Even when even FOVs changes, logically retain some or all geometric relationships between impact point and reference point in the material world, and represent in the plane of imaging detector.
As term as used herein, impact point typically in FOV by the position of observed supervision and diagnosis.Typically, monitor and diagnose based on theory and experimental knowledge, completing by the deciphering of the FPGA (Field Programmable Gate Array) with image-forming information.As above discuss, the example of impact point comprises valve, electrical connection, part machinery or power lead, electric switchgear or other target item.
From the optionally extraction of machinery and the image-forming information of electric component
Figure 23 illustrates mechanical P1, how to provide based on Bilateral Symmetry, automatically or semi-automatically configures the deciphering of the extraction of the spatial selectivity of target area.Figure 23 shows mechanical profile, it is characterized in that sector P1A and P1B are bilaterally symmetric, the statistical value that makes to be determined by half can have to by the similar expectation of second half definite statistical value.Machinery P1 probably has the statistically tested value of a considerable amount of pixels, for example, for the data array of imaging detector.For example, each image can have at fragment P1A and also 5 to 50000 values in section P1B.People it is contemplated that bilaterally symmetric many other fragments, as quartile or more sectional area.In a preferred embodiment, FPGA (Field Programmable Gate Array) automatically performs peak value retention value, mean value, intermediate value, minimum value and standard deviation statistically, the comparison of for example, other logic statistics between the left half and the right half of the Bilateral Symmetry object of mechanical P1, to distinguish whether metrical information comprises that whether cause and effect data or each fragment or two fragments are by the test of Gauss's normal state data colony.
By to carrying out such computing from FOV1, FOV2 and/or FOV3 about the numerical data at sector P1A and P1B or about the imaging data of the numerical data in other bilaterally symmetric area fragment, programmable logic device can be understood better and infer correct conclusion, and avoids false positive number or false negative result.For example, utilizing a kind of FAQs of electromagnetism photon detection is the impact of reflection.Can comprise the positional information of sky or the positional information on ground or another photon actual source from least a portion surface reflection from surperficial reflection.By observe article of for example mechanical P1 from multiple angles, unlikely few than at single view of the wrong chance of the instruction of reflecting in multiple views.The impact of this reflection can further be detected, understands and illustrate, and in subsequent calculations, discovery and recommendation, do suitably and process.If do not met with an accident, can issue alarm to operator or technician, with assessment result or carry out other inspection or test or measurement.These observations about multiple FOV perspective advantages are also applicable to other impact point of discussing respectively.
The sound that zero load is vibrated or ultrasonic measurement are directly linked to inner side, visual field or any specific of outside may be unpractiaca.But it is actual that the signal based on experience or knowwhy is associated with to one or more sound or hyperacoustic possible reason or source.Programmable logic device or people's logic of class can be combined this inference of the information from image data source to improve likelihood or the possibility that logic is understood with possible inference.
Preferred embodiment of the present invention can be used for selecting and the object in visual field with similar profile.For example, Figure 24 shows described electric component P2, and described electric component can be the switch enclosure that includes fuse.In visual field, other example of similar object comprises similar bearing on similar parts or the pipeline in switchyard or the similar fragment of pipe arrangement.The advantage that imaging transmitter has be can pick out there is shape for example, multiple article of the similar characteristic of size (allowing apart from lens), profile, temperature or other different and similar geometric configuratioies or amplitude pattern.
In this example, programmable logic device can be identified a series of 14 similar article, and for example fuse P2A is to P2N.In this case, the extraction of the spatial selectivity of view data can comprise and will in pixels statistics in each sector of all sectors for example containing fuse P2A, be reduced to for example scalar value of peak value retention value, intermediate value, mean value, minimum value, standard deviation or variance.Peak value keeps being used in a preferred embodiment.It is likely some fuse and untapped fuses in use.More possible fuses are defective, or appear at the electric wiring that is connected to fuse and have fault.The fuse of estimating each power supply may be higher than the temperature of periphery.The electric fuse that supplies of fault may be abnormal warm or extremely cold.May cannot distinguish and there is no the fault fuse of power supply and other fuse of not powering.Δ (delta) temperature computation can be by carrying out with FPGA (Field Programmable Gate Array) and posterior infromation, to assess the variation of resistance as mentioned below.
Figure 25 shows the pipe arrangement P3 in the visual field of image transmitter with valve.Described pipe arrangement P3 has fragment P3a, P3b and P3C.FPGA (Field Programmable Gate Array) can be used to infer where select fragment P3a, P3b and P3C.The region separately of these three fragments comprises multiple pixel values, on described multiple pixel values, can optionally extract application space, taking by the data decomposition of over-sampling as significant value.In this example, valve is characterised in that it typically has along with pipe arrangement is in upstream side and the downstream of both direction operation.Valve can normally open completely or normally cut out completely.It can change to other state or marginal somewhere from a state in operation.For the mankind in control system or programmable logic device, the valve of understanding and verifying and the normal state situation of pipe arrangement are important.This logic is also applicable to other process and comprises container, and is not only valve.Imaging transmitter can use imaging detector, the operation of the valve that detects and monitor with sound wave or ultrasonic detector or the operation of another process vessel.By comparing expression information and the valve portion from pipe arrangement upstream portion to pipe arrangement downstream part, can logically infer the information about valve mode of operation.The inference based on doing from the information of imaging detector can be confirmed or demolish sb.'s argument to sound wave or ultrasonic signal.
fouling.sometimes, pile up or remove material and can detect with imaging transmitter from the inner surface of pipe arrangement or tank or other container.This detection based on the impact that energy is shifted of the material that adds or remove, described energy jump routine is as at process material with include therein the pipe arrangement of material, the heat between tank or container is conducted or thermal convection.Be arranged on the imaging transmitter of described pipe arrangement or tank or other outside of containers, can monitor the surface energy of pipe arrangement or tank or other container.The baseline image of for example heat picture is typically as benchmark or reference spectrum image.Can, in incrustation scale or other accumulation or erosion or the final place occurring of other removal of estimating material, select one or more impact points.
detect and flow.stream in pipe arrangement or process vessel also can be by detecting or display with imaging transmitter.In a process, disclose the mobility status of pipe arrangement or other container or not the method for mobility status be along expection a flow path, identify two or more impact points.If occur to flow, then estimate that logic temperature or other heated spot resemble will follow this flow path, consider at the beginning of mobility status and the transient state interval of stopping period.For example, if pipe arrangement does not have in peripheral temperature situation, even a temperature instruction just can disclose to flow and occurring in that part of pipe arrangement.Enter the thermal convection of pipe arrangement inner fluid or heat conduction from the fluid transmission in pipe arrangement or transmission and will typically affect pipe arrangement that fluid flows through or the temperature on valve or outside of containers surface.
detect and stop up.when solid material gathers and becomes the obstacle that restriction is flowed or stopped up pipe arrangement or valve or other treatment mechanism, usually need to detect and identify stopping state.For example Figure 25 shows a method, for by the imaging transmitter measurement in entrance or upstream portion and vessel or compare with the outlet of valve or pipe arrangement or other mobile units or the imaging transmitter measured value of downstream part.When not stopping up and under stable situation, crossing described part by a large amount of fluids or other process media flow stronger than ought there is partially or completely to stop up impact on the measurement from imaging transmitter on the impact of the measured value from imaging transmitter.The for example imaging array of a heat picture detecting device can be used for detecting the heat conducting impact of convection current or conduction, and the heat conduction of described convection current or conduction is because near flowing chamber wall changes.
May be constrained to the ability that is disclosed in the stream in pipe arrangement or container or valve as transmitter around the isolation of pipe arrangement or container or valve.May be in some cases, observe temperature and heat conducting heat instruction by identification as the impact point of heat transmission fin or other Heat Conduction Material, wherein for example tube hanger of other Heat Conduction Material or pipe arrangement flange or support or other directly with the heat conduction object of isolating pipe arrangement or container or valve and being connected.
electric power transfer and distribution.electric power transfer section P4A and the P4B of automatic or automanual Piece Selection image-region are shown in Figure 26.The visual transmitter of the demonstrating electrical line of force has sky location context conventionally, but is not always to have.There is the similar electric power transmission of typical case and the outward appearance of dispense articles, make FPGA (Field Programmable Gate Array) can pick out the parts that electric power transfer is distributed set, described electric power transfer distributes set to be made up of for example power lead, power generation column, electric(al) insulator, transformer and more article.As previously mentioned, optionally extraction technique can be used for over-sampling view data to be decomposed into the significant scalar value of analyzing and understanding for further.Conventionally select peak value to keep, because the background inundatory low value data set often of day empty position.The linear characteristic of electric power transfer and distribution member and other diacritic geometry and electromagnetic image characteristic, by the use of FPGA (Field Programmable Gate Array) and people's logic of class of being assisted by FPGA (Field Programmable Gate Array), distinguish them easily with many other situations.
Cause the important indicating fault in the time using imaging transmitter to comprise from wind, rain, snow or condensation, the impact of working that the vibration of the movement of object, detector in visual field, the movement of imaging transmitter or the movement of impact point cause.For example, in the time sun power image being detected in background or in reflection, possible initiating failure instruction.The characteristic of reflection characteristic or Sunlight exposure characteristic or sky background can be calculated in interior completely and should be used to FPGA (Field Programmable Gate Array), for fear of from reflection be exposed to the misoperation that sky and ground cause, false positive number instruction, or the instruction of false negative.
temperature profile.the signal of the electromagnetism that object produces in combustion process, for example active catalyst in combustion process, can provide the instruction about the suitable form of for example catalysis material and the function of object.Be similar to Figure 26, when line of electric force shows the high response of exaggeration compared with background, the for example material of the catalyzer in combustion process may produce predictable geometrics shape, and described geometrics shape uses the imaging transmitter of the selectable electromagnetic spectrum ability of discovery of tool to be monitored easily.If the each independent element of catalysis material can be distinguished using the impact point as independent, as Figure 24, then can be applied to a large amount of pixels to extract as single or several scalar value as the statistical analysis technique of describing in Figure 24.Feature or attribute can further be associated with each scalar value.If multiple parts of catalysis medium cannot be distinguished, so suitably distribute the impact point that leaves and around for example entrance, in pars intermedia branch and exit point, the point of proximity of each point is better.Based on the Δ measured value between the such point of for example entrance and exit, comparison may logically be assigned to be analyzed and reports to the police.In addition, can be based on other characteristic for the absolute reference of analyzing and report to the police, the characteristic that for example process optimization, material degeneration, fuel efficiency, chemistry distribute or other catalyzer is relevant.
the automatic selection of reference point.programmable logic device can be used for one or more reference point of distinguishing and identifying.Two kinds of normally used reference point comprise: (1) position based on land, and as the position of ground location or plant, and sky empty position or the locus of (2).As the position based on land of ground location or locus can be compensation and confirm tested value and provide the reference of use near the orientation of inner or impact point position.According to different scenes, the logic behavior of the position of terrestrial reference position and sky based on a possibility associated will be distinguished.On location logic based on ground in or approach peripheral temperature.Conventionally with being associated compared with lower part of level or " landform " image.Based on sky or the location logic based on space in relatively cold temperature, conventionally approach geometrically the image of good guidance upper part near expansion.Have a variety of methods, those skilled in the art can operation technique, and these and other reference position is understood and distinguished to equipment and programmable logic device.
As shown in figure 27, in this example, reference point R1 and R2 represent respectively day empty position and ground location.FPGA (Field Programmable Gate Array) and/or people's logic of class can be used for selecting these points and be used for points of proximity R1 and R2, build respectively section R1A to R1d and R2a to R2d.The extraction of spatial selectivity can be used for analyzing, confirm and be identified for sky position measurement and ground location measure important reference value statistically.These reference values are used to understand the connotation of actual tested value by FPGA (Field Programmable Gate Array) and people's logic of class, because they may be subject to the impact of environmental characteristics, as the exposure of ground or sky.As mentioned in other embodiments, to carry out optionally and extract, and there is " the X-pattern " of four sections by structure, statistical study can possess a reference value effectively.In order to possess selectivity decimation value, left and right section can be compared, the part of top and bottom can be compared, with by logic testing.The concentric ring of Pixel Information also can be used for replacing or being increased to pie section effectively.Expectation can be from affined data acquisition obtains data community information on statistics Gauss normal distribution just as expecting optionally to extract and analyze from reference point.
In containment building or mine or another such closed region, even out of doors in place, also exist another kind of structure wherein can be identified situation as a reference, for example reflection of floor, wall, ceiling, post, hurdle, window, door, ground, pond or other object or object.The general reference value of selecting is in compensation or proofread and correct and use or for using at ratio or Difference Calculation.
Impact point typically near region for analyzing the focus of data, to obtain about near the meaningful information described region.Within typically impact point is positioned at the pixelated field of view for the first imaging transmitter.In a preferred embodiment, by locating identical impact point by pixelation imaging region field devices independently.
Data acquisition interval can be subcontinuous or be that be interrupted or that trigger or predetermined.
Dissimilar view data can be collected for the analysis in imaging sender system.For example, the first kind is that traditional process is typically processed the discrete image data being displayed in image or video format.In another example, second and the 3rd type be the pictorial data that selectivity extracts, these data are processed with the image information at room and time territory compression over-sampling.
Spread all over this instructions, the figure picture that term image is not only observed on display or the page with people is relevant.For imaging transmitter of the present invention, this word of image relates to the two-dimensional array of intensity or frequency spectrum data value, and it can present with " picture " form pattern.All images of the analog form that replacement human eye is seen, imaging transmitter is understood the data array after electromagnetic spectrum imager and on room and time, is processed numerical information.
A primary theme of different embodiment is that to utilize selectivity to extract the Image Data Compression of " over-sampling " of part be manageable information, described manageable information typical case is the measurement of the significant analytical parameters of being correlated with in logic, and wherein said meaning is to be associated with measurement or the description of the phenomenon with some importance of real world.
image stream and video data.the data of the first kind are typically used in the mankind's the image data stream of watching and understanding, for example, with reference to the display device D1 describing in Figure 22.Described image representation can be whole visual field, the region that only target limits, only higher than the image pixel of threshold level, for example, is equal to or greater than 60 DEG C, or isometric Plotted line, for example isotherm or define the class contour curve of described detector rank in some inside, border.The image of low resolution (40 × 40) can be after about 5 minutes, via wireless HART tMnetwork transmits, or requires higher resolution 80x80, or even higher.Can transmit a part of image and be superimposed upon on the top of the reference picture in the internal memory that is stored in Receiving Host, it provides the geometrical perspective figure of the demonstration and the analysis that can be used for special data.
optionally extract the spatial data from imaging transmitter.the data of Second Type comprise the information of the analytical parameters being associated with described impact point with reference point in image.The data of this Second Type generally include the optionally extraction of two-dimensional image data.For example, use the HART in imager has the visual field of single target point and 2 reference point, 4-20mA signal can send and represent near the selected scalar of selectivity decimation value of impact point that is close to.In addition digital HART, tMmain value can be transmitted together with the scalar value of 4-20mA signal correction, and the scalar value that deputy value can be relevant to the first reference point is transmitted together, and the scalar value that tertiary value can be relevant to the second reference point is transmitted together.Finally, the 4th place value can be transmitted together with the compensation result scalar value calculating, and the environment of wherein said main value or operation variance have obtained compensation on mathematics, for example, use from the information of the first or second reference point and compensate.
optionally extract the time data from imaging transmitter.the data of the 3rd type are the information of the analytical parameters that is associated with impact point in time-series image and reference point.The data characterization of the third type the delta data of a continuous time domain, typically for identifying feature or attribute or the function of quantitative and qualitative analysis, for example there is the stable case of quantitative values, there is the stable state of quantitative values, there is increase or the minimizing of quantitative speed or other value, there is the acceleration or deceleration of quantitative values, there is quantitative variation, have quantitative values shortage degree of confidence put letter or mistake, there is the super scope of the corresponding order of magnitude, and for example " Gauss " or " non-Gauss " or " exceeding control limit " or there is the statistic analysis result of another statistical measurement of corresponding metric.
The data of the 3rd type for determine validity or confidence level or from first or the feature of the investigation result of the data analysis of Second Type be necessary.In addition, exist and much can only use sequential time-domain analysis to detect mistake, investigation result and the confirmation that maybe can better detect.
Programmable logic device is to analyzing from the numerical data of described imaging transmitter.Preferred embodiment adopts the FPGA (Field Programmable Gate Array) that the numerical data in imaging transmitter is operated, by typically comprising that the mode of process that selectivity extracts reduces described data.Preferred embodiment also typically programmable logic device for receiving equipment, for example sending in the command center of the signal of described imaging transmitter.The data that FPGA (Field Programmable Gate Array) in receiving equipment is typically drawn into selectivity operate to separate the situation state in reading image or understand situation about changing.
The selectivity of over-sampling space or time data extracts.The example extracting from the selectivity of the space of imaging data or the actual value of time waveform data colony may include, but not limited to intermediate value, mode value, maximal value, minimum value, standard deviation or be selected from another actual value of colony.
Selectivity extraction technique of the present invention can include but not limited to, from the actual value of described colony, from the actual value of the conversion of described colony, and described the tested value of sampling interval colony.Selectivity decimation value can be quantitative qualitatively or both have both at the same time.
Selectivity extraction technique of the present invention can computed image data approximate region (area or volume), allow from view data relatively far away seldom or not contribution.Cartographic represenation of area typical earth surface is shown in the view data of approaching target point in given image or reference location.Volume represents it is that typically detector area expanded along with the time.For example, the expression of nonlinear polarity, wherein non-linear interval is used in room and time, but the geometric representation of polarity is only applied in detector space, it is a kind of example of cartesian geometry configuration, close on data and acquisition and be illustrated in the value being drawn into about data volume (cylindrical shape) array of the space radius of impact point for optionally extracting, and along with time domain " axle " is collected.
Preferred embodiment comprises that selectivity is taken into picture transmitter.A preferred embodiment is used selectivity to extract by the peak value retention value from multiple values in sampling interval.Here the abbreviation of peak value retention value used is PeakVue, and wherein " peak " generally refers to one type extreme, as from as described in selected maximal value or minimum value sampling interval colony, " Vue " refers to selected value.Should be understood that, any term PeakVue herein mentioning can be substituted by one or more other selectivity extraction techniques.
The example of selectivity decimation value may not be the actual value from time or spatial waveforms colony, and can change the measurement that represents colony or the selectivity decimation value that carrys out the measurement of transformation into itself's colony into, may include but not limited to mean value, standard deviation, variance, kurtosis, measure of skewness, mutual relationship, frequency distribution value, histogram population value, probability density distribution value and other significant measured value.
Selectivity decimation value also can be from the combination results of statistical value above-mentioned and other calculated value.Such combination, for quality and any abnormal possible cause of acquired data, can provide important clairvoyance.For example, when to the data analysis of crossing sampling, found that it follows Gauss normal distribution, then place and higher put letter in the information of being passed on by average measurement value.Greatest differences between intermediate value and average (or similarly basic calculate) has disclosed the cause and effect deviation distributing described in distortion.
For the purpose of illustration and description, aforementioned description of the preferred embodiment of the present invention is presented.They are not intended to exhaustive or limit the invention to disclosed certain form.According to above-mentioned instruction, significantly amendment or variation are possible.Select and describe these embodiment, to make great efforts providing best deciphering as the principle of the present invention and practical application thereof, and making thus those of ordinary skill in the art can utilize various embodiment of the present invention and various amendment as the special-purpose that is suitable for expection.In the time understanding according to width range fair, legal, that enjoy liberally, in the scope of the present invention that all such modifications and modification are all determined when by appended claims.

Claims (38)

  1. Processing derive from by contact with machine induction or treatment progress in the method for dynamic measuring data of simulating signal of analog sensor generation, described method comprises:
    (a) by should contact with mechanical sense or treatment progress in analog sensor generate simulating signal;
    (b) described simulating signal is converted to the digit data stream of over-sampling;
    (c) specify the sampling interval data set in the digit data stream of described over-sampling;
    (d) analyze at least a portion of described sampling interval data set, to determine one or more data set attribute that is selected from group, described group by intermediate value, mode value, standard deviation value (SDV), maximal value, value range, minimum value, root-mean-square value (RMS), statistics scattering value, momentum value, variance yields, skewness value, kurtosis value, peak shape factor (PSF) characteristic, parameter cause and effect than (PvC) characteristic and one or more different value;
    (e) the sequential sampling interval data set of analyzing in extraction step (d), to generate one or more scalar value corresponding with each sampling interval data set;
    (f) generate the waveform that comprises the described scalar value producing in step (e); And
    (g) one or more in described one or more data set attribute that preservation is relevant to described waveform.
  2. 2. method according to claim 1, further comprise based on one or more data set attribute and analyze described waveform, so that the sign of one or more situation to be provided, described situation comprises impact condition, friction condition, fault-signal situation, noise signal situation, arc discharge situation, corona situation, contact situation, discharge scenario, mobility status, corrosion condition, cavitation erosion situation, sliding condition, the closed situation of valve ruuning situation and valve.
  3. 3. method according to claim 1, after step (b), step (c) is front, filters over-sampling digit data stream to obtain the situation information of machine or treatment progress.
  4. 4. method according to claim 1, further comprises:
    Step (d) comprises each intermediate value and the mean value of determining multiple sampling interval data centralizations, and determines the difference of described intermediate value and mean value from multiple sampling interval data centralizations;
    (h) for one or more of described sampling interval data centralization, described difference is compared with the threshold values limit determine whether described sampling interval data set contains the possibility of Gauss's normal state data or non-Gauss's normal state data, and wherein said difference exceeds the possibility that threshold limit shows non-Gauss's normal state data;
    (i) distribute the Gauss's attribute or the non-Gauss's attribute that are associated with described sampling interval data set; And
    (j) storage and one or more Gauss's attribute being associated or non-Gauss's attribute in described sampling interval data set and waveform.
  5. 5. method according to claim 4, further comprise, indicate under the situation that dynamic sampling interval data collection is divided into non-Gauss's normal state data in the step comparing (h), further decryption set attribute distribution indicate the attribute of the situation that likely causes non-Gauss's normal state Data classification, and wherein said situation is from being changed the group forming and chosen by impact, sensor fault, fault, machine operation, noise, stable case, random occurrence, system event and environmental parameter.
  6. 6. method according to claim 4, further comprise, indicate under the situation that dynamic sampling interval data collection contains non-Gauss's normal state data in the step comparing (h), for the spectral analysis of measurement data use multiple in step (d) definite intermediate value.
  7. 7. method according to claim 4, wherein, step (f) comprises using from sampling interval data centralization to be selected for generation of at least one value of Gauss's normal state and the classification of non-Gauss's normal state and generates described waveform.
  8. 8. method according to claim 1, wherein, step (b) is included in the sampling rate of nyquist frequency carries out over-sampling to dynamic measuring data.
  9. 9. method according to claim 1, further comprises, obtains the characteristic of machine or the characteristic of method for collecting machine vibration data at least partly based on one or more scalar value or data set attribute.
  10. 10. method according to claim 1, further comprises:
    Step (d) comprising:
    (d1) determine multiple standard deviation values, each standard deviation value obtains from corresponding one of the data centralization of multiple over-samplings;
    (d2) determine multiple maximal values, each maximal value obtains from corresponding one of multiple sampling interval data centralizations;
    (d3) determine multiple minimum value, each minimum value obtains from corresponding one of multiple sampling interval data centralizations;
    (d4) determine multiple the first differences, each the first difference is determined by the difference between described maximal value and the described minimum value of definite described corresponding sampling interval data set;
    (d5) determine multiple the second differences, each the second difference is determined by the difference between described standard deviation value and described first difference of definite described corresponding sampling interval data set;
    (h), by one or more described the second difference and the threshold value comparison determined, with the possibility that determines whether that described dynamic measuring data comprises cause and effect data or Gauss's normal state data, the second difference that is wherein greater than described threshold value characterizes the possibility of cause and effect data; And
    (i) characterize in the situation that described dynamic measuring data comprises cause and effect data in described comparison step (h), appointment may cause the situation of described cause and effect data, and wherein said situation changes the group forming and selects from impact, sensor fault, fault, machine operation, noise, stable case, random occurrence, system event and environmental parameter.
  11. 11. methods according to claim 1, further comprise:
    Step (f) comprises
    (f1) generate the intermediate range waveform that comprises multiple mid-range value, each in multiple mid-range value in wherein said intermediate range waveform is all selected from corresponding one of multiple sampling interval data centralizations, wherein said multiple mid-range value comprises multiple intermediate values, multiple mean value, multiple RMS values or multiple mode value;
    (f2) generate the maximum magnitude waveform that comprises multiple maximum magnitude values, each in multiple maximum magnitude values in wherein said maximum magnitude waveform is all selected from corresponding one of multiple sampling interval data centralizations, wherein said multiple maximum magnitude value comprises multiple bare maximums, maximal value after multiple adjustment, multiple minimum value or be multiplely up to minimum peak-to-peak value;
    (f3) generate the statistical straggling waveform that comprises multiple statistical straggling values, each in multiple statistical straggling values in wherein said statistical straggling waveform is all selected from corresponding one of multiple over-sampling data centralizations, wherein said multiple statistical straggling value comprises multiple variance yields, multiple measure of skewness values, or multiple kurtosis value;
    (f4) generate the maximum waveform comprising after peaked adjustment after multiple adjustment, each in the maximal value after the multiple rectifications in the maximum waveform after wherein said rectification is all selected from corresponding one of multiple sampling interval data centralizations, and
    (f5) generate in the following manner the waveform merging:
    By another the corresponding value addition in the described waveform obtaining in comprising the value of in the described waveform obtaining in step (f1) to (f2) and comprising step (f1) to (f4), or
    In the corresponding value of another in the described waveform obtaining, deduct the value of comprising in the described waveform obtaining in step (f1) to (f2) from comprise step (f1) to (f4),
    The waveform table of wherein said merging is shown in the peak value in described sampling interval data set.
  12. 12. methods according to claim 1, further comprise:
    Step (d) comprising:
    (d1) determine multiple mode value, the value or the value scope that the most frequently repeat of each mode value based on appearing in corresponding one of multiple sampling interval data centralizations;
    (d2) determine multiple minimum value, each minimum value obtains from corresponding one of multiple sampling interval data centralizations;
    (d3) determine multiple maximal values, each maximal value obtains from corresponding one of multiple sampling interval data centralizations;
    (d4) determine multiple MODE-MIN differences, each MODE-MIN difference is determined by the difference between mode value and the minimum value of definite corresponding sampling interval data set;
    (d5) determine multiple MAX-MODE differences, each MAX-MODE difference is determined by the difference between maximal value and the mode value of definite corresponding sampling interval data set;
    (h), if be less than default threshold value in the above MODE-MIN difference of continuous sampling interval data set, at least one in definite described one or more sensors has fault; And
    (i), if be less than default threshold value in the above MAX-MODE difference of continuous sampled data set, at least one in definite described one or more sensors is in saturated conditions.
  13. 13. methods according to claim 12, further comprise and generate the scalar value that comprises first set and second set, wherein said first set comprises described MODE-MIN difference, and described second set comprises the correlation attribute value that characterizes sensor fault state or sensor unfaulty conditions.
  14. 14. methods according to claim 12, further comprise and generate the scalar value that comprises first set and second set, wherein said first set comprises described MAX-MODE difference, and described second set comprises the correlation attribute value that characterizes sensor state of saturation or sensor unsaturation state.
  15. 15. methods according to claim 1, further comprise:
    Step (a) comprises according to sorting according to value from minimum to max log, distributes with the cumulative data forming after sequence; And
    Step (d) further comprises, for one or more sampling interval data sets, determine as two or more be close on the absolute intermediate value that the cumulative data after described sequence distributes or under the intermediate value of mean value of data value.
  16. 16. methods according to claim 1, further comprise:
    Step (d) comprising:
    (d1) determine multiple maximal values, each maximal value obtains from corresponding one of multiple sampling interval data centralizations;
    (d2) determine one or more kurtosis momentum values based on described multiple maximal values;
    (h) determine form factor by deducting integer 3 at least one from described kurtosis momentum value;
    (i), in the time that described form factor equals zero, determine that described over-sampling dynamic measuring data is normal distribution;
    (j) in the time that described form factor is greater than 0, determine that described over-sampling dynamic measuring data is that spike distributes, and
    (k), in the time that described form factor is less than 0, the dynamic measurement data of determining described over-sampling is flat distribution.
  17. 17. methods according to claim 1, further comprise:
    Step (d) comprising:
    (d1) determine multiple mode value, the value or the value scope that the most frequently repeat of each mode value based on appearing in corresponding one of described sampling interval data centralization;
    (d2) determine multiple intermediate values, each intermediate value obtains from corresponding one of described sampling interval data centralization;
    (d3) determine multiple MODE-MED differences, each MODE-MED difference is determined by the difference between mode value and the intermediate value of definite corresponding sampling interval data set; And
    (h), if the absolute value of one or more MODE-MED differences is less than predetermined threshold value on continuous sampling interval data set, determine the stable measurement situation that exists.
  18. 18. methods according to claim 1, further comprise:
    Step (d) comprising:
    (d1) determine multiple intermediate values, each intermediate value obtains from corresponding one of described sampling interval data centralization;
    (d2) determine multiple maximal values, each maximal value from multiple samplings according to obtaining corresponding one of data centralization;
    And
    (h) determine multiple crest factors, intermediate value and the peaked difference of the sampling interval data set of each crest factor based on corresponding are determined.
  19. 19. methods according to claim 1, further comprise:
    Step (d) comprising:
    (d1) determine multiple maximal values, each maximal value obtains from corresponding one of described sampling interval data centralization;
    (d2) determine multiple minimum value, each minimum value obtains from corresponding one of described sampling interval data centralization;
    (d3) determine multiple MAX-MIN differences, each MAX-MIN difference is determined by the difference between maximal value and the minimum value of definite corresponding sampling interval data set;
    (d4) determine multiple standard deviation values, each standard deviation value obtains from corresponding of described sampling interval data centralization;
    (d5) determine multiple SDV differences, each SDV difference is determined by the difference between standard deviation value and the MAX-MIN difference of definite corresponding sampling interval data set;
    (h), by one or more described SDV differences and threshold, to determine whether dynamic measuring data comprises the possibility of cause and effect data or Gauss's normal state data, the difference that is wherein greater than described threshold value characterizes the possibility of cause and effect data.
  20. 20. methods according to claim 1, further comprise:
    Step (d) comprising:
    (d1) determine multiple maximal values, each maximal value obtains from corresponding one of described sampling interval data centralization;
    (d2) 3 or more value before the maximal value of definite immediately one or more described sampling interval data sets;
    (d3) 3 or more value after the maximal value of definite immediately one or more described sampling interval data sets;
    (d4), for one or more described sampling interval data sets, based on described maximal value, immediately 3 before described maximal value or more value and 3 or more value after described maximal value immediately, determine peak value form factor characteristic; And
    (h), based on described peak value form factor characteristic, determine the possible causal event relevant to described maximal value.
  21. 21. methods according to claim 1, further comprise:
    Step (d) comprising:
    (d1) determine multiple maximal values, each maximal value obtains from corresponding one of described sampling interval data centralization;
    (d2) determine multiple standard deviation values, each standard deviation value obtains from corresponding of described sampling interval data centralization;
    (d3) determine multiple parameter cause and effect ratio characteristics, each parameter cause and effect ratio characteristic obtains from corresponding one of described sampling interval data centralization;
    (d4) determine multiple peak value form factor characteristics, each peak value form factor characteristic obtains from corresponding of described sampling interval data centralization;
    (h) based on described maximal value, standard deviation value, parameter cause and effect ratio characteristic and peak value form factor characteristic, determine whether to exist one or more following situations:
    Due to the tired cracked situation existing of roller bearing parts, and
    The broken teeth situation existing due to geared parts fatigue failure.
  22. 22. methods according to claim 1, further comprise:
    Step (d) comprising:
    (d1) determine multiple minimum value, each minimum value obtains from corresponding one of described sampling interval data centralization;
    (d2) determine multiple intermediate values, each intermediate value obtains from corresponding one of described sampling interval data centralization;
    (d3) determine multiple mode value, each mode value obtains from corresponding one of described sampling interval data centralization;
    (d4) determine multiple mean value, each mean value obtains from corresponding one of described sampling interval data centralization;
    (d5) determine multiple standard deviation values, each standard deviation value obtains from corresponding of described sampling interval data centralization;
    (d6) determine multiple peak value form factor characteristics, each peak value form factor characteristic obtains from corresponding of described sampling interval data centralization; And
    (h), based on described minimum value, intermediate value, mode value, mean value, standard deviation value and peak value form factor characteristic, determine whether to exist and slide friction condition due to what lack of lubrication caused.
  23. 23. methods according to claim 1, further comprise:
    Step (d) comprising:
    (d1) determine multiple intermediate values, each intermediate value obtains from corresponding one of described sampling interval data centralization;
    (d2) determine multiple mode value, each mode value obtains from corresponding one of described sampling interval data centralization;
    (d3) determine multiple mean value, each mean value obtains from corresponding one of described sampling interval data centralization;
    (d4) determine multiple standard deviation values, each standard deviation value obtains from corresponding of described sampling interval data centralization;
    (d5) determine multiple peak value form factor characteristics, each peak value form factor characteristic obtains from corresponding of described sampling interval data centralization; And
    (h) based on described intermediate value, mode value, mean value, standard deviation value and peak value form factor characteristic, determine whether to exist the even running situation causing due to proper lubrication.
  24. 24. methods according to claim 1, further comprise:
    Step (d) comprising:
    (d1) determine multiple intermediate values, each intermediate value obtains from corresponding one of described sampling interval data centralization;
    (d2) determine multiple mean value, each mean value obtains from corresponding one of described sampling interval data centralization;
    (d3) determine multiple standard deviation values, each standard deviation value obtains from corresponding of described sampling interval data centralization;
    (d4) determine multiple peak value form factor characteristics, each peak value form factor characteristic obtains from corresponding of described sampling interval data centralization; And
    (h) based on described intermediate value, mean value, standard deviation value and peak value form factor characteristic, determine whether to exist misaligned situations.
  25. 25. methods according to claim 1, further comprise:
    Step (d) comprising:
    (d1) determine multiple maximal values, each maximal value obtains from corresponding one of described sampling interval data centralization;
    (d2) determine multiple standard deviation values, each standard deviation value obtains from corresponding of described sampling interval data centralization;
    (d3) determine multiple parameter cause and effect ratio characteristics, each parameter cause and effect ratio characteristic obtains from corresponding one of described sampling interval data centralization;
    (d4) determine multiple peak value form factor characteristics, each peak value form factor characteristic obtains from corresponding of described sampling interval data centralization; And
    (h), based on described maximal value, standard deviation value, parameter cause and effect ratio characteristic and peak value form factor characteristic, determine whether to exist the subsurface fatigue crack producing due to pipeline placement process middle sleeve resonance.
  26. 26. methods according to claim 1, further comprise:
    Step (d) comprising:
    (d1) determine multiple maximal values, each maximal value obtains from corresponding one of described sampling interval data centralization;
    (d2) determine multiple mean value, each average obtains from corresponding one of described sampling interval data centralization;
    (d3) determine multiple standard deviation values, each standard deviation value obtains from corresponding of described sampling interval data centralization;
    (d4) determine multiple peak value form factor characteristics, each peak value form factor characteristic obtains from corresponding of described sampling interval data centralization; And
    (h), based on described maximal value, mean value, standard deviation value and peak value form factor characteristic, determine whether owing to loading the excessive slip-stick that causes of static friction coefficient on interface.
  27. 27. methods according to claim 1, further comprise:
    Step (d) comprises with lower at least one item:
    (d1) determine multiple minimum value, each minimum value obtains from corresponding one of described sampling interval data centralization;
    (d2) determine multiple intermediate values, each intermediate value obtains from corresponding one of described sampling interval data centralization;
    (d3) determine multiple mode value, each mode value obtains from corresponding one of described sampling interval data centralization;
    (d4) determine multiple standard deviation values, each standard deviation value obtains from corresponding of described sampling interval data centralization; And
    (d5) determine multiple peak value form factor characteristics, each peak value form factor characteristic obtains from corresponding of described sampling interval data centralization; And
    (h) at least in part based on definite value in step (d), determine whether to exist one or more following situations:
    Occur near the shelf depreciation of high voltage electric equipment; And
    Occur near the leakage situation that produces fluid turbulent of pressurized leak.
  28. 28. methods according to claim 1, further comprise:
    Step (d) comprises with lower at least one item:
    (d1) determine multiple intermediate values, each intermediate value obtains from corresponding one of described sampling interval data centralization;
    (d2) determine multiple mode value, each mode value obtains from corresponding one of described sampling interval data centralization;
    (d3) determine multiple standard deviation values, each standard deviation value obtains from corresponding of described sampling interval data centralization; And
    (d4) determine multiple peak value form factor characteristics, each peak value form factor characteristic obtains from corresponding of described sampling interval data centralization; And
    (h), based on definite value in step (d), determine whether three-phase power line exists intermittent fault situation.
  29. 29. methods according to claim 1, further comprise:
    Step (e) comprises, for one or more sampling interval data set, generates the scalar value of the dynamic measuring data in the described data set of multiple expressions;
    (h), based on described multiple scalar value, determine the characteristic value of feature, quality or the characteristic of the dynamic measuring data in the described data set of one or more instructions; And
    Step (g) comprises preserves multiple scalar value and the one or more characteristic value relevant to the identifier of data set.
  30. 30. methods according to claim 1, further comprise:
    Step (d) comprising:
    (d1) determine multiple the first statistics scalar value, each the first statistics scalar value obtains from corresponding one of described sampling interval data centralization;
    (d2) determine multiple the second statistics scalar value, each the second statistics scalar value obtains from corresponding one of described sampling interval data centralization;
    (h) one or more in scalar value based on described the first statistics, determines that machine or treatment progress are in the first state instead of the second state;
    (i) one or more in scalar value based on described the second statistics, determines that machine or treatment progress are in the second state instead of the first state.
  31. 31. 1 kinds gather the method for over-sampling dynamic measuring data on the expansion time cycle with the sample frequency of fixing, described over-sampling dynamic measuring data comprises the multiple sampling interval data sets that collect by one or more sensors that append on machine or in process, each sampling interval data set is corresponding to specific sampling interval, and described method comprises:
    (a), during the period 1 within the described expansion time cycle, use the first sampling interval 1/F sR1gather described dynamic measuring data, to collect the sample of each sampling interval data centralization the first number during the period 1; And
    (b), during the second round within the described expansion time cycle, use than the first sampling interval 1/F sR1the second sampling interval 1/F that duration is longer sR2gather described dynamic measuring data, to collect the sample of each sampling interval data centralization the second number during second round, wherein the second number of sample is greater than the first number of sample.
  32. 32. 1 kinds of acquisition and processings are by multiple methods that append to the over-sampling vibration data collecting for the vibration transducer of the physical construction of material processed, wherein said physical construction operationally sends to vibration transducer by the energy of vibration from material, wherein said over-sampling vibration data comprises multiple sampling interval data sets, wherein each sampling interval data set is corresponding to specific sampling interval, and described method comprises:
    (a) receive the vibrational energy of the first vibration transducer in described multiple vibration transducer, wherein said vibrational energy is generated and is sent to described the first vibration transducer by described physical construction by the event occurring within just processed material;
    (b), based on described vibrational energy, described the first vibration transducer generates the first vibration signal;
    (c) the first over-sampling vibration data that described in over-sampling, the first vibration signal comprises multiple the first sampling interval data sets with generation;
    (d), for multiple the first sampling interval data centralizations each, determine the free maximal value of one or more choosings, minimum value, mean value, intermediate value, mode value, standard deviation value, be up to minimum zone value, the first scalar value in the group of kurtosis value, measure of skewness value and wavelength value composition;
    (e), based on described one or more the first scalar value, determine one or more the first characteristic values that provide event type to indicate;
    (f) generation is illustrated in the very first time stamp value of the time of the vibrational energy of the described event generation of the first vibration transducer place reception;
    (g) receive the vibrational energy of the second vibration transducer in described multiple vibration transducer, wherein said vibrational energy is sent to described the second vibration transducer by described physical construction;
    (h), based on described vibrational energy, described the second vibration transducer generates the second vibration signal;
    (i) the second vibration signal described in over-sampling, to generate the second over-sampling vibration data that comprises multiple the second sampling interval data sets;
    (j), for multiple the second sampling interval data centralizations each, determine the free maximal value of one or more choosings, minimum value, mean value, intermediate value, mode value, standard deviation value, be up to minimum zone value, the second scalar value in the group of kurtosis value, measure of skewness value and wavelength value composition;
    (k), based on described one or more the second scalar value, determine one or more the second characteristic values that provide event type to indicate;
    (l) generation is illustrated in the second timestamp value of the time of the vibrational energy of the described event generation of described the second vibration transducer place reception; And
    (m) by by one or more the First Eigenvalues and one or more Second Eigenvalue comparison, identical with the event type characterizing by one or more Second Eigenvalues to determine the event type characterizing by one or more the First Eigenvalues.
  33. 33. methods according to claim 32 further comprise:
    (n) determine the mistiming between very first time stamp and the second timestamp; And
    (o), at least in part based on the described mistiming, determine the position of event with respect to the first vibration transducer and the second vibration transducer.
  34. 34. methods according to claim 32, wherein said physical construction comprises one or more in pipeline, container, flour mill, disintegrating machine, mill and Metal Cutting instrument.
  35. 35. methods according to claim 32, the group that wherein said event selects free impact event, knock-on event, cavitation erosion event, empty formation event, material movement event, dry contact fuzzy event, boundary lubrication fuzzy event, leak of liquid event, fluid turbulent event, surface corrosion event, flutter event, resonance event and unstability event to form.
  36. 36. 1 kinds use the method for one or more current sensor acquisition and processing electric electromechanics flow datas, and described method comprises:
    (a) use one or more current sensor measurement simulating motor current signal information;
    (b), to be at least the sampling rate of 10 times of line frequency, described simulating motor current signal information is converted to the dynamo-electric flow data of digitalized electric of over-sampling;
    (c) from the dynamo-electric flow data of digitalized electric of over-sampling, generate a series of sampling interval data set, each sampling interval data set is corresponding to a sampling interval;
    (d) extract described sampling interval data set to obtain the scalar value after extracting;
    (e) optionally extract described sampling interval data set, with the data set characteristic of the group based on selecting free intermediate value, kurtosis value, maximal value, minimum value, standard deviation value and peak value form factor to form, obtain corresponding attribute; And
    (f) by the scalar value after extraction definite in step (d) and in step (e) definite attribute and the characteristic of described electric electromechanics flow data connect.
  37. The device of the vibration data in 37. 1 kinds of acquisition and processing machines or treatment progress, comprising:
    Append at least one vibration transducer on machine, described at least one vibration transducer generates at least one and has maximum corresponding frequencies F mAXanalog vibration signal, wherein F mAXbe greater than the event frequency that occurs in the event in described machine or treatment progress;
    At least one analog to digital converter, is being at least F mAXat least one analog vibration signal of over-sampling in the sample frequency of 7 times, to generate multiple sampling interval data sets, wherein each sampling interval data set is corresponding to specific sampling interval;
    The abstraction module that contains multiple parallel field programmable gate array, comprising:
    Primary scene programmable gate array, receive multiple sampling interval data sets and determine the first scalar value from each sampling interval data centralization, the group that described the first scalar selects free maximal value, minimum value, intermediate value, mode value, mean value, standard deviation value, parameter cause and effect ratio, ruuning situation value or peak value form factor value to form; And
    Secondary scene programmable gate array, receive multiple sampling interval data sets and determine the second scalar value from each sampling interval data centralization, the group that described the second scalar selects free maximal value, minimum value, intermediate value, mode value, mean value, standard deviation value, parameter cause and effect ratio, ruuning situation value or peak value form factor value to form.
  38. In the time processing by one or more over-sampling dynamic measuring data that appends to the sensor collection in machine or treatment progress, avoid the method for aliasing effect for 38. 1 kinds, wherein at sampling rate F sover-sampling dynamic measuring data described in up-sampling, the execution that wherein relates to the Nonlinear Processing that extracts described over-sampling dynamic measuring data will cause aliasing effect, described method comprises:
    (a) on integer up-sampling rate N, over-sampling dynamic measuring data is carried out to up-sampling by inserting N-1 individual 0 between the adjacent data sample in over-sampling dynamic measuring data, thereby generate up-sampling data;
    (b) there is cutoff frequency F by use s/ 2 low-pass filter carries out low-pass filtering to described up-sampling data and remove any spectral image producing in step (a), thereby generates not containing F sthe up-sampling data of the low-pass filtering of/2 above spectral images;
    (c) if N<L and L>1, show to exist part resampling rate, by L-1 the sample that retains each L sample and abandon between each L sample, the up-sampling data of described low-pass filtering are carried out to down-sampling, thereby generate not containing up-sampling frequency F sthe down-sampled data of the low-pass filtering of × spectral image (N/L);
    (d) carry out the Nonlinear Processing that relates to the up-sampling data that extract described low-pass filtering, there is thereby generate the F of being aliasing in sthe data of more than/2 distortion components;
    (e) use and there is cutoff frequency F s/ 2 low-pass filter carries out filtering to the data that generate in step (d), removes F thereby generate sthe data of more than/2 alias component; And
    (f) by N-1 the sample that retains each N sample and be discarded between each N sample, the data that generate in step (e) are carried out to down-sampling, thereby generated the non-linear aftertreatment data that wherein aliasing effect is alleviated.
CN201410173043.XA 2013-04-29 2014-04-25 Selective decimation and analysis of oversampled data Active CN104121985B (en)

Applications Claiming Priority (4)

Application Number Priority Date Filing Date Title
US201361816974P 2013-04-29 2013-04-29
US61/816,974 2013-04-29
US14/252,943 US9778080B2 (en) 2013-04-29 2014-04-15 Selective decimation and analysis of oversampled data
US14/252,943 2014-04-15

Publications (2)

Publication Number Publication Date
CN104121985A true CN104121985A (en) 2014-10-29
CN104121985B CN104121985B (en) 2020-07-14

Family

ID=51767492

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201410173043.XA Active CN104121985B (en) 2013-04-29 2014-04-25 Selective decimation and analysis of oversampled data

Country Status (1)

Country Link
CN (1) CN104121985B (en)

Cited By (31)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105806904A (en) * 2016-01-04 2016-07-27 杭州电子科技大学 Device and method for detecting efficacy of anti-cancer drug, ifosfamide
CN105823810A (en) * 2016-01-04 2016-08-03 杭州电子科技大学 Detection apparatus and detection method for chemotherapy effect of paclitaxel
CN106419840A (en) * 2016-09-13 2017-02-22 广州视睿电子科技有限公司 Signal processing method and device and monitoring device
CN106873034A (en) * 2017-03-28 2017-06-20 中国地震局工程力学研究所 The method and device of asymmetric waveform in identification STRONG MOTION DATA
CN106897728A (en) * 2015-12-21 2017-06-27 腾讯科技（深圳）有限公司 Method of Sample Selection, device and system based on service monitoring system
CN107103162A (en) * 2017-05-26 2017-08-29 中国人民解放军国防科学技术大学 A kind of vibration accelerated test method and system based on Theory of The Cumulative Fatigue Damage
CN107131905A (en) * 2016-02-26 2017-09-05 高准公司 Detect two or more gauge assemblies
CN107426054A (en) * 2017-07-28 2017-12-01 郑州云海信息技术有限公司 One kind is based on QPI protocol link layers checking system and its verification method
CN107851294A (en) * 2015-07-10 2018-03-27 汉阳大学校产学协力团 The preventive maintenance device and method based on state of large-scale runtime
CN108174126A (en) * 2017-12-28 2018-06-15 北京空间机电研究所 A kind of accurate choosing method in ccd signal sampling location based on visible images
CN108415352A (en) * 2017-02-09 2018-08-17 欧姆龙株式会社 Data converter, signal transfer method and signal transduction system
CN108470070A (en) * 2018-03-29 2018-08-31 重庆大学 A kind of multivariate time series distributed data storage and restoration methods
CN108490311A (en) * 2018-03-22 2018-09-04 中国南方电网有限责任公司超高压输电公司检修试验中心 Weak impact signal extraction based on power frequency sampling and separation method
CN108509558A (en) * 2018-03-23 2018-09-07 太原理工大学 A kind of sample count audio search method that resistance rapid-curing cutback is disturbed
CN108895018A (en) * 2018-04-20 2018-11-27 浙江理工大学 A kind of device and method monitoring the development process of centrifugal blade cavitation
CN109196560A (en) * 2016-01-14 2019-01-11 棱镜传感器公司 For the measuring circuit of X-ray detector and corresponding method and x-ray imaging system
CN109582706A (en) * 2018-11-14 2019-04-05 重庆邮电大学 The neighborhood density imbalance data mixing method of sampling based on Spark big data platform
CN109586767A (en) * 2017-09-29 2019-04-05 恩智浦有限公司 Combine point-to-point signal and collision detection method
CN109754817A (en) * 2017-11-02 2019-05-14 北京三星通信技术研究有限公司 signal processing method and terminal device
CN110956800A (en) * 2018-09-27 2020-04-03 杭州海康威视系统技术有限公司 Road section traffic data preprocessing method and device and electronic equipment
CN110969260A (en) * 2019-10-22 2020-04-07 成都信息工程大学 Unbalanced data oversampling method and device and storage medium
CN110967182A (en) * 2019-11-13 2020-04-07 鞍钢集团矿业有限公司 Cone crusher vibration data acquisition and preprocessing method
CN111063183A (en) * 2019-12-25 2020-04-24 宁波三星医疗电气股份有限公司 Bluetooth-based electric energy meter statistical method and device and storage medium
CN111994087A (en) * 2020-09-02 2020-11-27 中国第一汽车股份有限公司 Driving assisting method, system, vehicle and medium
CN113146274A (en) * 2021-04-02 2021-07-23 南京航空航天大学 Vibration amplitude sensorless detection and control system and method of vibration-assisted cutting device
CN113407734A (en) * 2021-07-14 2021-09-17 重庆富民银行股份有限公司 Construction method of knowledge map system based on real-time big data
CN113552545A (en) * 2021-07-21 2021-10-26 合肥中科环光技术有限公司 Method for comparing detection result consistency of radar equipment
CN113676506A (en) * 2020-05-13 2021-11-19 计算系统有限公司 Machine spectral data compression
WO2022053540A1 (en) * 2020-09-09 2022-03-17 iNDTact GmbH Method and device for evaluating a state of an object, and computer program
CN114938707A (en) * 2019-11-15 2022-08-23 Abb瑞士股份有限公司 Method and system for processing sensor data for transmission
CN116956650A (en) * 2023-09-21 2023-10-27 沈阳航天新光集团有限公司 Matlab app designer-based attitude control power system mechanical environment test condition design system

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5633811A (en) * 1994-12-09 1997-05-27 Computational Systems, Inc. Hand held data collector and analyzer system
US5799114A (en) * 1993-05-05 1998-08-25 Liberty Technologies, Inc. System and method for stable analysis of sampled transients arbitrarily aligned with their sample points
US5895857A (en) * 1995-11-08 1999-04-20 Csi Technology, Inc. Machine fault detection using vibration signal peak detector
CN1516814A (en) * 2001-06-16 2004-07-28 ά˹�ض���Ƶ���ع����޹�˾ Method of processing data
CN1524346A (en) * 2000-08-11 2004-08-25 �ʼҷ����ֵ������޹�˾ Method and arrangement for synchronizing a sigma-delta-modulator
CN102374876A (en) * 2010-08-16 2012-03-14 Csi技术股份有限公司 Integrated vibration measurement and analysis system

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5799114A (en) * 1993-05-05 1998-08-25 Liberty Technologies, Inc. System and method for stable analysis of sampled transients arbitrarily aligned with their sample points
US5633811A (en) * 1994-12-09 1997-05-27 Computational Systems, Inc. Hand held data collector and analyzer system
US5895857A (en) * 1995-11-08 1999-04-20 Csi Technology, Inc. Machine fault detection using vibration signal peak detector
CN1524346A (en) * 2000-08-11 2004-08-25 �ʼҷ����ֵ������޹�˾ Method and arrangement for synchronizing a sigma-delta-modulator
CN1516814A (en) * 2001-06-16 2004-07-28 ά˹�ض���Ƶ���ع����޹�˾ Method of processing data
CN102374876A (en) * 2010-08-16 2012-03-14 Csi技术股份有限公司 Integrated vibration measurement and analysis system

Cited By (47)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107851294B (en) * 2015-07-10 2022-07-12 汉阳大学校产学协力团 State-based preventive maintenance device and method for large-scale operation system
CN107851294A (en) * 2015-07-10 2018-03-27 汉阳大学校产学协力团 The preventive maintenance device and method based on state of large-scale runtime
CN106897728A (en) * 2015-12-21 2017-06-27 腾讯科技（深圳）有限公司 Method of Sample Selection, device and system based on service monitoring system
CN106897728B (en) * 2015-12-21 2019-12-17 腾讯科技（深圳）有限公司 Sample selection method, device and system based on business monitoring system
CN105806904A (en) * 2016-01-04 2016-07-27 杭州电子科技大学 Device and method for detecting efficacy of anti-cancer drug, ifosfamide
CN105823810A (en) * 2016-01-04 2016-08-03 杭州电子科技大学 Detection apparatus and detection method for chemotherapy effect of paclitaxel
CN109196560A (en) * 2016-01-14 2019-01-11 棱镜传感器公司 For the measuring circuit of X-ray detector and corresponding method and x-ray imaging system
CN107131905A (en) * 2016-02-26 2017-09-05 高准公司 Detect two or more gauge assemblies
CN106419840A (en) * 2016-09-13 2017-02-22 广州视睿电子科技有限公司 Signal processing method and device and monitoring device
CN106419840B (en) * 2016-09-13 2019-06-25 广州视睿电子科技有限公司 Signal processing method, device and custodial care facility
WO2018049762A1 (en) * 2016-09-13 2018-03-22 广州视睿电子科技有限公司 Signal processing method and device and monitoring apparatus
CN108415352A (en) * 2017-02-09 2018-08-17 欧姆龙株式会社 Data converter, signal transfer method and signal transduction system
US10775762B2 (en) 2017-02-09 2020-09-15 Omron Corporation Data converter, signal transmission method, and signal transmission system
CN106873034A (en) * 2017-03-28 2017-06-20 中国地震局工程力学研究所 The method and device of asymmetric waveform in identification STRONG MOTION DATA
CN106873034B (en) * 2017-03-28 2018-09-04 中国地震局工程力学研究所 Identify the method and device of asymmetric waveform in STRONG MOTION DATA
CN107103162A (en) * 2017-05-26 2017-08-29 中国人民解放军国防科学技术大学 A kind of vibration accelerated test method and system based on Theory of The Cumulative Fatigue Damage
CN107426054A (en) * 2017-07-28 2017-12-01 郑州云海信息技术有限公司 One kind is based on QPI protocol link layers checking system and its verification method
CN107426054B (en) * 2017-07-28 2020-10-20 苏州浪潮智能科技有限公司 QPI protocol-based link layer verification system and verification method thereof
CN109586767B (en) * 2017-09-29 2021-08-06 恩智浦有限公司 Joint point-to-point signal and collision detection method
CN109586767A (en) * 2017-09-29 2019-04-05 恩智浦有限公司 Combine point-to-point signal and collision detection method
CN109754817A (en) * 2017-11-02 2019-05-14 北京三星通信技术研究有限公司 signal processing method and terminal device
CN108174126B (en) * 2017-12-28 2020-09-18 北京空间机电研究所 CCD signal sampling position accurate selection method based on visible light image
CN108174126A (en) * 2017-12-28 2018-06-15 北京空间机电研究所 A kind of accurate choosing method in ccd signal sampling location based on visible images
CN108490311A (en) * 2018-03-22 2018-09-04 中国南方电网有限责任公司超高压输电公司检修试验中心 Weak impact signal extraction based on power frequency sampling and separation method
CN108509558B (en) * 2018-03-23 2021-11-05 太原理工大学 Anti-speed-variation-interference sampling counting audio retrieval method
CN108509558A (en) * 2018-03-23 2018-09-07 太原理工大学 A kind of sample count audio search method that resistance rapid-curing cutback is disturbed
CN108470070A (en) * 2018-03-29 2018-08-31 重庆大学 A kind of multivariate time series distributed data storage and restoration methods
CN108895018A (en) * 2018-04-20 2018-11-27 浙江理工大学 A kind of device and method monitoring the development process of centrifugal blade cavitation
CN110956800B (en) * 2018-09-27 2021-07-23 杭州海康威视系统技术有限公司 Road section traffic data preprocessing method and device and electronic equipment
CN110956800A (en) * 2018-09-27 2020-04-03 杭州海康威视系统技术有限公司 Road section traffic data preprocessing method and device and electronic equipment
CN109582706A (en) * 2018-11-14 2019-04-05 重庆邮电大学 The neighborhood density imbalance data mixing method of sampling based on Spark big data platform
CN110969260A (en) * 2019-10-22 2020-04-07 成都信息工程大学 Unbalanced data oversampling method and device and storage medium
CN110967182A (en) * 2019-11-13 2020-04-07 鞍钢集团矿业有限公司 Cone crusher vibration data acquisition and preprocessing method
CN110967182B (en) * 2019-11-13 2021-07-13 鞍钢集团矿业有限公司 Cone crusher vibration data acquisition and preprocessing method
CN114938707A (en) * 2019-11-15 2022-08-23 Abb瑞士股份有限公司 Method and system for processing sensor data for transmission
CN111063183B (en) * 2019-12-25 2021-03-19 宁波三星医疗电气股份有限公司 Bluetooth-based electric energy meter statistical method and device and storage medium
CN111063183A (en) * 2019-12-25 2020-04-24 宁波三星医疗电气股份有限公司 Bluetooth-based electric energy meter statistical method and device and storage medium
CN113676506A (en) * 2020-05-13 2021-11-19 计算系统有限公司 Machine spectral data compression
CN111994087B (en) * 2020-09-02 2021-11-05 中国第一汽车股份有限公司 Driving assisting method, system, vehicle and medium
CN111994087A (en) * 2020-09-02 2020-11-27 中国第一汽车股份有限公司 Driving assisting method, system, vehicle and medium
WO2022053540A1 (en) * 2020-09-09 2022-03-17 iNDTact GmbH Method and device for evaluating a state of an object, and computer program
CN113146274A (en) * 2021-04-02 2021-07-23 南京航空航天大学 Vibration amplitude sensorless detection and control system and method of vibration-assisted cutting device
CN113407734A (en) * 2021-07-14 2021-09-17 重庆富民银行股份有限公司 Construction method of knowledge map system based on real-time big data
CN113407734B (en) * 2021-07-14 2023-05-19 重庆富民银行股份有限公司 Method for constructing knowledge graph system based on real-time big data
CN113552545A (en) * 2021-07-21 2021-10-26 合肥中科环光技术有限公司 Method for comparing detection result consistency of radar equipment
CN116956650A (en) * 2023-09-21 2023-10-27 沈阳航天新光集团有限公司 Matlab app designer-based attitude control power system mechanical environment test condition design system
CN116956650B (en) * 2023-09-21 2023-11-28 沈阳航天新光集团有限公司 Matlab app designer-based attitude control power system mechanical environment test condition design system

Also Published As

Publication number Publication date
CN104121985B (en) 2020-07-14

Similar Documents

Publication Publication Date Title
CN104121985A (en) Selective decimation and analysis of oversampled data
US10175070B2 (en) Selective decimation and analysis of oversampled data
US11573558B2 (en) Methods and systems for sensor fusion in a production line environment
US10416634B2 (en) Methods and systems for the industrial internet of things
US20200096991A1 (en) Systems and methods for self-organizing data collection based on production environment parameter
CA3177392A1 (en) Industrial digital twin systems and methods with echelons of executive, advisory and operations messaging and visualization
CA3072045A1 (en) Methods and systems for detection in an industrial internet of things data collection environment with large data sets
WO2014109645A1 (en) Method, device and computer program for monitoring an industrial control system
CN103901835A (en) Apparatus and methods to monitor and control cyclic process units in a steady plant environment
CN104937616A (en) Systems and methods for collaborating in non-destructive testing system using location information
JP2013504037A (en) Advance manufacturing monitoring and diagnostic tools
JP5234441B2 (en) Information processing apparatus, information processing system, information processing method, and information processing program
Noman et al. A scale independent flexible bearing health monitoring index based on time frequency manifold energy & entropy
Carratù et al. Earthquake early detection as an IEEE1451 transducer network trigger for urban infrastructure monitoring and protection
US20110199861A1 (en) Method and system for detecting motorized objects
Dumbović et al. Forbush decrease prediction based on remote solar observations
CN103376130A (en) Environment monitoring system based on wireless communications
Ramezani et al. Automatic boiler tube leak detection with deep bidirectional LSTM neural networks of acoustic emission signals
JP2010027076A (en) Equipment diagnostic method
AU2008285140A1 (en) Method and apparatus for detecting pedestrians
Bouroumine et al. Smart diagnosis using an on-board middleware
Kuo et al. An Enhanced Structural Damage Identification Approach Using Statistical Analysis and Optimization
US20120245867A1 (en) Method and apparatus for detection of short stress waves
None Bayesian Support Vector Machine (BSVM)
Alkhadafe Computational intelligence for fault diagnosis in gearbox systems

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant