CN114330489A - Fault diagnosis method and system for monitoring equipment - Google Patents
Fault diagnosis method and system for monitoring equipment Download PDFInfo
- Publication number
- CN114330489A CN114330489A CN202111401240.9A CN202111401240A CN114330489A CN 114330489 A CN114330489 A CN 114330489A CN 202111401240 A CN202111401240 A CN 202111401240A CN 114330489 A CN114330489 A CN 114330489A
- Authority
- CN
- China
- Prior art keywords
- sample data
- fault diagnosis
- cloud server
- data
- equipment
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 41
- 238000003745 diagnosis Methods 0.000 title claims abstract description 29
- 238000012544 monitoring process Methods 0.000 title claims abstract description 22
- 230000002159 abnormal effect Effects 0.000 claims abstract description 29
- 238000012545 processing Methods 0.000 claims abstract description 29
- 230000036541 health Effects 0.000 claims abstract description 24
- 238000010224 classification analysis Methods 0.000 claims abstract description 6
- 238000004458 analytical method Methods 0.000 claims description 50
- 238000004364 calculation method Methods 0.000 claims description 25
- 239000011159 matrix material Substances 0.000 claims description 20
- 230000008569 process Effects 0.000 claims description 12
- 238000005070 sampling Methods 0.000 claims description 10
- 238000012806 monitoring device Methods 0.000 claims description 6
- 238000001914 filtration Methods 0.000 claims description 5
- 238000000611 regression analysis Methods 0.000 claims description 5
- 238000012314 multivariate regression analysis Methods 0.000 claims description 4
- 238000012549 training Methods 0.000 abstract description 4
- 238000005457 optimization Methods 0.000 abstract description 3
- 238000007621 cluster analysis Methods 0.000 abstract description 2
- 238000010801 machine learning Methods 0.000 abstract description 2
- 238000001228 spectrum Methods 0.000 description 31
- 238000001514 detection method Methods 0.000 description 8
- 238000005516 engineering process Methods 0.000 description 6
- 230000006870 function Effects 0.000 description 6
- 238000009826 distribution Methods 0.000 description 5
- 238000012423 maintenance Methods 0.000 description 5
- 230000000737 periodic effect Effects 0.000 description 5
- 230000005540 biological transmission Effects 0.000 description 4
- 230000001419 dependent effect Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 238000010183 spectrum analysis Methods 0.000 description 4
- 230000009471 action Effects 0.000 description 3
- 238000012417 linear regression Methods 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 3
- 238000005259 measurement Methods 0.000 description 3
- 238000007405 data analysis Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000005206 flow analysis Methods 0.000 description 2
- 238000012952 Resampling Methods 0.000 description 1
- 238000012300 Sequence Analysis Methods 0.000 description 1
- 230000001133 acceleration Effects 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000001364 causal effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 238000012942 design verification Methods 0.000 description 1
- 230000005284 excitation Effects 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention discloses a monitoring equipment fault diagnosis method and a monitoring equipment fault diagnosis system, wherein embedded edge equipment combining a Risc microprocessor core and a GPU is adopted, fault indexes of vibration sensor data are directly calculated on the embedded edge equipment, and trend prediction abnormity identification is carried out by using a more complex signal processing flow and a machine learning method besides statistical characteristics. And the produced health sample data and the abnormal sample data are uploaded to the cloud end to perform sample training and intelligent fault diagnosis, so that balance and sufficient resource optimization of cloud-side cooperative computing are realized. And calculating and detecting the key characteristic value at the edge equipment to generate abnormal sample data and health sample data, and uploading all the abnormal sample data and part of the health sample data to the cloud server, so that the cloud server performs classification and cluster analysis and performs fault diagnosis.
Description
Technical Field
The invention belongs to the field of mechanical equipment vibration online monitoring and fault diagnosis, and particularly relates to a monitoring equipment fault diagnosis method and system.
Background
Intelligent nuclear power plant production equipment is commonly interconnected and increasingly uses industrial internet of things (IIoT) sensors for status awareness. Especially, 5G technology application provides connection with large bandwidth, high reliability and low time delay for industrial internet, so that high-speed data transmission to a cloud computing data center becomes possible. Meanwhile, the cost of the sensor is continuously reduced, and the range of machines and other systems for adding real-time online monitoring instruments to production process equipment is greatly increased, and the number of the machines and other systems is increased more in the next few years.
For industries like nuclear power plants where unplanned shutdown is a major loss due to machine equipment failure, there are urgent technical needs and economic benefits driving the shift from traditional planned maintenance or post-mortem maintenance to predictive maintenance. Accordingly, more and more manufacturing enterprises are working to acquire more data from their devices for condition monitoring and design verification. The continuous monitoring data acquisition can more comprehensively and deeply analyze and judge the equipment state, thereby providing decision support for design improvement, operation optimization and state maintenance. In order to extract useful information from these sensors, sufficiently efficient processing algorithms are required to perform state estimation and fault prediction. These algorithms typically use high frequency sampled acceleration or current signals. The widespread use of the internet has brought about extensive coverage and continuous data connectivity. However, for power plant applications with bulky equipment groups, local connections still present problems due to the limited bandwidth of wired or mobile connections, and continuous, high frequency sampled data streams are still not easily handled. Therefore, local processing of the data at the acquisition end is necessary and becomes more important for the increase of the data volume of the vibration-like millisecond acquisition period.
In order to overcome the limitation of bandwidth and time delay on data flow, a common solution in the current vibration monitoring device is to acquire samples in a burst mode at high frequency in a short time. Such bursts are completed in intermittent time. However, this means that for equipment operating at non-constant speeds and varying loads, the chances of acquiring data under different load conditions are high. Due to the ever-changing nature of the system excitation, the response will change accordingly. These changes may affect the final monitored characteristic values. If the data samples taken at intermittent times are too distributed under different load and operating conditions, trend prediction studies become difficult. Therefore, not only continuous data acquisition but also continuous data processing becomes more important in order to be able to extract a high-quality status index.
The prior art has the defects that:
1) continuous data acquisition and full sample transmission are limited by bandwidth, so that the number of accessible sensors is limited, and the monitoring informatization application scene of industrial equipment is limited;
2) the data quality is not high, and the data representing the fault characteristics need to be screened from the mass data, so that the data maintenance amount is large and the continuous processing is difficult;
3) since data acquisition is a non-stationary random process (ergodic feature), the accuracy of fault characterization and prediction can deviate after mathematical transformation.
Disclosure of Invention
The invention aims to overcome the defects in the prior art and provide a monitoring equipment fault diagnosis method and system. And the produced health sample data and the abnormal sample data are uploaded to the cloud end to perform sample training and intelligent fault diagnosis, so that balance and sufficient resource optimization of cloud-side cooperative computing are realized. And calculating and detecting the key characteristic value at the edge equipment to generate abnormal sample data and health sample data, and uploading all the abnormal sample data and part of the health sample data to the cloud server, so that the cloud server performs classification and cluster analysis and performs fault diagnosis.
In order to solve the problems of the prior art, the invention discloses a fault diagnosis method for monitoring equipment,
acquiring a vibration signal of monitored equipment by utilizing a pre-constructed equipment layer;
transmitting the vibration signal to a pre-constructed edge layer, and outputting abnormal sample data and health sample data;
the abnormal sample data and the health sample data are transmitted to the cloud server, the abnormal sample data and the health sample data are classified and clustered through the cloud server, known fault categories are obtained according to the classification analysis results, and unknown fault categories are obtained according to the clustering analysis results.
Further, the air conditioner is provided with a fan,
the sampling frequency of the vibration signal of the monitored equipment collected by the equipment layer is greater than 25.6kS/s, the length of the waveform block is greater than 10s, and the signal data are continuously transmitted to the edge layer.
Further, the air conditioner is provided with a fan,
and all the abnormal sample data are transmitted to the cloud server.
Further, the air conditioner is provided with a fan,
the health sample data is at least partially transmitted to a cloud server.
Further, the air conditioner is provided with a fan,
the processing process of the edge layer on the vibration signals comprises the steps of sequentially filtering, order ratio analysis and eigenvalue calculation on the vibration signals through a RISC microprocessor to obtain a multi-element eigenvalue matrix; and calculating a multivariate eigenvalue matrix through a CPU (Central processing Unit) processor, and analyzing a calculation result by using a multivariate regression analysis method so as to obtain abnormal sample data and health sample data.
Accordingly, a monitoring device fault diagnosis system, comprising: the device layer, the edge layer and the cloud server; the equipment layer is used for collecting vibration signals of monitored equipment and transmitting the vibration signals to the edge layer, the edge layer is used for processing the vibration signals, outputting abnormal sample data and health sample data and transmitting the abnormal sample data and the health sample data to the cloud server, the cloud server is used for classifying and clustering the abnormal sample data and the health sample data, known fault categories are obtained according to classification analysis results, and unknown fault categories are obtained according to clustering analysis results.
Further, the air conditioner is provided with a fan,
the edge layer comprises a plurality of edge devices, each edge device comprises a RISC microprocessor and a CPU (Central processing Unit), the RISC microprocessor is used for generating a multi-element characteristic value matrix from the vibration signal, the CPU is used for calculating the multi-element characteristic value matrix, and the calculation result is analyzed through a multi-element regression analysis method to obtain abnormal sample data and health sample data.
The invention has the following beneficial effects:
the invention combines the advanced signal processing technology with the anomaly detection and feature fusion technology based on the data driving technology, and adopts the embedded edge device architecture to realize the balance of cloud server computing and edge computing.
Drawings
FIG. 1 is a schematic block diagram of an embodiment of the present invention;
FIG. 2 is a schematic process flow diagram of an edge device according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of a step ratio analysis process according to an embodiment of the present invention.
Detailed Description
The invention is further described below with reference to the accompanying drawings. The following examples are only for illustrating the technical solutions of the present invention more clearly, and the protection scope of the present invention is not limited thereby.
The invention provides a monitoring equipment fault diagnosis system, which solves the problem of the limitation of intermittent acquisition and discontinuous analysis of vibration monitoring data by using an advanced cloud-edge cooperative computing method and edge computing equipment, and provides a cloud-edge cooperative edge computing method combining an advanced signal processing method and anomaly detection. The current vibration signal acquisition is divided into two types: 1. one is that the hardware can achieve continuous collection, and data (such as 2s, 4s, 10s and 100s) are collected according to a set time length and sent to an upper computer for processing, namely continuous collection and continuous analysis, and the data are one block by one block in a time unit but are continuous; 2. one is that the hardware collects data (1min, 5min, 1h, 8h, etc.) according to a set time interval and sends the data to the upper computer for processing, which is discontinuous collection and analysis, so that the data is discontinuous and not continuous. The first method requires high real-time computing for emergency and software. In order to ensure continuous data flow analysis processing, the transmission of all data to the central cloud processing platform cannot be realized due to the limitation of network bandwidth and the influence of analysis processing real-time performance, and balanced data analysis processing must be realized between the cloud server and the edge devices of the sensor.
As shown in fig. 1, a schematic diagram of a cloud-edge cooperative edge computing architecture is adopted, an embedded technology and an anomaly detection technology are adopted for physical connection network cloud-edge cooperative edge computing, maximum computing is performed at a position closest to sensor measurement data, and an edge device overcomes the limitation of intermittent data acquisition and realizes continuous acquisition and analysis. The framework is divided into three layers, the first layer is an equipment layer, vibration waveform data of the monitored equipment is continuously transmitted to the second layer edge layer through the Internet of things (wired network, wireless network and mobile network), and the function of the edge equipment is to calculate and analyze vibration characteristic values and detect abnormal data. One edge device can correspond to one or more monitoring devices according to the distribution of the monitoring devices, the number of measuring points and the timeliness of calculation and analysis. The third layer is a cloud platform, and the cloud server receives high-quality data uploaded by the edge device group, is responsible for training model samples with large calculation amount and complexity, and performs sample classification and clustering so as to realize fault diagnosis. In order to ensure continuous data flow analysis of the edge equipment, calculation balance is realized between the cloud and the edge, all data are not transmitted to a central cloud processing platform, and only all abnormal samples and a certain proportion of normal samples are uploaded to the cloud platform.
As shown in fig. 2, the edge computing process and the architecture of the edge device, in the present invention, the edge computing device adopts an embedded very simple instruction set microprocessor (RISC microprocessor) and a graphics processor (GPU parallel computing) architecture. RISC microprocessor (reduced instruction set microprocessor) GPU (graphics processing unit)
RISC microprocessor and GPU processor, because the design goal, different, they have aimed at two kinds of different application scenarios separately. The RSISC microprocessor requires great versatility to handle a variety of different data types, as well as logic decisions, which introduce a large number of branch jumps and interrupts. These make the internal architecture of a RISC microprocessor exceptionally complex. GPUs are faced with a clean computing environment of highly uniform type, independent of large-scale data, and need not be interrupted.
The aim of the invention is to perform the maximum calculation at the location closest to the sensor measurement data. Thereby making maximum use of the processing power of the embedded RISC microprocessor and GPU processor. The method is characterized in that the RISC microprocessor is responsible for general calculation, and the GPU processor is specially responsible for processing multiple regression analysis to realize rapid matrix vector calculation.
The general calculation of the RISC microprocessor in the edge device involves the analysis of the basic time domain statistics of the original sensor vibration waveform to the complex filter sequence analysis, forming a multi-element eigenvalue matrix. Anomalies in these feature values will then be annotated using machine learning (multiple regression analysis). To optimize the edge device computation power, different types of analysis methods are coupled to the optimal processor type (signal processing programs are deployed on the risc microprocessor according to different computation types; anomaly detection and intelligent algorithms are deployed on the GPU processor). All signal processing calculations are done on the RSISC microprocessor. The parallel computation of multiple RISC microprocessors can process the waveform data of multiple channels at the same time. And then, the anomaly detection is carried out by using a multiple regression method through a GPU processor, and the method is suitable for fast matrix vector calculation.
The different analysis flows are computed in parallel on the edge device. And maximum calculation is carried out at the position closest to the measurement data of the sensor, and the edge equipment overcomes the limitation of intermittent data acquisition and realizes continuous acquisition and analysis. The cloud edge collaborative computing process is shown as a black box (edge device) in fig. 2.
The sensor continuously collects vibration waveforms, in order to guarantee proper analysis frequency range and frequency spectrum resolution precision, the sampling frequency is larger than 25.6kS/s, the length of a waveform block is larger than 10s, and data are continuously uploaded to edge equipment. The RISC microprocessor is responsible for general signal processing and calculation, and the vibration data generates key characteristic values, and the specific process comprises the following steps:
(1) filtering according to analysis requirements, and dynamically setting an analysis frequency range and a filter type;
(2) carrying out order ratio analysis on the vibration signals with the key phases, and carrying out non-key phase order ratio analysis on the vibration signals without the key phases; as shown in fig. 3, the purpose of order ratio analysis is to realize accurate equal angle sampling, that is, every time the rotating shaft rotates once, the sampling point is consistent with the angle difference on the rotating shaft, and the position on the rotating shaft is fixed, so as to realize accurate whole period synchronous sampling. The method comprises the steps of synchronously acquiring a rotary mechanical vibration signal and a key phase pulse signal which appears once after one rotation by adopting the fixed sampling frequency, selecting a vibration order number M, carrying out 2 multiplied by M interpolation on the key phase pulse by using an interpolation filter to obtain a time sequence of an equiangular signal sample, then carrying out low-pass filtering and interpolation resampling on the vibration signal to obtain an equiangular vibration sampling signal, and finally obtaining a vibration order ratio spectrum and each order of harmonic waveform by processing the equiangular vibration sampling signal.
(3) Calculating the eigenvalue to generate a multivariate characteristic matrix, including basic analysis, namely extracting the time domain characteristics of the vibration waveform: root mean square, peak factor, impulse factor, margin factor, kurtosis factor, form factor, and skewness calculations. Advanced analysis includes: order envelope analysis, cepstrum, power spectrum. Analysis method of vibration signal:
in monitoring and fault diagnosis of equipment, vibration state monitoring of the equipment is mostly adopted, so that vibration signals are effectively analyzed, and different analysis methods are used for obtaining characteristic parameters of the vibration signals, and the method is a main measure for realizing fault diagnosis of mechanical equipment. The commonly used vibration signal analysis methods include a time domain analysis method, a frequency domain analysis method, an order tracking analysis method, an empirical mode analysis method and an envelope demodulation analysis method, and the five analysis methods are described in detail one by one below.
Time domain analysis:
the vibration time domain parameter analysis is a simple method for fault detection and diagnosis of the wind generating set, and the time domain waveform is a signal subjected to denoising processing by the DSP and contains more information. In the time domain diagnosis, parameters such as mean value, root mean square value, kurtosis value, peak value, pulse factor and margin coefficient … … are adopted to diagnose whether the transmission component has mechanical fault by monitoring whether the characteristic parameters exceed the set value. The breadth domain parameters are generally divided into 2 types of indexes with dimension and without dimension. Mean, root mean square, etc. are dimensional time domain parameters. Dimensionless time domain parameters include a skewness coefficient, a form factor, a kurtosis coefficient, a pulse factor, and a margin coefficient … …, which briefly introduce the parameters for the major implementation involved in the time domain analysis.
(1) The average value, which may also be referred to as a dc component, is used to evaluate whether the signal is stable. The central fluctuation characterizing the variation of the vibration signal is a constant component of the signal expressed as
Wherein n is the total number of sampling points; x is the number ofiA sample function representing the vibration signal.
(2) The root mean square value is obtained by squaring the signal, then solving the average value and then squaring, and is useful for irregular signals. The expression is
(3) Kurtosis is a reliable parameter which can directly embody probability density, and the absolute value of the kurtosis is larger when the distribution form deviation of a probability density function is larger.
The kurtosis value may reflect the symmetry of the probability density graph. The larger the deviation of the distribution morphology of the probability density function, the larger the absolute value of the kurtosis value.
In addition, there are several more common time domain parameters,
frequency domain analysis:
the frequency spectrum analysis of the time domain vibration signal is one of the basic methods in the currently known fault characteristic research methods, and relatively comprehensive fault information can be obtained in the frequency spectrum. In the frequency domain, analysis is mainly performed from 3 basic spectrums of an amplitude spectrum, a power spectrum and a cepstrum. The function of the frequency spectrum is used for analyzing the natural frequency and the fault frequency of the inner ring and the outer ring of the bearing in an original signal and the Whistle frequency generated by the mutual Whistle of gears of the gearbox; the function of the cepstrum is to easily obtain the periodic components in the side bands of the spectrum and to determine the location where the fault occurred.
1. Amplitude spectrum analysis
The amplitude frequency spectrum is obtained by performing a fourier transform (FFT) on a processed vibration signal of an original signal acquired by a sensor, calculating and drawing a frequency spectrum of the time-domain vibration signal, and the expression of the fourier transform is as follows:
after Fourier transform is carried out on a periodic signal, the obtained amplitude spectrum is a discrete signal, and the frequency spectrum is composed of fundamental waves and various harmonics of the signal; the non-periodic signal is transformed into a continuous signal after Fourier transform, and the signal is continuously distributed in a certain frequency range. The amplitude spectrum can represent the effective value of the harmonic frequency time domain signal, and is a linear distribution of the amplitude of each harmonic of the time domain signal along with the frequency.
2. Power spectral analysis
The power spectrum is the situation that the distribution of the signal power is shown in the frequency domain, namely the energy of the vibration signal is shown. The power spectrum comprises 2 frequency spectrums of cross power spectrum and self power spectrum, and the frequency spectrum contains the same information as the amplitude spectrum, and is clearer than the prominent frequency of the amplitude spectrum because the information is the square of the amplitude. The expression of the self-power spectrum based on the magnitude spectrum is shown below.
From the above formula, the power spectrum is actually the square of the amplitude of the time domain signal at the harmonic frequency, and thus the obtained frequency spectrum is more prominent in the main frequency.
3. Cepstrum
The cepstrum is also called secondary spectrum. It can effectively detect periodic components in complex frequency spectrum. Cepstrum is commonly used in mechanical vibration and is used more in vibration signal analysis for the purpose of fault detection and diagnosis.
The power cepstrum can be defined as a spectrum obtained by performing inverse fourier transform on a result of an operation after logarithmic operation on a power spectrum, that is:
Cx(r)＝F-1[log s(f)]
the cepstrum has the following characteristics:
(1) by cepstrum analysis, different frequency components in the signal can be identified, and periodic components important for diagnosis can be found.
(2) The cepstrum can separate harmonic and sideband components.
Envelope demodulation analysis method:
the envelope demodulation analysis means that the vibration signal collected by hardware equipment is demodulated to generate an envelope, and the envelope is subjected to Fourier transform and then subjected to spectrum analysis. The main purpose of the envelope analysis of the signal is to analyze the energy variation of the high frequency signal accordingly. The envelope spectrum is mainly used for analyzing high-frequency resonance excited by fault impact of gear box teeth and bearings and detecting early fault problems of the gear box teeth and the bearings in time. The vibration signal is filtered, the low-frequency signal is filtered, the rest part is subjected to envelope demodulation, the low-frequency modulation signal can be extracted from the high-frequency signal, and the clear low-frequency signal can be obtained after low-pass filtering. The signal analysis method can resist the interference of low-frequency signals, improve the signal-to-noise ratio and enable fault characteristic signals to be displayed more obviously, thereby providing favorable help for fault diagnosis and analysis of mechanical equipment.
The acquired vibration signals can be obtained by calling corresponding calculation formulas and algorithms.
The GPU processor is responsible for trend prediction, i.e. anomaly detection. The multivariate feature matrix generated by the RISC microprocessor is subjected to multivariate regression analysis by the GPU, a large amount of rapid matrix vector calculation is needed, and the GPU is suitable for the repeated single rapid matrix operation. Multivariate regression analysis, which is a predictive modeling technique in big data analysis, studies the relationship between dependent variables (targets) and independent variables (predictors). This technique is commonly used for predictive analysis, time series modeling, and discovering causal relationships between variables.
Linear regression analysis is modeled to predict dependent variables through one or more independent variables,
multiple linear regression:
a plurality of multivariate regression models of independent variables (x) and dependent variables y.
y＝β0+β1x1+β2x2+...+βnxn+ε＝βTx
Wherein: beta is a0，β1，β2...βnIs the parameter value and epsilon is the error value. Beta, x is a matrix (eigenvector matrix),
estimating a multiple regression equation, namely, obtaining a characteristic matrix x by calculating and analyzing the acquired vibration data samples, and then estimating beta0，β1，β2...βnIs the parameter value.
The method of estimation is to obtain a set of estimates
Parameter values such that dependent variable y and predicted value
The sum of squares of (a) is minimal.
The operation involves linear algebra and matrix algebra, is a single, repeated and heavy calculation process, and is therefore completed in the GPU.
And performing trend prediction by using a multiple regression analysis calculation result to form abnormal sample data and a healthy data sample, wherein the abnormal sample data is completely uploaded to the cloud server, and the healthy sample data can be uploaded to the cloud server in a software configuration custom ratio. The cloud server is mainly responsible for sample training, classifying and clustering the samples, and performing fault diagnosis, wherein the purpose of trend prediction is as follows: for early equipment failure, it cannot be determined, without experience, whether the data is normal or not, or is not sensitive to finding a failure, based on the characteristic quantities of the previous signal analysis. At this time, the characteristic values of each group of signal samples form a characteristic matrix to perform linear regression calculation, namely, a regression model is established according to the samples when the equipment is normal, a subsequent new sample is input into the model to calculate a target value, and if the deviation between the target value and the true value is overlarge, the group of samples are considered to be abnormal and have faults.
It is noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Also in the description of the present invention, it is to be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise", and the like, indicate orientations and positional relationships based on those shown in the drawings, and are used only for convenience of description and simplicity of description, and do not indicate or imply that the equipment or element so referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore, should not be considered as limiting the present invention. In the drawings of the present invention, the filling pattern is only for distinguishing the layers, and is not limited to any other way.
Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that changes, modifications, substitutions and alterations can be made in these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.
Claims (7)
1. A fault diagnosis method for monitoring equipment is characterized by comprising the following steps:
acquiring a vibration signal of monitored equipment by utilizing a pre-constructed equipment layer;
transmitting the vibration signal to a pre-constructed edge layer, and outputting abnormal sample data and health sample data;
the abnormal sample data and the health sample data are transmitted to the cloud server, the abnormal sample data and the health sample data are classified and clustered through the cloud server, known fault categories are obtained according to the classification analysis results, and unknown fault categories are obtained according to the clustering analysis results.
2. The fault diagnosis method for the monitoring equipment according to claim 1, characterized in that:
the sampling frequency of the vibration signal of the monitored equipment collected by the equipment layer is greater than 25.6kS/s, the length of the waveform block is greater than 10s, and the signal data are continuously transmitted to the edge layer.
3. The fault diagnosis method for the monitoring equipment according to claim 1, characterized in that:
and all the abnormal sample data are transmitted to the cloud server.
4. The fault diagnosis method for the monitoring equipment according to claim 1, characterized in that:
the health sample data is at least partially transmitted to a cloud server.
5. The fault diagnosis method for the monitoring equipment according to claim 1, characterized in that:
the processing process of the edge layer on the vibration signals comprises the steps of sequentially filtering, order ratio analysis and eigenvalue calculation on the vibration signals through a RISC microprocessor to obtain a multi-element eigenvalue matrix; and calculating a multivariate eigenvalue matrix through a CPU (Central processing Unit) processor, and analyzing a calculation result by using a multivariate regression analysis method so as to obtain abnormal sample data and health sample data.
6. A monitoring device fault diagnostic system, comprising: the device layer, the edge layer and the cloud server; the equipment layer is used for collecting vibration signals of monitored equipment and transmitting the vibration signals to the edge layer, the edge layer is used for processing the vibration signals, outputting abnormal sample data and health sample data and transmitting the abnormal sample data and the health sample data to the cloud server, the cloud server is used for classifying and clustering the abnormal sample data and the health sample data, known fault categories are obtained according to classification analysis results, and unknown fault categories are obtained according to clustering analysis results.
7. The fault diagnosis system of a monitoring device according to claim 1, wherein:
the edge layer comprises a plurality of edge devices, each edge device comprises a RISC microprocessor and a CPU (Central processing Unit), the RISC microprocessor is used for generating a multi-element characteristic value matrix from the vibration signal, the CPU is used for calculating the multi-element characteristic value matrix, and the calculation result is analyzed through a multi-element regression analysis method to obtain abnormal sample data and health sample data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111401240.9A CN114330489A (en) | 2021-11-24 | 2021-11-24 | Fault diagnosis method and system for monitoring equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111401240.9A CN114330489A (en) | 2021-11-24 | 2021-11-24 | Fault diagnosis method and system for monitoring equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114330489A true CN114330489A (en) | 2022-04-12 |
Family
ID=81046579
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111401240.9A Pending CN114330489A (en) | 2021-11-24 | 2021-11-24 | Fault diagnosis method and system for monitoring equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114330489A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115643159A (en) * | 2022-11-18 | 2023-01-24 | 奇点创新(江苏)智能科技有限公司 | Equipment abnormity early warning method and system based on edge calculation |
| CN117273547A (en) * | 2023-11-17 | 2023-12-22 | 建平慧营化工有限公司 | Production equipment operation data processing method based on edge calculation |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106991502A (en) * | 2017-04-27 | 2017-07-28 | 深圳大数点科技有限公司 | A kind of equipment fault forecasting system and method |
| CN107891199A (en) * | 2017-12-01 | 2018-04-10 | 中州大学 | Spark discharge condition checkout gear and recognition methods |
| CN110008898A (en) * | 2019-04-02 | 2019-07-12 | 中国计量大学 | Industrial equipment data edges processing method based on symbol and convolutional neural networks |
| US20200051419A1 (en) * | 2017-10-11 | 2020-02-13 | Analog Devices Global Unlimited Company | Cloud-based machine health monitoring |
| CN112601326A (en) * | 2020-11-20 | 2021-04-02 | 上海亚明照明有限公司 | Lamp maintenance device and method |
| CN113110386A (en) * | 2021-04-19 | 2021-07-13 | 重庆大学 | GIS/GIL equipment mechanical vibration state on-line monitoring and mechanical fault cloud diagnosis system |
-
2021
- 2021-11-24 CN CN202111401240.9A patent/CN114330489A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106991502A (en) * | 2017-04-27 | 2017-07-28 | 深圳大数点科技有限公司 | A kind of equipment fault forecasting system and method |
| US20200051419A1 (en) * | 2017-10-11 | 2020-02-13 | Analog Devices Global Unlimited Company | Cloud-based machine health monitoring |
| CN107891199A (en) * | 2017-12-01 | 2018-04-10 | 中州大学 | Spark discharge condition checkout gear and recognition methods |
| CN110008898A (en) * | 2019-04-02 | 2019-07-12 | 中国计量大学 | Industrial equipment data edges processing method based on symbol and convolutional neural networks |
| CN112601326A (en) * | 2020-11-20 | 2021-04-02 | 上海亚明照明有限公司 | Lamp maintenance device and method |
| CN113110386A (en) * | 2021-04-19 | 2021-07-13 | 重庆大学 | GIS/GIL equipment mechanical vibration state on-line monitoring and mechanical fault cloud diagnosis system |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115643159A (en) * | 2022-11-18 | 2023-01-24 | 奇点创新(江苏)智能科技有限公司 | Equipment abnormity early warning method and system based on edge calculation |
| CN117273547A (en) * | 2023-11-17 | 2023-12-22 | 建平慧营化工有限公司 | Production equipment operation data processing method based on edge calculation |
| CN117273547B (en) * | 2023-11-17 | 2024-01-30 | 建平慧营化工有限公司 | Production equipment operation data processing method based on edge calculation |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109460618B (en) | Rolling bearing residual life online prediction method and system | |
| CN107016235B (en) | Equipment running state health degree evaluation method based on multi-feature adaptive fusion | |
| CN114330489A (en) | Fault diagnosis method and system for monitoring equipment | |
| Wang et al. | Traversal index enhanced-gram (TIEgram): A novel optimal demodulation frequency band selection method for rolling bearing fault diagnosis under non-stationary operating conditions | |
| CN113834657B (en) | Bearing fault early warning and diagnosis method based on improved MSET and frequency spectrum characteristics | |
| CN111504645B (en) | Rolling bearing fault diagnosis method based on frequency domain multipoint kurtosis | |
| CN107886171B (en) | PMU data-based breaker state online diagnosis method and system | |
| CN111922095A (en) | Vibration diagnosis method for abnormal torsional vibration fault of roller of cold rolling mill | |
| CN110987438B (en) | Method for detecting periodical vibration impact signals of hydraulic generator in variable rotating speed process | |
| CN110224394B (en) | Fourier decomposition algorithm suitable for extracting characteristics of non-stationary power oscillation signal | |
| CN102155988A (en) | Equipment monitoring and diagnosing method | |
| CN107748209B (en) | Structural damage detection method | |
| JP2019067197A (en) | Method for detecting trouble sign | |
| CN109596349A (en) | A kind of decelerator trouble diagnostic method based on VMD and PCT | |
| CN110398362B (en) | Robot RV reducer fault diagnosis and positioning method | |
| Liu et al. | An online bearing fault diagnosis technique via improved demodulation spectrum analysis under variable speed conditions | |
| CN106951359A (en) | A kind of system health degree determination method and device | |
| CN110056640A (en) | Speed reducer wireless malfunction diagnostic method based on acceleration signal and edge calculations | |
| CN117270514A (en) | Production process whole-flow fault detection method based on industrial Internet of things | |
| CN114048879A (en) | Fault detection method, device, system, computer equipment and readable storage medium | |
| CN114486252B (en) | Rolling bearing fault diagnosis method of vector mode maximum envelope | |
| Ferracuti et al. | MSPCA with KDE thresholding to support QC in electrical motors production line | |
| Deng et al. | A fractional hilbert transform order optimization algorithm based DE for bearing health monitoring | |
| CN115310490A (en) | Rotating equipment fault analysis method based on multi-domain feature and sensitive feature selection | |
| CN115238726A (en) | Rotating machinery state monitoring method based on performance degradation space fluctuation evaluation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |