WO2023226213A1

WO2023226213A1 - Device fault detection method based on baseline data space

Info

Publication number: WO2023226213A1
Application number: PCT/CN2022/113629
Authority: WO
Inventors: 王红星; 王新梦; 李海龙; 姜一博
Original assignee: 山东瑞美油气装备技术创新中心有限公司; 烟台杰瑞石油装备技术有限公司
Priority date: 2022-05-24
Filing date: 2022-08-19
Publication date: 2023-11-30
Also published as: CN114996923A

Abstract

Provided in the present invention are a device fault detection method based on a baseline data space. The method comprises: determining a working condition parameter, which has a preset correlation with a working condition of a device; according to a fixed-length time period of the working condition parameter of the device, discretizing a set of original data of the acquired working condition parameter of the device, and generating a set of discrete working condition parameters that include a plurality of pieces of fixed-length time period data; according to the set of discrete working condition parameters and acquired signal data of the device in a normal state, constructing a baseline frequency domain data space of the device in a normal state, and setting a fault determination threshold value; and calculating an over-limit difference value assessment index, comparing the over-limit difference value assessment index with the fault determination threshold value, and determining whether the device has a fault. In the present application, a complex working condition is decomposed and discretized, so as to reduce the influence of multiple working conditions on fault feature extraction, fault identification, etc., and a fault is visually responded to by means of an over-limit difference value space, such that related service personnel can understand and grasp fault information and the deterioration degree of the fault more easily

Description

Equipment fault detection method based on baseline data space

References to related applications

This application claims the rights and interests of Chinese Patent Application No. 202210570497.5 submitted to the State Intellectual Property Office of the People's Republic of China on May 24, 2022, the entire content of which is hereby incorporated by reference in its entirety.

Technical field

The present application relates to the field of mechanical equipment fault detection, and more specifically, to an equipment fault detection method based on baseline data space.

Background technique

At present, there is extensive research on fault detection of mechanical equipment. Universities, scientific research institutions, and major industrial units are all interested in this direction, and there are many related studies. However, there are relatively few intelligent means for actual industrial implementation and are not widespread. It can be seen that intelligent detection of equipment faults is still in the theoretical research stage, and there is still a long way to go before widespread industrial implementation. Intelligent equipment fault detection and prediction are currently mainly affected by many factors, such as: lack of effective data accumulation in actual production, lack of fault data accumulation, complex multi-source signal overlapping coverage of industrial production site environmental conditions, widely used neural networks and other methods whose results can be explained Poor sex is not suitable for field engineers to understand and approve, etc.

The development of intelligent equipment fault detection is currently limited by many influencing factors, such as: complex on-site working conditions make it difficult to extract fault features, poor quality of effective data accumulation affects model prediction results, and model detection prediction results have poor interpretability and low accuracy resulting in correlation The business engineer cannot directly understand the cause and basis of the alarm...

Contents of the invention

In order to solve the above problems, the present invention provides an equipment fault detection method based on baseline data space. The detection method can disassemble and discretize complex working conditions, and reduce the impact of multiple working conditions on fault feature extraction, fault identification, etc. Moreover, by intuitively responding to faults through the over-limit difference space, it is easier for relevant business personnel to understand and grasp fault information and fault degradation degree.

In order to achieve the above purpose, the equipment fault detection method based on the baseline data space according to the present application includes: determining operating condition parameters that have a predetermined correlation with the operating conditions of the equipment; and a fixed time period according to the operating condition parameters of the equipment Discretize the obtained original data set of operating condition parameters of the equipment to generate a discrete operating condition parameter set containing multiple fixed time period data; according to the discrete operating condition parameter set and the obtained equipment normal signal data in the normal state, construct the baseline frequency domain data space of the equipment in the normal state, and set the fault judgment threshold; and calculate the over-limit difference evaluation index, and compare the over-limit difference evaluation index with the fault judgment The threshold is compared to determine whether the device is faulty.

Further, discretizing the original data set according to the fixed time period of the working condition parameters of the equipment and generating a discrete working condition parameter set containing multiple fixed time period data includes: based on the original data Set, lag a time point, and obtain a new data set at the lag time point, where the new data set is the same length as the original data set; calculate the difference sequence between the new data set and the original data set, and generate Set the difference sequence, set the error tolerance threshold σ, and set the data value of the difference sequence <σ to zero; and set the time length threshold t_lim, intercept the truncated data where the difference sequence is continuously zero for a duration >t_lim, and generate all The set of discrete working condition parameters.

Further, constructing a baseline frequency domain data space under the normal state of the equipment based on the discrete working condition parameter set and the acquired signal data includes: obtaining a baseline time domain data set and A baseline frequency domain data set; performing bandpass filtering on the baseline frequency domain data set to generate a second baseline frequency domain data set corresponding to the operating condition parameters in the frequency band of interest; and constructing the second baseline frequency domain data set based on the second baseline frequency domain data set Described baseline frequency domain data space.

Further, obtaining a baseline time domain data set and a baseline frequency domain data set according to the discrete working condition parameter set includes: discretizing and segmenting the signal data corresponding to the time node in the signal data in the normal state, and Generate the baseline time domain data set based on the parameter values of the working condition parameters corresponding to the time period in which the segmented data marks are located; and perform fast Fourier transform on the baseline time domain data set to obtain the baseline Frequency domain data set.

Further, constructing the baseline frequency domain data space based on the second baseline frequency domain data set includes: expanding each segment of spectrum data in the second baseline frequency domain data set according to the frequency dimension one spectrum at a time to obtain a frequency matrix; Calculate the statistical index values in columns of the frequency matrix to obtain the baseline frequency domain upper limit space; perform Hilbert transform on the baseline frequency domain upper limit space and find the upper envelope to obtain the baseline frequency domain data space.

Further, the statistical index values include: mean, median, maximum value and extreme value.

Further, the equipment fault detection method further includes: based on the baseline frequency domain data space, according to the parameter values of the target unknown working condition parameters of the equipment, using an interpolation algorithm to obtain the interpolation value under the target unknown working condition parameters. Baseline frequency domain data space.

Further, the equipment fault detection method also includes: setting an evaluation index threshold of the interpolation baseline frequency domain data space; calculating the degree of separation between the baseline frequency domain data space and the interpolation baseline frequency domain data space; The degree of separation is compared with the evaluation index threshold to determine whether the interpolation baseline frequency domain data space is suitable.

Further, the equipment fault detection method further includes: based on the comparison result between the separation degree and the evaluation index threshold, determining that the interpolation baseline frequency domain data space is appropriate, and adding the interpolation baseline frequency domain data space to all the interpolation baseline frequency domain data spaces. The baseline frequency domain data space is used as a part of the baseline frequency domain data space; or based on the comparison result between the separation degree and the evaluation index threshold, it is determined that the interpolation baseline frequency domain data space is inappropriate, and the target is supplemented The parameter value of the unknown working condition parameter is close to the signal data under the parameter value; and based on the baseline frequency domain data space and the supplemented signal data, an interpolation algorithm is used to obtain the interpolated baseline frequency domain data under the target unknown working condition parameter. space.

Further, the working condition parameters are parameters that are not highly correlated with the spectrum structure distribution. Based on the baseline frequency domain data space and the parameter values of the target unknown working condition parameters of the equipment, an interpolation algorithm is used to obtain the The interpolation baseline frequency domain data space under the target unknown working condition parameters includes: superposing the baseline frequency domain data spaces in sequence to construct a joint distribution matrix; using the interpolation algorithm according to the parameter values of the target unknown working condition parameters of the equipment. , perform interpolation calculation on the joint distribution matrix to obtain the interpolated baseline frequency domain data space under the target unknown working condition parameters.

Further, the working condition parameters are parameters that are highly correlated with the spectrum structure distribution. Based on the baseline frequency domain data space, according to the parameter values of the target unknown working condition parameters of the equipment, an interpolation algorithm is used to obtain the target. Interpolating the baseline frequency domain data space under unknown working condition parameters includes: superposing the baseline frequency domain data space in sequence to construct a joint distribution matrix; transforming the frequency values in the joint distribution matrix into order values to obtain the order The joint distribution matrix in the second dimension; according to the parameter value of the target unknown working condition parameter of the equipment, use the interpolation algorithm to perform interpolation calculation on the joint distribution matrix in the second order dimension to obtain the interpolation value under the target unknown working condition parameter. Baseline data order space; perform inverse transformation on the order value of the interpolation baseline data order space under the target unknown working condition parameters, restore it back to the frequency value, and obtain the interpolated baseline frequency domain data under the target unknown working condition parameters. space.

Further, the degree of separation is the sum of absolute values of amplitude differences between the baseline frequency domain data space and the interpolation baseline frequency domain space.

Further, the equipment fault detection method also includes: performing a cleaning process on the acquired original data set to supplement missing values and remove zero values, and use the cleaned data set as the original data set.

Further, the missing value is supplemented by averaging the nearest neighbor data before and after the position of the missing value.

Further, the equipment fault detection method further includes: obtaining signal data of the normal state of the equipment, where the signal data includes vibration signal data.

further,

Over-limit difference evaluation index = sum (abs (baseline upper limit data of baseline frequency domain data space - data to be detected)),

Among them, abs (baseline upper limit data of the baseline frequency domain data space - data to be detected) means finding the absolute value of the difference between the baseline upper limit data of the baseline frequency domain data space and the data to be detected, sum (abs (baseline of the baseline frequency domain data space) Upper limit data - data to be detected)) represents the sum of the absolute values of the differences between the baseline upper limit data and the data to be detected in the baseline frequency domain data space.

According to another aspect of the present application, a computer device is provided, including a memory and a processor. The memory stores a computer program that can be run on the processor. When the processor executes the computer program, the above-mentioned baseline-based implementation is implemented. Steps of device failure detection method in data space.

According to another aspect of the present application, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the above-mentioned equipment fault detection method based on the baseline data space is implemented.

According to the equipment fault detection method of this application, by discretizing, processing and storing historical working conditions according to strong correlation parameters, it is conducive to refining fault detection scenarios and limiting the impact of different working condition factors on the effectiveness of fault detection results. , simplifying the fault detection conditions and steps for each working condition. Moreover, by setting the over-limit difference evaluation index, the fault can be responded to intuitively, making it easier for relevant business personnel to understand and grasp the fault information and fault degradation degree.

Moreover, according to the equipment fault detection method of this application, the original data of historical working condition parameters is truncated by using a difference sequence with equal step size, so that the discretized working condition parameter data and the corresponding time period can be quickly and accurately obtained.

Moreover, according to the equipment fault detection method of the present application, based on the historical discrete working condition parameters, the original baseline data vibration data set (baseline time domain data set) and the baseline data vibration frequency of the equipment's normal operation under different discrete working condition parameters are constructed. Domain data set (baseline frequency domain data set) and baseline data space provide reliable evaluation standards and basis methods for equipment fault detection.

In addition, based on the baseline data space under historical limited discrete working condition parameters, the interpolation fitting of the unknown working condition baseline data space is automatically generated, enriching and improving the discrete working condition and baseline data space, and avoiding the lack of historical working conditions affecting the baseline data space. Limitations on performing fault detection.

Description of the drawings

The description and drawings that constitute a part of this application are used to provide a further understanding of the present invention. The illustrative embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the attached picture:

Figure 1 shows a flow chart of a device fault detection method based on baseline data space according to an embodiment of the present application;

Figure 2 shows a flow chart of a device fault detection method based on baseline data space according to a preferred embodiment of the present application.

Detailed ways

In order to make the purpose, technical solutions and advantages of the present application more clear, the present application will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present application and are not used to limit the present application.

This application provides an equipment fault detection method based on the baseline data space, which constructs a baseline data space based on the strong correlation parameter data of the equipment's normal operating conditions and vibration signal time domain and frequency domain data.

According to this application, as shown in Figure 1, an equipment fault detection method based on the baseline data space is provided. The method includes: determining operating condition parameters that have a predetermined correlation with the operating conditions of the equipment (S101); according to the operating conditions of the equipment The fixed time period of the parameters discretizes the original data set of the obtained working condition parameters of the equipment to generate a discrete working condition parameter set containing multiple fixed time period data (S102); according to the discrete working condition parameter set and the obtained The signal data of the equipment in the normal state is constructed, the baseline frequency domain data space of the equipment in the normal state is constructed, and the fault judgment threshold is set (S103); and the over-limit difference evaluation index is calculated, and the over-limit difference evaluation index is combined with the fault judgment Compare the thresholds to determine whether the device is faulty (S104).

According to the equipment fault detection method of this application, by disassembling and discretizing complex working conditions, the impact of multiple working conditions on fault feature extraction, fault identification, etc. can be reduced, and faults can be intuitively responded to through the over-limit difference space, and more It is easy for relevant business personnel to understand and grasp fault information and fault deterioration degree.

According to this application, as shown in Figure 2, the following description mainly involves: discrete working condition parameter processing and standardized mining (discretizing working conditions and constructing baseline data space according to working conditions), construction of baseline data space, unknown working conditions Baseline data space interpolation fitting is automatically generated and baseline data space fault detection effectiveness is verified. The above-mentioned modules include specific construction steps of each module, and the correlation and combination of each module can generate a baseline data spatial database under each discrete working condition.

The baseline data space data under complex discrete working conditions can directly calculate the over-limit difference evaluation index for the equipment operation data under the same working condition, and based on the over-limit difference evaluation index level, determine whether the current status of the equipment is in a fault state. According to this application, on the one hand, the baseline data space collects the basic vibration data information of the equipment's normal operation under various historical discrete working conditions, and on the other hand, it can interpolate and fit the corresponding baseline data space for the current new unknown working conditions. Therefore, the method of this application is not limited by the number of historical working conditions, and reflects the equipment fault status through the over-limit difference evaluation index, which is relatively intuitive and easier for field operators or fault diagnosis engineers to understand the basis and fault status of the equipment.

Below, the equipment fault detection method based on the baseline data space is explained in detail.

1. Discrete working condition parameter processing and standardized mining

Determine the working condition parameters that are highly relevant to the complex working conditions of the equipment.

The working conditions are quantified and discretized according to the determined parameters, that is, the values of equipment operation-related parameters that are strongly related to the working conditions are discretized.

Taking the fracturing truck-mounted plunger pump at the fracturing well site as an example, the parameters that affect the operating conditions of the fracturing pump include: pressure, rotation speed, sand ratio... Among the above parameters, the main influencing factor is rotation speed. According to this application, a baseline data space is constructed using fracturing pump reduction gearbox failure detection as a scenario to detect gearbox equipment failure, and the discrete working condition parameters (i.e., the working condition parameters to be discretized , hereafter referred to as the discrete working condition parameter) is the driving motor or engine speed, that is, the working condition parameter is determined to be the rotating speed.

Obtain the original data collection of the equipment's working condition parameters.

According to an embodiment of the present application, the original data set JDS={(time _i , speed _i )} of main relevant operating parameters of the equipment operating conditions is obtained, where i=0,1,2,...,n. time _i is the time data column of the original data of the parameters affected by the equipment working conditions, speed _i is the original data column of the parameters affected by the equipment working conditions. Taking the fracturing pump as an example, the main influencing parameters of the equipment working conditions are the drive motor or engine speed, speed _i is the pressure Crack pump motor or engine speed.

According to an embodiment of the present application, JDS data cleaning of the original data set is performed, that is, missing values are supplemented and zero values are eliminated on (time _i , speed _i ) data. According to an embodiment of the present application, missing values are mainly supplemented by averaging the nearest neighbor data before and after the missing value position, and zero-valued data are directly eliminated. Taking the fracturing pump as an example, the main purpose is to supplement the missing value mean and eliminate the zero value of the data for the speed _i . And, use the cleaned data set as the original data set.

According to an embodiment of the present application, based on the original data set JDS, the tim _i column is lagged by one time point granularity, and the new JDS data of the lagged time point is obtained JDS_new={(time_new _i+1 , speed_new _i+1 )}, And delete the row where the last time point data of the original data set JDS is located, so that JDS and JDS_new are equal in length, and the data in the same row differs by one time point. The new JDS data JDS_new is data that lags behind the original data set JDS by one time point. For example, the original data set JDS starts from 1s, and the new JDS data JDS_new starts from a time point after the original data set JDS, such as 2s.

Calculate the difference sequence speed_diff _i+1 ₌ speed_new _i+1 -speed _i for the speed_new i+1 column and speed _i column of JDS_new and JDS, and generate the difference sequence data set JDS_diff={(time_new _i+1 , speed_diff _i+1 )}, and set the fault tolerance threshold σ (generally <5, which can be set according to the specific situation), and set the data value of speed_diff _i+1 <σ to zero.

Starting from time_new ₁ of the difference sequence data set JDS_diff, perform speed_diff _i+1 continuous data traversal, set the discrete working condition parameter stable time length threshold t_lim (generally >10min, can be set according to the specific situation), intercept speed_diff _i+1 The truncated data start time time_s _k and end time time_e _k with continuous zero duration >t_lim and the working condition fixed parameter speed _k corresponding to the current time period are generated to generate a discrete working condition parameter set JDS _resut ={(time_s _k ,time_e _k ,speed _k )}, realize the discretization of the original working conditions according to the fixed time period according to the working condition parameters, where,

And k is the number of fixed-time periods, len(JDS_diff) represents the total length of the JDS_diff data set, that is, the total time length of the data set, time_s _k represents the current speed start time, time_e _k represents the current speed end time, speed _k represents the current rotation speed. And use JDS _resut according to the table structure and field naming to create structured database tables and store data. The database type is not limited.

According to this application, the original data set is discretized according to the fixed time period of the equipment's working condition parameters, and a discrete working condition parameter set containing multiple fixed time period data is generated.

2. Build a baseline data space

Obtain the signal data of the device in normal state, which includes vibration signal data.

According to an embodiment of the present application, the vibration signal source data V _signal ={AI1, AI2,...,AIi}, AIi=(AI _time , AI _signal ) collected by the multi-channel vibration sensor at the working target position under the normal state of the equipment is obtained. Among them, i=1,2,3,…, AI _time represents the vibration signal collection time of the channel, and AI _signal represents the original collection data of the vibration signal of the channel.

Taking the fracturing pump reduction box as an example, a total of five channels of vibration sensors are installed. The installation positions are: reduction box input side H, reduction box input opposite side V, reduction box parallel stage input opposite side axial direction, reduction box parallel stage The big-end input side H and the planetary housing H of the reduction gearbox. The following is a description of the process of constructing, verifying and interpolating the entire baseline data space using a single channel. The calculation process for other channels is the same and can be fully referred to.

Of course, you can also use the existing vibration signal data set in the normal state of the device.

According to the obtained discrete working condition parameter set, the baseline time domain data set and the baseline frequency domain data set are obtained.

According to an embodiment of the present application, according to each time period in the discrete working condition parameter set, it is determined whether there is original vibration signal collection data in the time period. If not, skip the working condition parameter time period; if so, follow the JDS _resut Each discrete working condition parameter in the V _signal records the start time and end time in the information data in detail. The discrete time period vibration signal corresponding to the time node under the discrete working condition parameter is discretized for a certain channel data AIi in the V signal, and each segment is segmented. The parameter value of the discrete working condition parameter corresponding to the time period where the segment data mark is located, and the discrete baseline time domain data set AIi _k = (AI _{time_k} , AI _{signal_k} , AI _{speed_k} ) is obtained, where AI _{time_k} represents the vibration signal collection time, AI _{signal_k} represents the baseline frequency domain amplitude, AI _{speed_k} represents the parameter value of the discrete operating condition parameter corresponding to the baseline time domain data set, and AIi _k is named according to the table structure and fields to create a structured database table and store data. The database type does not matter. limit. Taking the fracturing pump as an example, the intercepted AIi _k data set is the original data of the vibration signal in the time period corresponding to each discrete fixed speed.

For the vibration data of each discrete working condition parameter time period of a certain channel, fast Fourier transform is performed sequentially according to the fixed number of spectral lines fr and the fixed number of sampling points fd (the specific values can be customized) to obtain the baseline frequency domain data set Fre_AIi _k = (AI _{fre_k} , AI _{signal_k} , AI _{speed_k} ). Among them, AI _{fre_k} represents the baseline frequency domain frequency value, AI _{signal_k} represents the baseline frequency domain amplitude, and AI _{speed_k} represents the parameter value of the discrete working condition parameter corresponding to the baseline frequency domain data set. And _{Fre_AIik} will be named according to the table structure and fields to create a structured database table and store data. The database type is not limited. Taking the fracturing pump as an example, set the fixed number of spectral lines fr = 12800 and the fixed number of sampling points fd = 51200. The intercepted Fre_AIi _k data set is the baseline frequency domain data set corresponding to the time period of each discrete fixed speed.

Band-pass filter the obtained baseline frequency domain data set to generate a second baseline frequency domain data set corresponding to the operating condition parameters of the frequency band of interest.

According to an embodiment of the present application, a band-pass filter with upper and lower frequency limits [fre _start , fre _end ] is set, the above upper and lower limit data are customized according to the analysis requirements, and the obtained baseline frequency domain data set _{Fre_AIik} is performed Band-pass filtering is performed to generate the baseline frequency domain data set Fre′_AIi _k corresponding to the discrete parameters of the frequency band of interest, that is, the second baseline frequency domain data set.

A baseline frequency domain data space is constructed based on the second baseline frequency domain data set.

According to an embodiment of the present application, traverse k = 1, 2, 3, ..., and obtain each segment of spectrum data in the second baseline frequency domain data set Fre′_AIi _k , that is, the two-dimensional list Fre_AIi′ _k = (AI′ _{fre_k} ,AI′ _{signal_k} ) is expanded spectrum by spectrum according to the frequency dimension to obtain the frequency matrix, as shown below:

Among them, each row in the above matrix is a complete spectrum in Fre_AIi′ _k , the frequency range is 0 ~ m hz, and the number of rows is the number of sampling time periods of the signal vibration data corresponding to the fixed number of sampling points fd.

According to an embodiment of the present application, Fre_AIi′ _k can be obtained by calculating a certain statistical index by column of the above matrix (which can be any of the following indicators but is not limited to: mean, median, maximum value, extreme value, etc.) The baseline data upper limit waveform (baseline frequency domain upper line space), that is

AI' _{fre_k} = (fre_k ₀ , fre_k ₁ ,..., fre_k _m ). in,

represents the upper limit frequency domain amplitude of the baseline data under different discrete parameters, and AI′ _{fre_k} is the upper limit frequency domain frequency value of the baseline data under different discrete parameters.

For example, according to an embodiment of the present application, the median is used as a statistical index,

Respectively, find the median by column for the Fre_AIi′ _k matrix.

According to another embodiment of the present application, the maximum value is used as the statistical index,

Respectively, find the maximum value by column for the Fre_AIi′ _k matrix.

right

in baseline cap data

Perform Hilbert transformation and obtain its upper envelope to obtain the baseline frequency domain data space (i.e., baseline data space) Fre_hil_AIi′ _k .

Set the fault judgment threshold γ, calculate the over-limit difference evaluation index, and compare the over-limit difference evaluation index with the fault judgment threshold to determine whether the equipment is faulty, where,

When the over-limit difference evaluation index is less than or equal to the fault judgment threshold, the data to be tested belongs to the normal operating state; when the over-limit difference evaluation index is greater than the fault judgment threshold, the data to be tested belongs to the fault state.

According to an embodiment of the present application, based on the baseline frequency domain data space Fre_hil_AIi′ _k obtained as above, the over-limit difference evaluation index DIFF of the data to be detected and the vibration signal baseline data space under the same discrete working condition parameters is constructed to evaluate the data to be tested. Whether the data is in a faulty state, i.e.:

in,

Indicates the frequency domain data amplitude of the data to be detected. When DIFF≤γ, the data to be tested is in a normal operating state; when DIFF>γ, the data to be measured is in a fault state.

Therefore, the above method can be relatively intuitive and easier for field workers or fault diagnosis engineers to understand the basis and fault status of equipment faults.

Moreover, according to this application, historical working conditions are discretized, processed and stored based on strong correlation parameters, which is conducive to refining fault detection scenarios, limiting the impact of different working condition factors on the effectiveness of fault detection results, and simplifying each process. Condition fault detection conditions and steps; and, using difference sequence and equal step size to truncate the original data of historical working condition parameters, the discretized working condition parameter data and the corresponding time period can be quickly and accurately obtained; and based on each historical discrete working condition parameter, a The baseline data vibration time domain data set, baseline data vibration frequency domain data set, and baseline frequency domain data space of the normal operation of equipment under different discrete working conditions parameters provide reliable evaluation standards and basis methods for equipment fault detection.

When the number of historical operating conditions is limited and the obtained signal data is insufficient, the already obtained baseline data space can be used to generate more data to improve or supplement the historical operating condition signal data.

According to a preferred embodiment of the present application, based on the obtained baseline frequency domain data space and the parameter values of the target unknown working condition parameters of the equipment, an interpolation algorithm is used to obtain the interpolated baseline frequency domain data space under the target unknown working condition parameters.

The method of generating the unknown working condition baseline data space based on the obtained baseline frequency domain data space is described below by way of an embodiment.

3. Automatic generation of spatial interpolation fitting of unknown working condition baseline data

Traverse k=1,2,3,..., and Fre_hil_AIi′ _k single-channel baseline frequency domain data space, that is, a two-dimensional list

Overlay in turn to construct a 2*k-dimensional joint distribution data frame (joint distribution matrix), as shown below:

Because parameters that are highly correlated with the spectrum structure distribution, such as rotational speed, affect the frequency, they cannot be directly interpolated in the baseline frequency domain data space, while the operating condition parameters are other parameters that are not highly correlated with the spectrum structure distribution. They can be directly fitted in the baseline frequency domain data space. Interpolation fitting is performed on the baseline frequency domain data space. If the discrete operating condition parameter in the above step is the rotation speed, it is necessary to transform the frequency value to the order value. If it is other parameters that are not highly related to the spectrum structure distribution, it does not need to be performed. Conversion of frequency values to order values.

The transformation step from frequency value to order value is mainly based _on the joint _distribution data frame (joint _distribution matrix) Fre_hil2_AIi′ _k , for AI′ _{fre_i} (where i =0,1,2,…,k), perform the following calculation, that is, convert the frequency value into an order value, the calculation formula is as follows:

Among them, AI′ _{JC_i} represents each order value of the baseline data space under a single constant rotation speed. Therefore, the joint distribution data frame (joint distribution matrix) Fre_hil2_AIi′ _k can be transformed into a joint distribution data frame (joint distribution matrix) under the order dimension, that is

For Fre_hil2_AIi _′ _k or JC_mean_AIi′ (when the discrete working condition parameter is speed), set the interpolation resolution to H (this resolution can Customized according to the situation), use relevant interpolation algorithms (interpolation algorithms can include but are not limited to multi-spline interpolation algorithms, Lagrangian interpolation algorithms, etc.) according to each frequency dimension AI′ _{fre_i} or each order dimension AI′ _{JC_i} to speed _x Interpolation calculation is performed on each frequency amplitude or each order under the target to obtain the baseline data frequency domain space Fre_hil_AIi′ _X or the baseline data order space JC_mean_AIx′ under the unknown working condition parameters of the target.

If the discrete operating condition parameter is rotational speed, the conversion operation from order value to frequency value needs to be performed. If it is other operating condition parameters that are not closely related to the spectrum structure distribution, the conversion operation from order value to frequency value can be skipped.

The transformation operation from the order value to the frequency value mainly uses the following formula to inversely transform the order value of the baseline data under the interpolation target working condition parameters, restore it back to the frequency value, and obtain the baseline frequency domain data space under the interpolation target working condition parameters. Fre_hil_AIi′ _X , the specific calculation formula is as follows:

Therefore, according to the above method, when the discrete operating condition parameter is the rotation speed, it is necessary to perform the transformation step from the frequency value to the order value for AI′ _{fre_i} in the obtained Fre_hil2_AIi′ _k , and the joint distribution data frame (joint Distribution matrix) is interpolated to obtain the baseline data order space JC_mean_AIx′ under the unknown working condition parameters of the target, and the order value of the baseline data order space JC_mean_AIx′ obtained after interpolation is transformed from the order value to the frequency value. , to obtain the interpolated baseline data frequency domain space Fre_hil_AIi′ _X.

When the discrete operating condition parameters are other parameters that are not highly related to the spectrum structure distribution, there is no need to perform the conversion step from frequency value to order value and interpolate the baseline data for AI′ _{fre_i} in the obtained Fre_hil2_AIi′ _k The order value of the order space JC_mean_AIx′ performs the transformation operation from the order value to the frequency value, and Fre_hil2_AIi′ _k is interpolated to obtain the interpolated baseline data frequency domain space Fre_hil_AIi′ _X .

The interpolation effect is judged by calculating the separation index (the sum of the absolute values of the amplitude differences sum_diff) corresponding to the frequency in the original baseline data space and the interpolated baseline frequency domain data space.

Specifically, the evaluation index threshold β is set. When sum _diff ≤ β (β is customized based on experience), the interpolation effect is considered to be better; otherwise, the interpolation effect is poor.

If the interpolation effect is appropriate or good, the interpolated baseline frequency domain data space is added to the existing baseline frequency domain data space and used as a part of the existing baseline frequency domain data space.

If the interpolation effect is inappropriate or poor, continue to supplement the parameter values of the equipment's working condition parameters, such as using historical vibration source signal data under the nearby parameter value of speed _i , and based on the baseline frequency domain data space and the supplemented signal Data, use the above-mentioned interpolation algorithm to obtain the interpolation baseline frequency domain data space under the target unknown working condition parameters, and judge the quality of the interpolation effect again based on the evaluation index threshold β until the interpolation baseline frequency domain data space with good interpolation effect is obtained. , and the interpolated baseline frequency domain data space is added to the existing baseline frequency domain data space and used as a part of the existing baseline frequency domain data space.

According to this application, based on the baseline frequency domain data space under historical limited discrete working condition parameters, the interpolation fitting generation of the unknown new working condition baseline data space is performed, enriching and improving the discrete working condition and baseline data space, and avoiding the lack of historical working conditions. Limitations affecting the baseline data space for fault detection.

According to the above method, the baseline frequency domain data space is obtained. The obtained baseline frequency domain data space can be verified through the method described below to confirm whether the obtained baseline frequency domain data space is valid in the process of detecting equipment faults. .

4. Validation of baseline data space fault detection

Obtain the original data of normal vibration signals corresponding to the target channel and target working condition parameters and the vibration frequency domain data in the equipment fault state.

The sample data is divided according to a certain proportion to generate the baseline frequency domain data space and the frequency domain data to be detected (including normal state baseline frequency domain data and fault state frequency domain data).

Calculate the over-limit difference in frequency amplitude corresponding to the frequency domain data target to be detected and the baseline frequency domain data space, and generate the frequency domain over-limit difference distribution of the data to be detected.

Compare the distribution of out-of-limit differences with the fault judgment threshold, and analyze and compare the spatial distribution of the data to be detected and the baseline data through visual charts.

Based on the obtained comparison results (i.e., data labels), the detection results are verified for accuracy.

The following uses several specific embodiments to illustrate the effectiveness of the obtained baseline frequency domain data space and the interpolated baseline frequency domain data space under unknown target working condition parameters.

5. Embodiment of verification of effectiveness of baseline data space fault detection and unknown working condition baseline data space interpolation fitting effectiveness

Using the vibration signal data collected by the fracturing pump at the fracturing well site as the sample source data, three sets of comparative examples (Example 1 to Example 3) were set up to verify the effectiveness of spatial fault detection in the baseline frequency domain data: the same unit is normal Status data verification, normal status data verification of different units, and equipment fault status verification of different units. In addition, Embodiment 4 is set up to verify the validity of spatial interpolation fitting of baseline frequency domain data under unknown working conditions.

The three sets of comparative vibration signal data are all data collected by the vibration sensor of the fracturing pump reduction box. The working condition parameter is the rotation speed. The specific sensor installation position is: Example 1 and Example 2 use the reduction box to input the contralateral V channel data. Example 3 And Embodiment 4 uses a reduction gear box to input H channel data. And the above data collection frequency is all 51.2k HZ. In the fast Fourier transform, the fixed number of spectral lines fr=12800 and the fixed number of sampling points fd=51200 are set. And based on the accumulated calculation experience of historical data, the over-limit difference evaluation index threshold γ = 1.5 is set. Next, the above-mentioned baseline frequency domain data space is constructed for the three sets of data, and the over-limit difference evaluation index is calculated. The result looks like this:

Example 1:

From the historical operation data of a certain equipment on a certain platform, the discrete constant speed working condition of 1200rpm was screened from November 4 to November 14, 2021. The accumulated normal operating time of 8 hours was used as the control group to generate the constant speed working condition. The baseline frequency domain data space; from October 23 to October 30, 2021, the accumulated 8 hours of normal operating time data are divided into four groups of test groups, and each group of test groups generates segments under the fixed speed working condition. 7200 sets of frequency domain data. Next, the over-limit difference evaluation index is calculated for the test group data based on the baseline frequency domain data space generated by the control group. The results are as follows:

Table 1

It can be seen from Table 1 that among the four groups of test data, the second group has the largest number of samples with the maximum over-limit difference value greater than zero, but the maximum value of the over-limit difference evaluation index in all sample data is about 0.3, and the level of over-limit difference It is very low, and the same is true for other groups. The maximum values of the over-limit difference evaluation indicators are all at a low level and lower than the threshold γ. That is to say, the test data is completely covered by its baseline frequency domain data space, is in normal operating status, and is consistent with the actual situation of the data. Therefore, equipment fault detection based on the baseline data space is effective for the normal status data of the same unit.

Example 2:

From the historical operation data of two different equipment on a certain platform, the discrete fixed speed is 1100rpm working condition from October 7 to October 17, 2021. The vibration signal data of the accumulated normal state operating time for 8 hours is used as the control group to generate the fixed speed. Baseline frequency domain data space under rotational speed conditions; from October 23 to October 30, 2021, the vibration signal data of 8 hours of accumulated normal operating time was divided into four test groups, and each test group generated the fixed rotation speed There are a total of 7200 sets of segmented frequency domain data below. Next, the over-limit difference evaluation index is calculated for the test group data based on the baseline frequency domain data space generated by the control group. The experimental results are as follows:

Table 2

It can be seen from Table 2 that among the four groups of test data, the maximum value of the over-limit difference evaluation index is greater than zero. The fourth group has the largest number of samples, but the maximum value of the over-limit difference evaluation index in all sample data is about 0.2, which exceeds the limit. The level of the difference evaluation index is very low, and the same is true for other groups. The maximum values of the over-limit difference evaluation index are all at a low level and lower than the threshold γ. That is, the test data is completely covered by its baseline data space, is in normal operating status, and is consistent with the actual data situation. Therefore, it shows that equipment fault detection based on baseline data space is effective for different unit normal state data.

Example 3:

From the historical operation data of two different pieces of equipment on a certain platform, the discrete fixed speed working condition is 1080rpm from November 6 to November 16, 2021. The vibration signal data of the accumulated normal operating time of 8 hours is used as the control group to generate the fixed speed. Baseline frequency domain data space under rotational speed; from December 20 to December 29, 2021, the vibration signal data of the cumulative operating time of 4 hours was averaged into 2 test groups. Among them, the corresponding equipment status of the test group is damage to the bearing roller of the reduction box, that is, the equipment failure status. Each test group generates a total of 7200 sets of segmented frequency domain data at a certain rotation speed. Next, the over-limit difference evaluation index is calculated for the test group data based on the baseline frequency domain data space generated by the control group. The experimental results are as follows:

table 3

As can be seen from Table 3 above, the spectrum structure of the equipment fault state at the same speed completely exceeds the space limit of the baseline frequency domain data. The maximum value of the over-limit difference evaluation index reaches more than 10, which is significantly higher than the threshold γ and all sample data are basically over the limit. , the baseline frequency domain data space effectively responds to the damage to the bearing roller of the equipment reduction box, and the response results are consistent with the actual data fault status. Therefore, it shows that equipment fault detection based on baseline data space is effective for different unit fault status data.

In summary, it can be seen from the above three experiments that whether it is the same unit or different units, under the same working conditions, the baseline data space can intuitively and effectively detect the equipment fault state, and has no relevant response to the normal state of the equipment.

Example 4:

Screen the sample data of different discrete constant speed working conditions from the historical operation data of a certain equipment on a certain platform, perform the above baseline data space construction steps, and generate Fre_hil_AIi′ _k under different speed working conditions. The discrete operating speeds are [1000, 1100, 1200, 1300, 1400, 1500, 1600, 1700, 1800, 1900, 2000, 2100]. A total of twelve sets of discrete speed operating conditions correspond to 12 sets of baseline data spaces. And traverse and eliminate the rotational speeds in the above list in sequence, set the interpolation resolution H=12, and use the amplitudes of each order of the baseline data space under the remaining discrete rotational speed conditions to interpolate the amplitudes of each order of the eliminated rotational speed conditions ( The interpolation method uses the cubic spline interpolation algorithm), and then performs an inverse order transformation to obtain the spectrum interpolation data, and calculates the sum_diff evaluation index (i.e., the separation index) of the sum of the absolute values of the amplitude differences between the interpolated data and the original data. , setting the threshold β = 5.5 based on the statistical results of historical accumulated experimental data and expert experience. Among them, the calculation results of each group of interpolated sample data sum_diff are as follows:

Table 4

As can be seen from Table 4 above, in the 12 sets of interpolation experiments, the sum_diff evaluation indicators are all less than β, that is, the interpolation effect is relatively good, the accuracy of the interpolation data is relatively high, and the baseline data space under the interpolation speed condition has relative practical application significance and value. .

According to this application, by discretizing, processing and storing historical working conditions according to strong correlation parameters, it is conducive to refining fault detection scenarios, limiting the impact of different working condition factors on the effectiveness of fault detection results, and simplifying each working condition. Fault detection conditions and steps.

In addition, the original data of historical working condition parameters are truncated by using difference sequence equal step size, which can quickly and accurately obtain the discretized working condition parameter data and the corresponding time period.

Furthermore, based on the historical discrete working condition parameters, the original baseline data vibration data set, the baseline data vibration frequency domain data set, and the baseline data space of the normal operation of the equipment under different discrete working condition parameters are constructed, which provides a reliable basis for equipment fault detection. evaluation criteria and basis methods.

Moreover, by setting the over-limit difference evaluation index, the fault can be responded to intuitively, making it easier for relevant business personnel to understand and grasp the fault information and fault degradation degree.

Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage. In the media, when executed, the computer program may include the processes of the above method embodiments. Any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and/or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous chain Synchlink DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

The above descriptions are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and changes. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.

Claims

An equipment fault detection method based on baseline data space, characterized in that the equipment fault detection method includes:

Determining operating condition parameters that have a predetermined correlation with the operating conditions of the equipment;

Discretize the obtained original data set of the working condition parameters of the equipment according to the fixed time period of the working condition parameters of the equipment, and generate a discrete working condition parameter set containing multiple fixed time period data;

According to the discrete working condition parameter set and the obtained signal data of the equipment in the normal state, construct a baseline frequency domain data space of the equipment in the normal state, and set a fault judgment threshold; and

Calculate the over-limit difference evaluation index, and compare the over-limit difference evaluation index with the fault judgment threshold to determine whether the equipment is faulty.
The equipment fault detection method according to claim 1, characterized in that the original data set is discretized according to the fixed time period of the operating condition parameters of the equipment, and a data set containing multiple fixed time period data is generated. A set of discrete operating condition parameters, including:

Based on the original data set and lagging by one time point, a new data set of the lagged time point is obtained, where the new data set is the same length as the original data set;

Calculate the difference sequence between the new data set and the original data set, generate a difference sequence set, set a fault tolerance threshold σ, and set the data values of the difference sequence <σ to zero; and

Set the time length threshold t_lim, intercept the truncated data whose difference sequence is continuously zero for a duration >t_lim, and generate the discrete working condition parameter set.
The equipment fault detection method according to claim 1 or 2, characterized in that, based on the discrete working condition parameter set and the acquired signal data, a baseline frequency domain data space in the normal state of the equipment is constructed, including:

According to the discrete working condition parameter set, a baseline time domain data set and a baseline frequency domain data set are obtained;

Perform bandpass filtering on the baseline frequency domain data set to generate a second baseline frequency domain data set corresponding to the operating condition parameters in the frequency band of interest; and

The baseline frequency domain data space is constructed based on the second baseline frequency domain data set.
The equipment fault detection method according to claim 3, characterized in that, according to the discrete working condition parameter set, a baseline time domain data set and a baseline frequency domain data set are obtained, including:

The signal data corresponding to the time nodes in the signal data in the normal state are discretized into segments, and the parameter values of the working condition parameters corresponding to the time periods in which the segmented data are marked are generated to generate the baseline time domain. data set; and

Perform fast Fourier transform on the baseline time domain data set to obtain the baseline frequency domain data set.
The equipment fault detection method according to claim 3, wherein constructing the baseline frequency domain data space based on the second baseline frequency domain data set includes:

Expand each segment of spectrum data in the second baseline frequency domain data set spectrum by spectrum according to the frequency dimension to obtain a frequency matrix;

Calculate the statistical index values by column of the frequency matrix to obtain the upper limit space of the baseline frequency domain;

Perform Hilbert transform on the upper limit space of the baseline frequency domain and find the upper envelope to obtain the baseline frequency domain data space.
The equipment fault detection method according to claim 5, characterized in that the statistical index values include: mean, median, maximum value and extreme value.
The equipment fault detection method according to claim 1, characterized in that the equipment fault detection method further includes: based on the baseline frequency domain data space, according to the parameter values of the target unknown working condition parameters of the equipment, using interpolation algorithm to obtain the interpolated baseline frequency domain data space under unknown working condition parameters of the target.
The equipment fault detection method according to claim 7, characterized in that the equipment fault detection method further includes:

Setting the evaluation index threshold of the interpolated baseline frequency domain data space;

Calculate the degree of separation between the baseline frequency domain data space and the interpolated baseline frequency domain data space;

The separation degree is compared with the evaluation index threshold to determine whether the interpolation baseline frequency domain data space is suitable.
The equipment fault detection method according to claim 8, characterized in that the equipment fault detection method further includes:

Based on the comparison result between the separation degree and the evaluation index threshold, it is determined that the interpolated baseline frequency domain data space is suitable, and the interpolated baseline frequency domain data space is added to the baseline frequency domain data space and used as the baseline frequency domain data space. part of the domain data space; or

Based on the comparison result between the separation degree and the evaluation index threshold, it is determined that the interpolation baseline frequency domain data space is inappropriate, and the signal data at a parameter value close to the parameter value of the target unknown working condition parameter is supplemented; and

Based on the baseline frequency domain data space and the supplemented signal data, an interpolation algorithm is used to obtain the interpolated baseline frequency domain data space under the target unknown working condition parameters.
The equipment fault detection method according to any one of claims 7 to 9, characterized in that the working condition parameter is a parameter that has little correlation with the spectrum structure distribution, based on the baseline frequency domain data space, according to the The parameter values of the target unknown working condition parameters of the equipment are used, and the interpolation algorithm is used to obtain the interpolated baseline frequency domain data space under the target unknown working condition parameters, including:

The baseline frequency domain data space is sequentially superimposed to construct a joint distribution matrix;

According to the parameter values of the target unknown working condition parameters of the equipment, an interpolation algorithm is used to perform interpolation calculation on the joint distribution matrix to obtain the interpolated baseline frequency domain data space under the target unknown working condition parameters.
The equipment fault detection method according to any one of claims 7 to 9, characterized in that the working condition parameter is a parameter with high correlation with the spectrum structure distribution, based on the baseline frequency domain data space, according to the The parameter values of the target unknown working condition parameters of the equipment are used to obtain the interpolated baseline frequency domain data space under the target unknown working condition parameters using an interpolation algorithm, including:

The baseline frequency domain data space is sequentially superimposed to construct a joint distribution matrix;

Transform the frequency values in the joint distribution matrix into order values to obtain a joint distribution matrix in the order dimension;

According to the parameter values of the target unknown working condition parameters of the equipment, use an interpolation algorithm to perform interpolation calculations on the joint distribution matrix in the order dimension to obtain the interpolated baseline data order space under the target unknown working condition parameters;

The order value of the interpolation baseline data order space under the target unknown working condition parameters is inversely transformed and restored back to the frequency value to obtain the interpolated baseline frequency domain data space under the target unknown working condition parameters.
The equipment fault detection method according to claim 8, wherein the degree of separation is the sum of absolute values of amplitude differences between the baseline frequency domain data space and the interpolation baseline frequency domain space.
The equipment fault detection method according to claim 1, characterized in that the equipment fault detection method further includes: performing a cleaning process of supplementing missing values and removing zero values on the acquired original data set, and using the cleaned data Collection as a collection of primitive data.
The equipment fault detection method according to claim 13, characterized in that the missing value is supplemented by averaging the nearest neighbor data before and after the position of the missing value.
The equipment fault detection method according to claim 1, characterized in that the equipment fault detection method further includes:

Obtain signal data of the normal state of the device, where the signal data includes vibration signal data.
The equipment fault detection method according to claim 1, characterized in that the over-limit difference evaluation index=sum(abs(baseline upper limit data of baseline frequency domain data space-data to be detected))

Among them, abs (baseline upper limit data of the baseline frequency domain data space - data to be detected) means finding the absolute value of the difference between the baseline upper limit data of the baseline frequency domain data space and the data to be detected, sum (abs (baseline of the baseline frequency domain data space) Upper limit data - data to be detected)) represents the sum of the absolute values of the differences between the baseline upper limit data and the data to be detected in the baseline frequency domain data space.
A computer device, including a memory and a processor. The memory stores a computer program that can be run on the processor. It is characterized in that when the processor executes the computer program, any one of claims 1 to 16 is realized. The steps of the method described in the item.
A computer-readable storage medium with a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps of the method described in any one of claims 1 to 16 are implemented.