Fault identification method, device, equipment and storage medium

By conducting exhaustive arrangement and combination of historical health data of industrial equipment and multivariate statistical analysis, fault codes are generated to identify fault types, solving the problems of low fault detection efficiency and difficulty in automatic identification in the existing technology, and achieving more efficient fault identification and automatic judgment.

CN114490829BActive Publication Date: 2024-12-06CNOOC INFORMATION TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210121119.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-02-09
Publication Date
2024-12-06
Estimated Expiration
2042-02-09

AI Technical Summary

Technical Problem

In the prior art, it is difficult to automatically identify the fault type through a computer for industrial process fault detection, and the training model is unstable, making it difficult to summarize the fault characteristics concisely, resulting in low fault recognition efficiency.

Method used

By obtaining the historical health data of the device, it is exhaustively arranged and combined, multiple training sets and test sets are generated, and multiple statistical analysis models are used for training and testing, the target training model is selected, and the fault code is created to identify the fault type.

Benefits of technology

It realizes fault recognition that is easier to execute by computers, optimizes training set data, and automatically recognizes fault types, improving the accuracy and efficiency of fault detection.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114490829B_ABST
    Figure CN114490829B_ABST
Patent Text Reader

Abstract

The present application discloses a fault identification method, device, equipment and storage medium, the method comprising: exhaustively permuting and combining the historical health data in all devices based on a preset number to obtain multiple training sets and corresponding multiple test sets; using different training sets and corresponding test sets to train and test the multivariate statistical analysis model to be trained to screen out the target trained model; creating corresponding feature codes based on the historical equipment status characteristics of different types of faults in a preset time period before the occurrence of the fault to obtain a first fault code; acquiring real-time data and inputting it into the target trained model for fault detection, and then creating corresponding feature codes based on the real-time equipment status characteristics in the real-time data to obtain a second fault code; comparing the first fault code with the second fault code to identify the current fault type. Through the technical solution of the present application, faults can be detected in a timely manner and the fault type can be effectively identified.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of industrial control fault diagnosis, and in particular to a fault identification method, device, equipment and storage medium. Background Art

[0002] With the development of industrial processes, the scale of production is getting larger and larger, and the process is becoming more and more complicated. In industrial processes, when equipment fails during operation but is not detected and processed in a timely manner, it will not only reduce production efficiency and increase maintenance costs, but even threaten human life safety in serious cases. Therefore, in order to prevent accidents, it is very important to introduce a real-time and effective fault detection system in the industrial site. Due to the rapid development of communication, sensing and other technologies, more and more industrial processes are equipped with data monitoring and acquisition systems, storing a large amount of real-time industrial production data. Therefore, industrial process fault detection based on data-driven, especially based on multivariate statistical methods, has received more and more attention.

[0003] However, in the prior art, when training the model: after the engineer selects a specific data segment of the health data, it is directly used as the training set input for model training, which is seriously affected by the collected data. The health data of different time periods are used as the training set for training, and the model results obtained are very different. In order to obtain a detection model with good performance, different data segments have to be repeatedly selected for training. However, the training effect of the selected data segment is not necessarily good, and it is ultimately difficult to find an ideal model; when identifying a fault: the data at all times of the fault are analyzed, and the contribution rate of each variable feature of the data at each time to the fault is calculated. The contribution rate information is recorded by drawing a contribution graph. It is difficult to classify the information of a certain fault, and the characteristics of a certain fault cannot be concisely summarized. It is not possible to simply determine whether it is the same type of fault by computer instructions.

[0004] In summary, how to make it easier for computers to execute, optimize training set data, automatically identify faults and determine fault types are issues that need to be resolved. Summary of the invention

[0005] In view of this, the purpose of the present invention is to provide a fault identification method, device, equipment and storage medium, which can be more easily executed by a computer, optimize training set data, automatically identify faults and determine the fault type. The specific scheme is as follows:

[0006] In a first aspect, the present application discloses a fault identification method, comprising:

[0007] Acquire historical health data of the device over a continuous period of time, and exhaustively permutate and combine all the historical health data based on a first preset number and a second preset number to obtain a plurality of training sets and a plurality of test sets corresponding to the plurality of training sets one by one; wherein each of the training sets includes the first preset number of the historical health data, and each of the test sets includes the second preset number of the historical health data;

[0008] Using different training sets and corresponding test sets to train and test the multivariate statistical analysis model to be trained, respectively, so as to select a target trained model from a plurality of trained models based on the test results;

[0009] Based on historical equipment status characteristics of different types of faults within a preset time period before occurrence, corresponding characteristic codes are created to obtain a first fault code;

[0010] Acquire real-time data, and input the real-time data into the target trained model for fault detection, and then create a corresponding feature code based on the real-time device status feature in the real-time data to obtain a second fault code;

[0011] The first fault code is compared with the second fault code to identify the current fault type.

[0012] Optionally, the using different training sets and corresponding test sets to train and test the multivariate statistical analysis model to be trained respectively, so as to select a target trained model from a plurality of trained models based on the test results, includes:

[0013] Standardizing the different training sets, and then performing dimensionality reduction on the standardized different training sets based on a multivariate statistical method to obtain a plurality of trained models;

[0014] The multiple trained models are tested using the test sets corresponding to different training sets to obtain corresponding test results, and a target trained model is screened out from the multiple trained models based on the test results and preset model screening rules.

[0015] Optionally, the filtering out a target trained model from the multiple trained models based on the test result and a preset model filtering rule includes:

[0016] Determine the target variable deviation statistics corresponding to different trained models at different times using the test results;

[0017] The trained models whose target variable deviation statistics at each moment are not greater than the preset control limit threshold are statistically obtained from the multiple trained models, and then the trained model closest to the preset reliability is screened out from the statistically obtained trained models to obtain the target trained model.

[0018] Optionally, the creating corresponding feature codes based on historical device status features within a preset time period before different types of faults occur to obtain a first fault code includes:

[0019] Using preset time point determination rules to determine target time points from preset time periods before different types of faults occur;

[0020] Obtain the numbers corresponding to different historical device status features to obtain corresponding feature codes;

[0021] The fault contribution rates of the feature codes at the target time point are calculated, and the fault contribution rates are sorted to obtain a first fault code.

[0022] Optionally, the method of determining the target time point from a preset time period before different types of faults occur using a preset time point determination rule includes:

[0023] If the variable deviation statistics corresponding to each moment of the different types of faults do not exceed the preset range of the preset control limit threshold, the moment corresponding to the variable deviation statistics with the largest value is determined as the target time point;

[0024] If there are more than a preset number of moments among all moments of the different types of faults whose corresponding variable deviation statistical values ​​exceed the preset range of the preset control limit threshold, the moment corresponding to the variable deviation statistical value that earliest exceeds the preset control limit threshold is determined as the target time point.

[0025] Optionally, acquiring real-time data and inputting the real-time data into the target trained model for fault detection includes:

[0026] Acquire real-time data, and input the real-time data into the target trained model, so as to calculate the target variable deviation statistic corresponding to the real-time data by using the target trained model, and count the time corresponding to when the target variable deviation statistic corresponding to the real-time data is greater than the preset control limit threshold, so as to obtain the data abnormal time point;

[0027] If the ratio between the number of the data abnormal time points and the total number of time points is greater than a preset ratio, it is determined that the device has a fault.

[0028] Optionally, comparing the first fault code with the second fault code to identify the current fault type includes:

[0029] If the first fault code is consistent with the second fault code, the data of the device at the time of the fault is recorded to obtain a fault probability trend graph and prompt the display of the current fault type;

[0030] If the first fault code is inconsistent with the second fault code, the device alarm displays an abnormality.

[0031] In a second aspect, the present application discloses a fault identification device, comprising:

[0032] A data acquisition module, used to acquire historical health data of the device within a continuous time, and based on a first preset number and a second preset number, perform exhaustive permutations and combinations of all the historical health data to obtain multiple training sets and multiple test sets corresponding to the multiple training sets one by one; wherein each of the training sets includes the first preset number of the historical health data, and each of the test sets includes the second preset number of the historical health data;

[0033] A model training module, used to train and test the multivariate statistical analysis model to be trained using different training sets and corresponding test sets, so as to select a target trained model from multiple trained models based on the test results;

[0034] A first fault code determination module, configured to create corresponding feature codes based on historical device status features of different types of faults within a preset time period before the occurrence of the faults, so as to obtain a first fault code;

[0035] A fault detection module, used for acquiring real-time data and inputting the real-time data into the target trained model for fault detection;

[0036] a second fault code determination module, configured to create a corresponding feature code based on the real-time device status feature in the real-time data to obtain a second fault code;

[0037] A fault identification module is used to compare the first fault code with the second fault code to identify the current fault type.

[0038] In a third aspect, the present application discloses an electronic device, comprising:

[0039] Memory, used to store computer programs and operating data;

[0040] The processor is used to execute the computer program to implement the aforementioned fault identification method.

[0041] In a fourth aspect, the present application discloses a computer-readable storage medium for storing a computer program; wherein the computer program implements the aforementioned fault identification method when executed by a processor.

[0042] In the present application, historical health data of the equipment in a continuous time is first obtained, and based on a first preset number and a second preset number, all the historical health data are exhaustively arranged and combined to obtain multiple training sets and multiple test sets corresponding to the multiple training sets; wherein each of the training sets includes the first preset number of historical health data, and each of the test sets includes the second preset number of historical health data; the multivariate statistical analysis model to be trained is trained and tested using different training sets and corresponding test sets, respectively, so as to screen out a target trained model from multiple trained models based on the test results; then, based on the historical equipment status characteristics of different types of faults within a preset time period before the occurrence, a corresponding feature code is created to obtain a first fault code; real-time data is obtained, and the real-time data is input into the target trained model for fault detection, and then a corresponding feature code is created based on the real-time equipment status characteristics in the real-time data to obtain a second fault code; the first fault code is compared with the second fault code to identify the current fault type. It can be seen that when determining the target trained model, directly input a large amount of historical health data for automatic selection, test the training set through exhaustive permutations and combinations, and obtain the best trained model in an enumerated manner, which is not only simple to operate but also has better results; when performing fault identification, the fault code is used to record the different equipment status characteristics corresponding to different types of faults, so that the fault identification process is more intelligent and can be automatically processed by the computer. By comparing the first fault code of different types of faults before the occurrence and the second fault code obtained in the real-time data, the current fault type can be identified. It is easier to be executed by the computer, optimize the training set data, automatically identify faults and judge the fault type. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0044] Figure 1 A flow chart of a fault identification method disclosed in this application;

[0045] Figure 2 This is a fault identification warning display interface effect diagram disclosed in this application;

[0046] Figure 3 A structural schematic diagram of a fault identification method disclosed in this application;

[0047] Figure 4 A specific fault identification method flow chart disclosed in this application;

[0048] Figure 5 A schematic diagram of a specific fault identification method disclosed in this application;

[0049] Figure 6 A system framework diagram of a fault identification method disclosed in this application;

[0050] Figure 7 This is a schematic diagram of the structure of a fault identification device disclosed in this application;

[0051] Figure 8 This is a structural diagram of an electronic device disclosed in this application. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0053] Currently, in the process of industrial process fault detection, repeatedly selecting different data segments for training when training models may not necessarily produce good results, and it is difficult to find an ideal model; when identifying faults, the characteristics of a certain fault cannot be concisely summarized, and it is not possible to simply determine whether it is the same type of fault by computer instructions.

[0054] To this end, the present application provides a fault identification solution that can be more easily executed by a computer, optimize training set data, automatically identify faults and determine the type of fault.

[0055] The embodiment of the present invention discloses a fault identification method, see Figure 1 As shown, the method includes:

[0056] Step S11: Obtain the historical health data of the device over a continuous period of time, and based on a first preset number and a second preset number, exhaustively permutate and combine all the historical health data to obtain multiple training sets and multiple test sets corresponding one-to-one to the multiple training sets; wherein each of the training sets includes the first preset number of historical health data, and each of the test sets includes the second preset number of historical health data.

[0057] In the embodiment of the present application, it is necessary to train a health data model. In this way, when fault detection is performed, the real-time data of the acquisition device can be calculated and compared with the data in the trained health data model to detect whether the device has a fault.

[0058] Specifically, first obtain the historical health data of the device over a continuous period of time. For example, collect a large amount of health data from the historical data of the device as samples, where the samples do not need to be selected in units of days, and data for several months can be collected directly. Then, based on the first preset number and the second preset number, perform exhaustive permutations and combinations of all the historical health data to obtain multiple training sets and multiple test sets corresponding one-to-one to the multiple training sets. It should be pointed out that before obtaining multiple training sets and multiple test sets corresponding one-to-one to the multiple training sets, the data when the device is in the parking state should be excluded first, that is, data segments with a large number of consecutive 0 elements.

[0059] In an embodiment of the present application, each of the training sets includes the first preset number of historical health data, and each of the test sets includes the second preset number of historical health data. For example, the historical health data of the device for one month is collected, and then 3-5 days of historical health data within one month are selected as the training set in the form of permutations and combinations, and the remaining historical health data are used as the test set to verify the test results of the training set. It can be understood that exhausting all permutations and combinations can enable the test set to screen out the best training set to obtain the best trained model.

[0060] Step S12: using different training sets and corresponding test sets to train and test the multivariate statistical analysis model to be trained, so as to select a target trained model from multiple trained models based on the test results.

[0061] In an embodiment of the present application, the multiple trained models are tested using the test sets corresponding to different training sets to obtain corresponding test results, and a target trained model is screened out from the multiple trained models based on the test results and preset model screening rules.

[0062] In the embodiment of the present application, when training the model with the training set, the different training sets are first standardized to become a matrix X with a mean of 0 and a variance of 1; then the different training sets after the standardization are reduced in dimension based on the multivariate statistical method to obtain multiple trained models. It should be pointed out that the multivariate statistical method includes but is not limited to PCA (Principal Components Analysis), ICA (Independent Component Analysis), etc. The embodiment of the present application takes the PCA dimensionality reduction data model as an example to project the data into a low-dimensional space, thereby compressing the data information more efficiently, and will not be described later. When the principal component analysis method is used to reduce the dimension of the standardized training set, a new matrix containing the corresponding principal component information in the principal component space after dimensionality reduction can be obtained, and the difference between the new matrix after the compressed information and the original matrix is ​​defined as the residual matrix; the principal component subspace will retain the main information of the observed variables, while the residual subspace mainly retains the noise and interference in the process.

[0063] Furthermore, after obtaining the multivariate statistical analysis model to be trained after information processing, namely the principal component model, we can then statistically determine whether the data at each moment is abnormal. 2 Statistics and Q statistics are often used in fault detection based on multivariate statistics as the degree of deviation of response data. 2 The statistic is obtained by calculating the Mahalanobis distance of the score vector of the principal component in space, which reflects the degree to which each principal component deviates from the model in terms of change trend and amplitude. It is a measure of the internalization of the model and can be used to monitor multiple principal components at the same time. The Q statistic is the Euclidean distance of the residual vector in the residual space, which describes the degree of deviation of the measured value of the input variable from the principal component model and is a measure of the external changes of the model. Therefore, the test set data corresponding to the training set can be projected into the principal component space after dimensionality reduction, and then T at each moment can be calculated. 2 Statistics and Q statistics. If the statistics exceed the preset control limit threshold, it is considered a false alarm. In this way, when the multivariate statistical analysis model to be trained is tested with the test set, the closer the probability of no false alarm is to the preset confidence within the algorithm, the better the model effect.

[0064] In a specific embodiment, T can be calculated 2 The control limit thresholds for the statistic and the Q statistic at a 99% confidence level, T 2 The threshold of the statistic can be calculated using the F distribution of probability statistics, and the threshold of the Q statistic can be calculated using the χ 2 The distribution is calculated.

[0065] In the embodiment of the present application, when the model performance is tested with the test set, the test results are first used to determine the target variable deviation statistics corresponding to different trained models at different times; it can be understood that the test results include false alarms and no false alarms, so the determined target variable deviation statistics are T under the condition of no false alarm. 2 Statistics and Q statistics. Further, from the multiple trained models, the trained models whose target variable deviation statistics at each moment are not greater than the preset control limit threshold are counted, and then the trained models closest to the preset reliability are screened out from the statistically trained models to obtain the target trained model. In this way, the best training model can be obtained by screening out the model with no false alarms that is closest to the preset reliability within the algorithm.

[0066] Step S13: Based on historical device status characteristics of different types of faults within a preset time period before occurrence, corresponding characteristic codes are created to obtain a first fault code.

[0067] In the embodiment of the present application, after training a healthy target trained model, it is necessary to train fault features, and create corresponding feature codes based on the historical equipment status features of different types of faults in a preset time period before the occurrence of the faults, so as to obtain the first fault code. In this way, the fault features of different types of faults can be recorded through the first fault code.

[0068] Step S14: Acquire real-time data, and input the real-time data into the target trained model for fault detection, and then create a corresponding feature code based on the real-time device status features in the real-time data to obtain a second fault code.

[0069] In the embodiment of the present application, after obtaining the first fault code recording the fault characteristics of different types of faults, the healthy target trained model can be used for fault detection. Specifically, first obtain real-time data, and input the real-time data into the target trained model, so as to use the target trained model to calculate the target variable deviation statistics corresponding to the real-time data. It can be understood that when training a healthy data model, since all collected are a large number of historical health data samples, when the target variable deviation statistics corresponding to different trained models at different times exceed the preset control limit threshold, it is regarded as a false alarm; then, in the embodiment of the present application, if the corresponding target variable deviation statistics of the acquired real-time data are greater than the preset control limit threshold after being input into the target trained model, it indicates that the corresponding moment data is abnormal.

[0070] Furthermore, the time corresponding to when the target variable corresponding to the real-time data deviates from the statistical value greater than the preset control limit threshold is counted to obtain the data abnormal time point; if the ratio between the number of data abnormal time points and the total number of time points is greater than the preset ratio, it is determined that the device has a fault.

[0071] In an embodiment of the present application, when it is determined that a fault exists in the equipment during fault detection, feature analysis is performed in the same manner as when training fault features to obtain a second fault code corresponding to the fault in the real-time data, that is, a corresponding feature code is created based on the real-time equipment status features in the real-time data to obtain a second fault code.

[0072] Step S15: Compare the first fault code with the second fault code to identify the current fault type.

[0073] In the embodiment of the present application, the first fault code is a characteristic code that records the fault characteristics of different types of faults obtained during the training of fault characteristics, and the specific fault type of the device can be determined by the first fault code. In this way, the second fault code corresponding to the fault in the real-time data is compared with the first fault code to identify the fault type of the current device.

[0074] It should be pointed out that if the first fault code is consistent with the second fault code, it is determined to be the same type of fault, and the data of the device at the time of the fault is recorded to obtain a fault probability trend chart and prompt the display of the current fault type; if the first fault code is inconsistent with the second fault code, the device alarm displays an abnormality.

[0075] For example, Figure 2 This is a warning display interface effect diagram provided by an embodiment of the present invention. It can monitor the probability trend change of equipment failure, that is, the monitoring data is integrated every hour, and the historical data of the last 48 hours is taken for calculation each time. 2 and the alarm rate of Q. It should be noted that T 2 The alarm rate is calculated as the value exceeding T 2 The number of abnormal points exceeding the threshold is divided by the total number of time points; the Q alarm rate is calculated by dividing the number of abnormal points exceeding the Q threshold by the total number of time points; when T 2Or when the alarm rate of Q is greater than 10%, an alarm message pops up, so the maximum value of the two alarm rate values ​​is used as the abnormal fault alarm rate, so that a recent abnormal fault alarm rate can be obtained every hour. Recording the abnormal fault alarm rate every hour and making a fault probability trend change graph can more intuitively reflect the status trend changes of the equipment to maintenance personnel. Compared with the traditional method of using statistical change graphs to show the equipment status trend and using alarm rates to determine whether a fault has occurred, for product users, the threshold for reading interface charts is lower, and there is no need to understand the concept of statistics. Furthermore, directly displaying the alarm rate as a trend graph makes the understanding of the current fault level of the equipment more intuitive.

[0076] like Figure 3 The figure shows a schematic diagram of the structure of a method provided by an embodiment of the present invention. It shows the basic structure of the embodiment of the present application, that is, the method consists of four steps: health model training, fault feature training, fault detection, and fault identification. Specifically, the historical health data in the acquisition device is used to train the model, and then the fault feature training is performed through the fault data to obtain a fault code that can characterize the fault type; when the real-time data is detected, if there is a fault, a fault code is also obtained, and the two fault codes are compared. If they are consistent, they are determined to be the same type of fault, and the device status and alarm information are obtained to complete the fault identification process.

[0077] In the present application, historical health data of the equipment in a continuous time is first obtained, and based on a first preset number and a second preset number, all the historical health data are exhaustively arranged and combined to obtain multiple training sets and multiple test sets corresponding to the multiple training sets; wherein each of the training sets includes the first preset number of historical health data, and each of the test sets includes the second preset number of historical health data; the multivariate statistical analysis model to be trained is trained and tested using different training sets and corresponding test sets, respectively, so as to screen out a target trained model from multiple trained models based on the test results; then, based on the historical equipment status characteristics of different types of faults within a preset time period before the occurrence, a corresponding feature code is created to obtain a first fault code; real-time data is obtained, and the real-time data is input into the target trained model for fault detection, and then a corresponding feature code is created based on the real-time equipment status characteristics in the real-time data to obtain a second fault code; the first fault code is compared with the second fault code to identify the current fault type. It can be seen that when determining the target trained model, directly input a large amount of historical health data for automatic selection, test the training set through exhaustive permutations and combinations, and obtain the best trained model in an enumerated manner, which is not only simple to operate but also has better results; when performing fault identification, the fault code is used to record the different equipment status characteristics corresponding to different types of faults, so that the fault identification process is more intelligent and can be automatically processed by the computer. By comparing the first fault code of different types of faults before the occurrence and the second fault code obtained in the real-time data, the current fault type can be identified. It is easier to be executed by the computer, optimize the training set data, automatically identify faults and judge the fault type.

[0078] The present application discloses a specific fault identification method, see Figure 4 As shown, the method includes:

[0079] Step S21: Obtain the historical health data of the device over a continuous period of time, and based on a first preset number and a second preset number, exhaustively permutate and combine all the historical health data to obtain multiple training sets and multiple test sets corresponding one-to-one to the multiple training sets; wherein each of the training sets includes the first preset number of historical health data, and each of the test sets includes the second preset number of historical health data.

[0080] Step S22: using different training sets and corresponding test sets to train and test the multivariate statistical analysis model to be trained, so as to select a target trained model from multiple trained models based on the test results.

[0081] For more specific processing procedures of the above steps S21 and S22, reference may be made to the corresponding contents disclosed in the above embodiments, which will not be described in detail here.

[0082] Step S23: using preset time point determination rules to determine a target time point from preset time periods before different types of faults occur.

[0083] In the embodiment of the present application, after training a healthy target post-training model, it is necessary to train fault features. First, it is necessary to collect data before the fault occurs, and obtain a preset time period before a specific fault occurs in the equipment, such as a specified impeller eccentricity, bearing misalignment, etc., such as historical data within 48 hours as a sample, and standardize the data according to the standard of the training set, that is, to make it a matrix with a mean of 0 and a variance of 1. Then find the key time point, and no longer record the fault features of all time points of the fault data. The operation is simple and easy to record.

[0084] In the process of determining the target time point, it is necessary to calculate the variable deviation statistics of each moment in the historical data before the occurrence of different types of faults. In the embodiment of the present application, the T of each moment of the data is calculated. 2 Statistics and Q statistics, T 2 The statistic reflects the degree of deviation of the data in the principal component space, and the Q statistic reflects the degree of deviation in the residual space. Then, the target time point is determined using the preset time point determination rule. Specifically, if the variable deviation statistical values ​​corresponding to each moment of the different types of faults do not exceed the preset range of the preset control limit threshold, the moment corresponding to the variable deviation statistical value with the largest value is determined as the target time point; if the variable deviation statistical values ​​corresponding to more than a preset number of moments of the different types of faults exceed the preset range of the preset control limit threshold among all moments, the moment corresponding to the variable deviation statistical value that first exceeds the preset control limit threshold is determined as the target time point. For example: if T of all time points 2 If the Q statistic does not exceed the control limit threshold by ten times, the time point with the highest statistic is selected; if it exceeds the control limit threshold by ten times, the time point that exceeds the control limit threshold by ten times the earliest is selected.

[0085] Step S24: obtaining serial numbers corresponding to different historical equipment status features to obtain corresponding feature codes; calculating the fault contribution rate of the feature codes at the target time point, and sorting the fault contribution rates to obtain a first fault code.

[0086] In the embodiment of the present application, the numbers corresponding to different historical equipment status features are obtained to obtain the corresponding feature codes. It should be noted that different types of faults correspond to different historical equipment status features, such as temperature, liquid level, pressure, etc. The corresponding feature codes can be obtained by numbering the different historical equipment status features collected; then the contribution rate of the feature codes corresponding to the different historical equipment status features to the fault at the target time point is calculated. It can be understood that the contribution rate reflects the degree of contribution of each feature of the data to the statistic. The feature variable with a larger contribution rate value is considered to be a variable that is more likely to cause a fault. The calculation formula is as follows:

[0087]

[0088] Where x represents the array after the standardized features of each variable at a specific moment; x k represents an array in which all elements are 0 except the kth element which is the same as x; P is the load matrix obtained by principal component analysis of the healthy data set X. Λ is the diagonal matrix of eigenvalues ​​obtained by decomposing the eigenvalues ​​of X; after calculation, the contribution of each variable feature to the fault is obtained.

[0089] In the embodiment of the present application, after the contribution of each variable feature to the fault is obtained, the feature code of the fault is recorded to obtain the first fault code. In this process, the contribution rate of each state feature of the target time point data to the fault is calculated, and the first few serial numbers are sorted from large to small according to the contribution rate as fault codes. The specific number can be set arbitrarily and is not specifically limited here. For example, the serial numbers of the first three variable features are sorted from large to small according to the contribution rate as the first fault code. When the contribution rate of the variable serial number is less than 0.1, the serial number 0 is used to fill it. For example, after a feature analysis, the feature serial numbers with the largest contribution rate are 5, 3, and 2, respectively, and the contribution rates are 1.22, 0.63, and 0.05, respectively. At this time, feature 2 is not retained because the contribution rate is too small, and the saved fault code is [5, 3, 0]. The first fault code is an integer array with three elements, which records the fault characteristics of different types of faults.

[0090] Step S25: Acquire real-time data, and input the real-time data into the target trained model for fault detection, and then create a corresponding feature code based on the real-time device status features in the real-time data to obtain a second fault code.

[0091] Step S26: Compare the first fault code with the second fault code to identify the current fault type.

[0092] Among them, for more specific processing procedures of the above steps S25 and S26, reference can be made to the corresponding contents disclosed in the above embodiments, which will not be repeated here.

[0093] like Figure 5 As shown, a specific fault identification schematic diagram disclosed in the present application includes the overall process of fault identification in the embodiment of the present application. The entire process requires training of health models, training of fault features, fault detection and fault identification.

[0094] Training health data models: Collect health history data, divide into multiple training sets and test sets, train PCA dimensionality reduction data models, calculate control limit thresholds of statistics, use test sets to verify model performance, and find the optimal training model.

[0095] Training fault features: Collect historical data before the fault occurs, find key time points, calculate the contribution rate of each feature, and record fault codes.

[0096] Fault detection: collect real-time data, project the data into PCA dimension reduction space, calculate T 2 With the Q statistic, determine whether the statistic exceeds the control limit threshold, determine whether the statistic exceeds the threshold, and whether the alarm rate exceeds the threshold. It should be pointed out that when collecting real-time data, the variables of the m bit numbers collected at each time point, that is, the different historical equipment status feature data, are formed into an array of length m, and the data is standardized according to the standard of the training set. It can be understood that if the alarm rate exceeds the threshold, the fault identification step is entered. If the alarm rate does not exceed the threshold, the display device is operating normally and the data at the next moment is collected.

[0097] Fault identification: If the real-time alarm rate exceeds the threshold, the contribution rate of each feature at the key time point is calculated, the fault code is recorded, and it is determined whether the fault code is consistent with the fault code of the specific fault. If it is consistent, the alarm will display the specific fault; if it is inconsistent, the alarm will display the abnormal condition.

[0098] For example, Figure 6 As shown, a system framework diagram for industrial equipment provided by an embodiment of the present invention includes the system hierarchy and module framework, including the industrial field layer, data acquisition layer, fault identification algorithm layer and human-computer interaction layer. When the method is actually used on site, the environment in which the algorithm is located is that the "data acquisition layer" first collects data from the equipment in the industrial field through the distributed control system (DCS) and stores it in the historical database, and then the historical data and real-time data are transmitted to the fault identification algorithm layer, which is used to monitor and identify faults in real time. The final algorithm result is displayed in the "human-computer interaction layer" and presented to the user in the form of an early warning interface.

[0099] In the present application, the historical health data of the device in a continuous time is first obtained, and based on a first preset number and a second preset number, all the historical health data are exhaustively arranged and combined to obtain multiple training sets and multiple test sets corresponding to the multiple training sets; wherein each of the training sets includes the historical health data of the first preset number, and each of the test sets includes the historical health data of the second preset number; different training sets and corresponding test sets are used to train and test the multivariate statistical analysis model to be trained, respectively, so as to screen out the target trained model from the multiple trained models based on the test results. type; then determine the target time point from the preset time period before the occurrence of different types of faults using the preset time point determination rule; obtain the numbers corresponding to different historical equipment status features to obtain the corresponding feature codes; calculate the fault contribution rate of the feature code at the target time point, and sort the fault contribution rate to obtain the first fault code; obtain real-time data, and input the real-time data into the target trained model for fault detection, and then create the corresponding feature code based on the real-time equipment status features in the real-time data to obtain the second fault code; compare the first fault code with the second fault code to identify the current fault type. It can be seen that when determining the target trained model, directly input a large amount of historical health data for automatic selection, test the training set through exhaustive permutations and combinations, and obtain the best trained model in an enumerated manner, which is not only simple to operate but also has better effects; when performing fault identification, the different equipment status features corresponding to different types of faults are recorded through fault codes, so that the fault identification process is more intelligent and can be automatically processed by a computer. By comparing whether the first fault code of different types of faults before the occurrence is consistent with the second fault code obtained in the real-time data, the current fault type can be identified. It is easier for computers to perform, optimize training set data, automatically identify faults and determine the type of fault.

[0100] Accordingly, the present application embodiment discloses a fault identification device, see Figure 7 As shown, the device comprises:

[0101] The data acquisition module 11 is used to acquire the historical health data of the device within a continuous time, and based on a first preset number and a second preset number, exhaustively arrange and combine all the historical health data to obtain multiple training sets and multiple test sets corresponding to the multiple training sets one by one; wherein each of the training sets includes the historical health data of the first preset number, and each of the test sets includes the historical health data of the second preset number;

[0102] A model training module 12, used to train and test the multivariate statistical analysis model to be trained using different training sets and corresponding test sets, so as to select a target trained model from multiple trained models based on the test results;

[0103] A first fault code determination module 13, for creating corresponding feature codes based on historical device status features of different types of faults within a preset time period before the occurrence of the faults, so as to obtain a first fault code;

[0104] A fault detection module 14 is used to obtain real-time data and input the real-time data into the target trained model for fault detection;

[0105] A second fault code determination module 15, configured to create a corresponding feature code based on the real-time device status feature in the real-time data to obtain a second fault code;

[0106] The fault identification module 16 is used to compare the first fault code with the second fault code to identify the current fault type.

[0107] Among them, for more specific working processes of the above-mentioned modules, please refer to the corresponding contents disclosed in the aforementioned embodiments, which will not be repeated here.

[0108] It can be seen that through the above scheme of this embodiment, first, the historical health data of the equipment in a continuous time is obtained, and based on the first preset number and the second preset number, all the historical health data are exhaustively arranged and combined to obtain multiple training sets and multiple test sets corresponding to the multiple training sets; wherein each of the training sets includes the first preset number of historical health data, and each of the test sets includes the second preset number of historical health data; the multivariate statistical analysis model to be trained is trained and tested using different training sets and corresponding test sets, respectively, so as to screen out the target trained model from the multiple trained models based on the test results; then, based on the historical equipment status characteristics of different types of faults within a preset time period before the occurrence, a corresponding feature code is created to obtain a first fault code; real-time data is obtained, and the real-time data is input into the target trained model for fault detection, and then a corresponding feature code is created based on the real-time equipment status characteristics in the real-time data to obtain a second fault code; the first fault code is compared with the second fault code to identify the current fault type. It can be seen that when determining the target trained model, directly input a large amount of historical health data for automatic selection, test the training set through exhaustive permutations and combinations, and obtain the best trained model in an enumerated manner, which is not only simple to operate but also has better results; when performing fault identification, the fault code is used to record the different equipment status characteristics corresponding to different types of faults, so that the fault identification process is more intelligent and can be automatically processed by the computer. By comparing the first fault code of different types of faults before the occurrence and the second fault code obtained in the real-time data, the current fault type can be identified. It is easier to be executed by the computer, optimize the training set data, automatically identify faults and judge the fault type.

[0109] Furthermore, an embodiment of the present application discloses an electronic device, Figure 8 This is a structural diagram of an electronic device 20 according to an exemplary embodiment, and the content in the diagram cannot be considered as any limitation on the scope of use of the present application.

[0110] Figure 8 The present invention provides a schematic diagram of the structure of an electronic device 20 provided in an embodiment of the present application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 is used to store a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the fault identification method disclosed in any of the aforementioned embodiments.

[0111] In this embodiment, the power supply 23 is used to provide working voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and the external device, and the communication protocol it follows is any communication protocol that can be applied to the technical solution of the present application, and is not specifically limited here; the input and output interface 25 is used to obtain external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs and is not specifically limited here.

[0112] In addition, the memory 22 as a carrier for resource storage may be a read-only memory, a random access memory, a disk or an optical disk, etc. The resources stored thereon may include an operating system 221, a computer program 222 and data 223, etc. The data 223 may include various data. The storage method may be temporary storage or permanent storage.

[0113] The operating system 221 is used to manage and control the hardware devices on the electronic device 20 and the computer program 222, which can be Windows Server, Netware, Unix, Linux, etc. In addition to including a computer program that can be used to complete the fault identification method performed by the electronic device 20 disclosed in any of the aforementioned embodiments, the computer program 222 can further include a computer program that can be used to complete other specific tasks.

[0114] Further, the embodiment of the present application discloses a computer-readable storage medium, wherein the computer-readable storage medium mentioned here includes a random access memory (Random Access Memory, RAM), a memory, a read-only memory (Read-Only Memory, ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a magnetic disk or an optical disk, or any other form of storage medium known in the technical field. Wherein, when the computer program is executed by the processor, the aforementioned fault identification method is implemented. For the specific steps of the method, reference can be made to the corresponding contents disclosed in the aforementioned embodiment, and no further description will be given here.

[0115] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0116] The steps of the fault identification or algorithm described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0117] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0118] The above is a detailed introduction to a fault identification method, device, equipment and storage medium provided by the present invention. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for general technical personnel in this field, according to the idea of ​​the present invention, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A fault identification method, characterized in that: include: Acquire historical health data of the device over a continuous period of time, and exhaustively permutate and combine all the historical health data based on a first preset number and a second preset number to obtain a plurality of training sets and a plurality of test sets corresponding to the plurality of training sets one by one; wherein each of the training sets includes the first preset number of the historical health data, and each of the test sets includes the second preset number of the historical health data; Using different training sets and corresponding test sets to train and test the multivariate statistical analysis model to be trained, respectively, so as to select a target trained model from a plurality of trained models based on the test results; Based on historical equipment status characteristics of different types of faults within a preset time period before occurrence, a corresponding characteristic code is created to obtain a first fault code; Acquire real-time data, and input the real-time data into the target trained model for fault detection, and then create a corresponding feature code based on the real-time device status feature in the real-time data to obtain a second fault code; The first fault code is compared with the second fault code to identify the current fault type.

2. The fault identification method according to claim 1, characterized in that: The method of respectively using different training sets and corresponding test sets to train and test the multivariate statistical analysis model to be trained, so as to select a target trained model from a plurality of trained models based on the test results, includes: Standardizing the different training sets, and then performing dimensionality reduction on the standardized different training sets based on a multivariate statistical method to obtain a plurality of trained models; The multiple trained models are tested using the test sets corresponding to different training sets to obtain corresponding test results, and a target trained model is screened out from the multiple trained models based on the test results and preset model screening rules.

3. The fault identification method according to claim 2, characterized in that: The step of selecting a target trained model from the plurality of trained models based on the test result and a preset model screening rule includes: Determine the target variable deviation statistics corresponding to different trained models at different times using the test results; The trained models whose target variable deviation statistics at each moment are not greater than the preset control limit threshold are statistically obtained from the multiple trained models, and then the trained model closest to the preset reliability is screened out from the statistically obtained trained models to obtain the target trained model.

4. The fault identification method according to claim 3, characterized in that: The method of creating corresponding feature codes based on historical device status features within a preset time period before different types of faults occur, so as to obtain a first fault code, includes: Using preset time point determination rules to determine target time points from preset time periods before different types of faults occur; Obtain the numbers corresponding to different historical device status features to obtain corresponding feature codes; The fault contribution rates of the feature codes at the target time point are calculated, and the fault contribution rates are sorted to obtain a first fault code.

5. The fault identification method according to claim 4, characterized in that: The method of using a preset time point determination rule to determine a target time point from a preset time period before different types of faults occur includes: If the variable deviation statistics corresponding to each moment of the different types of faults do not exceed the preset range of the preset control limit threshold, the moment corresponding to the variable deviation statistics with the largest value is determined as the target time point; If there are more than a preset number of moments among all moments of the different types of faults whose corresponding variable deviation statistical values ​​exceed the preset range of the preset control limit threshold, the moment corresponding to the variable deviation statistical value that earliest exceeds the preset control limit threshold is determined as the target time point.

6. The fault identification method according to claim 3, characterized in that: The acquiring of real-time data and inputting the real-time data into the target trained model for fault detection comprises: Acquire real-time data, and input the real-time data into the target trained model, so as to calculate the target variable deviation statistic corresponding to the real-time data by using the target trained model, and count the time corresponding to when the target variable deviation statistic corresponding to the real-time data is greater than the preset control limit threshold, so as to obtain the data abnormal time point; If the ratio between the number of the data abnormal time points and the total number of time points is greater than a preset ratio, it is determined that the device has a fault.

7. The fault identification method according to any one of claims 1 to 6, characterized in that: The comparing the first fault code with the second fault code to identify the current fault type includes: If the first fault code is consistent with the second fault code, the data of the device at the time of the fault is recorded to obtain a fault probability trend graph and prompt the display of the current fault type; If the first fault code is inconsistent with the second fault code, the device alarm displays an abnormality.

8. A fault identification device, characterized in that: include: A data acquisition module, used to acquire historical health data of the device within a continuous time, and based on a first preset number and a second preset number, perform exhaustive permutations and combinations of all the historical health data to obtain multiple training sets and multiple test sets corresponding to the multiple training sets one by one; wherein each of the training sets includes the first preset number of the historical health data, and each of the test sets includes the second preset number of the historical health data; A model training module, used to train and test the multivariate statistical analysis model to be trained using different training sets and corresponding test sets, so as to select a target trained model from multiple trained models based on the test results; A first fault code determination module, configured to create corresponding feature codes based on historical device status features of different types of faults within a preset time period before the occurrence of the faults, so as to obtain a first fault code; A fault detection module, used for acquiring real-time data and inputting the real-time data into the target trained model for fault detection; a second fault code determination module, configured to create a corresponding feature code based on the real-time device status feature in the real-time data to obtain a second fault code; A fault identification module is used to compare the first fault code with the second fault code to identify the current fault type.

9. An electronic device, characterized in that: include: Memory, used to store computer programs and operating data; A processor, configured to execute the computer program to implement the fault identification method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: Used to store a computer program; wherein when the computer program is executed by a processor, the fault identification method according to any one of claims 1 to 7 is implemented.