Sample library construction method, fault diagnosis method and device

Abstract

The present invention provides a sample library construction method, a fault diagnosis method and a device, and relates to the field of fault diagnosis technology, wherein the sample library construction method includes: obtaining equipment historical fault records and equipment historical operation data corresponding to the target equipment type; based on the fault field in the equipment historical fault records, retrieving and obtaining the operation basic data corresponding to each sample equipment in the equipment historical operation data; the operation basic data includes fault operation basic data and normal operation basic data; extracting features from the operation basic data to obtain initial samples, and clustering the initial samples to determine the fault sample set and normal sample set corresponding to the target equipment type; based on the fault sample set and the normal sample set, determining the sample library corresponding to the target equipment type. The method and device provided by the present invention improve the data reliability and accuracy of the sample library, and realize the automatic construction of the sample library according to the equipment historical fault records and equipment historical operation data.

Description

Technical Field

[0001] The present invention relates to the technical field of fault diagnosis, and in particular to a sample library construction method, a fault diagnosis method and a device. Background Art

[0002] During fault monitoring or fault diagnosis, a mechanism analysis is usually performed on the device or system under fault conditions to obtain typical data characteristics under the fault conditions, which are used as the basis for judging the working status of the device or system.

[0003] In the prior art, a sample library is usually generated based on the fault records of a device or system to perform fault monitoring or fault diagnosis on the device or system. Since most fault records are recorded manually, there is subjectivity in the recording of time and data, resulting in poor data reliability and low accuracy in the sample library. Summary of the invention

[0004] The present invention provides a sample library construction method, a fault diagnosis method and a device, which are used to solve the technical problems in the prior art of establishing a sample library through artificial records, and having poor data reliability and low accuracy.

[0005] The present invention provides a sample library construction method, comprising:

[0006] Obtain the equipment historical fault records and equipment historical operation data corresponding to the target equipment type;

[0007] Based on the fault field in the historical fault record of the device, retrieve and obtain the basic operation data corresponding to each sample device in the historical operation data of the device; the basic operation data includes fault operation basic data and normal operation basic data;

[0008] Extracting features from the basic operation data to obtain initial samples, and clustering the initial samples to determine a fault sample set and a normal sample set corresponding to the target device type;

[0009] Based on the fault sample set and the normal sample set, a sample library corresponding to the target device type is determined.

[0010] According to the sample library construction method provided by the present invention, based on the fault field in the historical fault record of the device, searching and obtaining the basic operation data corresponding to each sample device in the historical operation data of the device includes:

[0011] Based on the device number in the fault field in the device historical fault record, retrieve the historical operation data corresponding to each sample device in the device historical operation data;

[0012] Based on the fault recording time in the fault field and the preset time deviation, determining the fault operation basic data corresponding to each sample device in the historical operation data corresponding to each sample device;

[0013] Based on the historical operation data corresponding to each sample device and the fault operation basic data corresponding to each sample device, the normal operation basic data corresponding to each sample device is determined.

[0014] According to the sample library construction method provided by the present invention, clustering the initial samples to determine the fault sample set and the normal sample set corresponding to the target device type includes:

[0015] Clustering initial normal samples in the initial samples to obtain a plurality of first candidate categories, and determining a normal sample set corresponding to the target device type based on the number of initial normal samples in each first candidate category;

[0016] Clustering initial fault samples in the initial samples to obtain a plurality of second candidate categories;

[0017] If the similarity between any second candidate category and the normal sample set is less than a preset similarity threshold, the second candidate category is determined as the third candidate category;

[0018] Based on the number of initial fault samples in each third candidate category, a fault sample set corresponding to the target device type is determined.

[0019] According to the sample library construction method provided by the present invention, determining the sample library corresponding to the target device type based on the fault sample set and the normal sample set includes:

[0020] Samples are randomly removed from the fault sample set and the normal sample set respectively, so that the number of the fault sample set remains the same as the number of the normal sample set.

[0021] The present invention provides a fault diagnosis method, comprising:

[0022] Obtain the operating data of the equipment to be diagnosed;

[0023] Inputting the operating data into a fault diagnosis model to obtain a fault diagnosis result output by the fault diagnosis model;

[0024] Wherein, the training sample library of the fault diagnosis model is determined based on the sample library construction method.

[0025] According to the fault diagnosis method provided by the present invention, after the step of inputting the operating data into the fault diagnosis model to obtain the fault diagnosis result output by the fault diagnosis model, the method comprises:

[0026] Obtaining a correction result corresponding to the fault diagnosis result;

[0027] The fault diagnosis model is updated based on the correction result and the operation data.

[0028] The present invention provides a sample library construction device, comprising:

[0029] An acquisition unit, used to acquire equipment historical fault records and equipment historical operation data corresponding to the target equipment type;

[0030] A retrieval unit, configured to retrieve and obtain basic operation data corresponding to each sample device in the historical operation data of the device based on the fault field in the historical fault record of the device; the basic operation data includes fault operation basic data and normal operation basic data;

[0031] A clustering unit, configured to extract features from the basic operation data to obtain initial samples, and cluster the initial samples to determine a fault sample set and a normal sample set corresponding to a target device type;

[0032] An analyzing unit is used to determine a sample library corresponding to the target device type based on the fault sample set and the normal sample set.

[0033] The present invention provides an electronic device, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the sample library construction method or the fault diagnosis method when executing the program.

[0034] The present invention provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the sample library construction method or the fault diagnosis method are implemented.

[0035] The present invention provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the steps of the sample library construction method or the fault diagnosis method are implemented.

[0036] The present invention provides a sample library construction method, a fault diagnosis method and a device. According to the fault field in the historical fault record of the equipment, the basic operation data corresponding to each sample equipment is retrieved in the historical operation data of the equipment; the feature extraction of the basic operation data is performed to obtain the initial sample, and the initial sample is clustered to determine the fault sample set and the normal sample set corresponding to the target equipment type, so as to obtain the sample library corresponding to the target equipment type, and the historical operation data of the equipment is retrieved through the fault field in the historical fault record of the equipment to obtain the basic operation data with accurate time record and comprehensive data record, thereby avoiding the influence of human subjective factors on the sample library construction, improving the data reliability and accuracy of the sample library, and realizing the separation of normal samples and fault samples by clustering the initial samples, further improving the accuracy of the sample library, and in addition, realizing the automatic construction of the sample library according to the historical fault record of the equipment and the historical operation data of the equipment without human participation. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions in the present invention or 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 some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0038] Figure 1 This is a flow chart of the sample library construction method provided by the present invention;

[0039] Figure 2 The second flowchart of the sample library construction method provided by the present invention;

[0040] Figure 3 A schematic diagram of a flow chart of a fault diagnosis method provided by the present invention;

[0041] Figure 4 A schematic diagram of the structure of a sample library construction device provided by the present invention;

[0042] Figure 5 This is a schematic structural diagram of an electronic device provided by the present invention. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are 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.

[0044] In the related art, the fault diagnosis method trains the fault diagnosis model by constructing a sample library to improve the analysis and prediction ability of the fault diagnosis model. Most of the samples in the sample library come from human records and have uncertainty, which makes the trained fault diagnosis model have low accuracy.

[0045] Figure 1 This is a flow chart of a sample library construction method provided by the present invention, as shown in FIG. Figure 1 As shown, the method includes:

[0046] Step 110, obtaining the device historical fault records and device historical operation data corresponding to the target device type.

[0047] Specifically, the target device type is the type of device that needs to be diagnosed. Here, the target device can be a hardware device, such as an operating machine, a data server, a machine tool, etc.; it can also be a software device, such as an instrument monitoring system, an operation control system, etc. For example, for an excavator, its device type can be divided into multiple types according to different classification standards. For example, according to the driving mode, a fuel excavator and an electric excavator can be obtained.

[0048] The probability of equipment failure or the type of failure of equipment of the same equipment type are similar. Collecting the failure records or operation data of equipment of the same equipment type is of great reference significance for the failure warning and failure diagnosis of the equipment of this type.

[0049] The equipment historical fault record is a record of equipment failures during use. The equipment historical fault record generally includes the equipment number, fault time, working time, fault location, fault phenomenon, fault code displayed on the equipment, etc. For example, for a certain equipment manufacturer, the equipment historical fault record can be obtained through the fault case record of the manufacturer's customer relationship management system.

[0050] The historical operation data of the equipment is a record of various parameters or indicators of the equipment during operation in chronological order. It is obtained by real-time monitoring and recording of the parameters or indicators of the equipment. The various parameters of the equipment during operation can be collected and recorded through the equipment's own control system or remote monitoring system. For example, for a certain excavator, its own control system can collect parameters such as hydraulic oil pressure in real time, and save the parameters such as hydraulic oil pressure locally, and can send them to the remote monitoring system through the wireless network for data backup.

[0051] Step 120, based on the fault field in the historical fault record of the device, retrieve the basic operation data corresponding to each sample device in the historical operation data of the device; the basic operation data includes fault operation basic data and normal operation basic data.

[0052] Specifically, because the equipment historical fault records are usually recorded manually, the data is uncertain and has poor reliability. For example, when a lubricating oil failure occurs in a certain excavator, the maintenance personnel usually wait until the failure is eliminated before recording the failure. During the recording process, the maintenance personnel cannot accurately record the time when the failure occurred. In addition, the failure phenomena recorded by the maintenance personnel are mostly the appearances observed by themselves, and the parameter changes or indicator changes before and after the excavator failure cannot be accurately recorded, making the data in the equipment historical fault records subjective and arbitrary.

[0053] The historical operation data of the equipment is recorded by the equipment's own control system or remote monitoring system. The recorded parameters or indicators are arranged in chronological order, with high accuracy and data reliability. However, the amount of data in the historical operation data of the equipment is large, and it is difficult to directly obtain fault-related data from the historical operation data of the equipment. For example, in large equipment, for a certain important parameter, the event recording system of the equipment itself records it in units of 1 millisecond, and 3.6 million data records will be generated in 1 hour.

[0054] Considering that the data in the equipment's historical fault records have poor reliability, but the fault records are real, and the data in the equipment's historical operation data has high reliability, but the data volume is extremely large, valid data can be obtained by querying key fields.

[0055] According to the fault field in the historical fault record of the equipment, the basic operation data corresponding to each sample equipment can be retrieved in the historical operation data of the equipment. The fault field is the key field information related to the equipment failure, which may include the equipment number, fault recording time, working time, fault location, fault phenomenon and fault code.

[0056] The sample equipment here refers to equipment that has experienced failure. The basic operation data refers to the operation data of each sample equipment during the entire life cycle of the machine. The basic operation data can be divided into basic operation data for failure and basic operation data for normal operation.

[0057] For example, in the equipment historical fault record, the equipment number of the sample equipment that has failed can be used as a search keyword to retrieve the basic operation data corresponding to each sample equipment in the equipment historical operation data.

[0058] Step 130 , extracting features from the basic operation data to obtain initial samples, and clustering the initial samples to determine a fault sample set and a normal sample set corresponding to the target device type.

[0059] Specifically, the amount of running basic data is huge. If it is directly used to build a sample library, the amount of data in the sample library will be large and the use efficiency will be low. Therefore, feature extraction can be performed based on the running basic data, and the initial sample can be established based on the feature extraction results.

[0060] Feature engineering can be used to extract features from basic operating data. The extracted features can reflect the changing characteristics of parameters or indicators of sample equipment under fault conditions and normal conditions. The extracted features can be used as initial samples.

[0061] Feature engineering refers to the use of a series of engineering methods to filter out better data features from raw data to improve the training effect of the model. Feature engineering extracts the category features, numerical features, time features, and spatial features of the raw data. Specific methods include data preprocessing, feature extraction, feature selection, and feature construction.

[0062] There may still be data contamination in the initial samples. For example, the initial fault samples may be contaminated with normal samples, and the initial normal samples may be contaminated with fault samples. Therefore, the initial samples can be clustered, and the normal samples in the initial fault samples can be removed according to the clustering results, and the fault samples in the initial normal samples can be removed to obtain a higher accuracy fault sample set and normal sample set.

[0063] Feature clustering methods may include K-Means clustering, mean shift clustering, density-based clustering, maximum expectation clustering using a Gaussian mixture model, hierarchical clustering, and spectral clustering, etc. The embodiment of the present invention does not specifically limit the selection of feature clustering methods.

[0064] Step 140: Determine a sample library corresponding to the target device type based on the fault sample set and the normal sample set.

[0065] Specifically, a sample library corresponding to the target device type can be established based on the fault sample set and the normal sample set. The sample library includes two parts, which constitute positive and negative samples respectively. By adjusting the number of fault sample sets and normal sample sets, the balance of positive and negative sample data can be achieved. When the sample library is used to train the model, the model can simultaneously enhance the recognition of fault states and normal states, thereby improving the analysis and prediction capabilities of the model.

[0066] The sample library construction method provided by the embodiment of the present invention retrieves the basic operation data corresponding to each sample device in the historical operation data of the device according to the fault field in the historical fault record of the device; performs feature extraction on the basic operation data to obtain an initial sample, and clusters the initial sample to determine the fault sample set and the normal sample set corresponding to the target device type, thereby obtaining a sample library corresponding to the target device type; retrieves the historical operation data of the device through the fault field in the historical fault record of the device to obtain the basic operation data with accurate time record and comprehensive data record, thereby avoiding the influence of human subjective factors on the sample library construction, improving the data reliability and accuracy of the sample library, and realizing the separation of normal samples and fault samples by clustering the initial samples, further improving the accuracy of the sample library; in addition, the sample library is automatically constructed according to the historical fault record of the device and the historical operation data of the device without human intervention.

[0067] Based on the above embodiment, step 120 includes:

[0068] Based on the device number in the fault field in the device historical fault record, retrieve the historical operation data corresponding to each sample device in the device historical operation data;

[0069] Based on the fault recording time in the fault field and the preset time deviation, determining the fault operation basic data corresponding to each sample device in the historical operation data corresponding to each sample device;

[0070] Based on the historical operation data corresponding to each sample device and the fault operation basic data corresponding to each sample device, the normal operation basic data corresponding to each sample device is determined.

[0071] Specifically, the sample device where the fault occurs may be determined according to the device number in the fault field in the device historical fault record, and then the operation history data corresponding to each sample device may be retrieved from the device historical operation data according to the device number.

[0072] Since the fault record time in the fault field of each sample device is recorded manually, there is a deviation from the actual time when the sample device fails. A preset time deviation can be set to determine a time period with the fault record time as the midpoint and the preset time deviation as the fluctuation range, and the operation history data corresponding to the time period is extracted from the operation history data corresponding to each sample device as the fault operation basic data corresponding to each sample device. The preset time deviation can be selected as needed, for example, the time deviation can be selected as 1 hour, 5 hours or 10 hours, etc.

[0073] Then, after removing the faulty operation basic data from the operation history data corresponding to each sample device, the operation history data corresponding to the time period in which the parameters or indicators have no obvious abnormal changes are selected as the normal operation basic data corresponding to each sample device.

[0074] Based on any of the above embodiments, step 130 includes:

[0075] Clustering initial normal samples in the initial samples to obtain multiple first candidate categories, and determining a normal sample set corresponding to the target device type based on the number of initial normal samples in each first candidate category;

[0076] Clustering initial fault samples in the initial samples to obtain multiple second candidate categories;

[0077] If the similarity between any second candidate category and the normal sample set is less than a preset similarity threshold, the second candidate category is determined as the third candidate category;

[0078] Based on the number of initial fault samples in each third candidate category, a fault sample set corresponding to the target device type is determined.

[0079] Specifically, the acquired initial samples include initial fault samples and initial normal samples. For the device, it works in a normal state most of the time, so normal samples are very easy to obtain, and most of the samples collected for the device are normal samples. On the contrary, it is more difficult to obtain fault samples than normal samples. In the embodiment of the present invention, step 130 can effectively improve the success rate of obtaining fault samples, which is specifically achieved by the following method:

[0080] First, since the initial normal samples may be mixed with a small amount of faulty samples, a clustering method can be used to cluster the initial normal samples to obtain multiple first candidate categories. Among the multiple first candidate categories, the candidate categories with a large number of samples are more likely to be normal samples, and the candidate categories with a small number of samples are more likely to be faulty samples. Therefore, a first sample number threshold can be set, and candidate categories with a sample number less than the first sample number threshold can be eliminated as faulty samples, and candidate categories with a sample number greater than or equal to the first sample number threshold can be retained, and they can be merged as a normal sample set corresponding to the target device type.

[0081] Secondly, a clustering method can be used to cluster the initial fault samples to obtain multiple second candidate categories. Since each second candidate category may also have normal samples, if only the sample quantity selection method is used, the normal samples may be regarded as fault samples. At this time, each second candidate category can be compared with the normal sample for similarity. The higher the similarity, the greater the probability that the second candidate category is a normal sample, and the lower the similarity, the greater the probability that the second candidate category is a fault sample. The similarity can be compared using a class distance comparison method. Class distances include Euclidean distance, angle cosine, and correlation coefficient, etc. The embodiment of the present invention does not specifically limit the selection type of class distance.

[0082] A preset similarity threshold may be set according to actual needs. When the similarity between any second candidate category and the normal sample set is less than the preset similarity threshold, the second candidate category is determined as the third candidate category.

[0083] Again, if the third candidate category is not similar to the normal sample set, the samples in the third candidate category can be considered as faulty samples. The faulty sample set corresponding to the target device type can be determined based on the number of initial faulty samples in each third candidate category. For example, the candidate category with the largest number of initial faulty samples among multiple third candidate categories can be used as the faulty sample set corresponding to the target device type, or multiple third candidate categories can be combined to form the faulty sample set corresponding to the target device type.

[0084] In the above steps, the clustering method may be K-Means clustering, mean shift clustering, density-based clustering, maximum expectation clustering using a Gaussian mixture model, hierarchical clustering, and spectral clustering, etc. The embodiment of the present invention does not specifically limit the selection of the clustering method.

[0085] Based on any of the above embodiments, step 140 includes:

[0086] Samples are randomly removed from the fault sample set and the normal sample set respectively, so that the number of the fault sample set remains the same as that of the normal sample set.

[0087] Specifically, in order to keep the ratio of positive and negative samples in the sample library balanced, the number of fault sample sets and normal sample sets may be adjusted.

[0088] A random elimination method may be used to remove samples from the fault sample set and the normal sample set, respectively, so that the number of the fault sample set remains the same as the number of the normal sample set.

[0089] Based on any of the above embodiments, Figure 2 The second flow chart of the sample library construction method provided by the present invention is as follows: Figure 2 As shown, the method includes:

[0090] Step 1: For the target device, prepare the device's historical fault records (such as customer relationship management system fault case records, etc.) and the device's historical operation data (such as the device's operation indicators or operation parameters) within a certain time range;

[0091] Step 2: Extract key field information related to the fault in the historical fault records of the equipment, such as equipment number, fault recording time, working time, fault location, fault phenomenon, fault code, etc.;

[0092] Step 3. Take the intersection of the equipment number in the historical fault record and the equipment historical operation data to obtain the target equipment for constructing the sample library. Take the time point in the fault case record, consider the deviation between the recording time and the actual fault occurrence time, locate the operation data of the machine within a period of time before and after the time point (such as within 10 hours before and after), and automatically extract the data in this time period as the basic data of the initial fault sample. At the same time, take the time period with no fault records and no obvious abnormalities in the operation data, and extract the basic data of the initial normal sample;

[0093] Step 4: Extract feature quantities from the initial normal sample basic data and the fault sample basic data through feature engineering to obtain the initial sample. Since the initial fault sample may be mixed with normal samples, it is necessary to separate the normal sample from the fault sample. Based on the clustering method, separate the normal samples from the initial fault sample, and clean up the individual fault state samples in the initial normal sample to complete the cleaning of the initial sample and form an effective normal sample set and fault sample set;

[0094] Step 5: Considering the balance of sample size, the number of normal sample sets and fault sample sets are adjusted to be basically consistent by removing random samples to obtain a sample library.

[0095] The sample library construction method provided by the embodiment of the present invention avoids the influence of human subjective factors on the sample library construction, adopts objective fault records as the basic data of fault samples, and takes into account the uncertainty of the fault case recording time, thereby ensuring the authenticity and reliability of the final sample library from the source, and automatically completes the cleaning of the initial sample library using a machine learning method, realizing the separation of normal samples and fault samples, which not only improves the efficiency of sample library organization, but also further ensures the accuracy of the sample library.

[0096] Based on any of the above embodiments, Figure 3 A schematic diagram of a fault diagnosis method provided by the present invention is shown in FIG. Figure 3 As shown, the method includes:

[0097] Step 310, obtaining the operating data of the device to be diagnosed;

[0098] Step 320, inputting the operating data into the fault diagnosis model to obtain the fault diagnosis result output by the fault diagnosis model;

[0099] The training sample library of the fault diagnosis model is determined based on the above sample library construction method.

[0100] Specifically, the device to be diagnosed may be a hardware device, such as an operating machine, a data server, a machine tool, etc.; or a software device, such as an instrument monitoring system, an operation control system, etc. The fault diagnosis result is a possible fault of the device to be diagnosed. The fault diagnosis result includes at least one of the fault type, fault occurrence time, fault location, and fault code.

[0101] The operating data of the device to be diagnosed can be input into the fault diagnosis model, and the fault diagnosis model can predict the possible faults of the device to be diagnosed to obtain the fault diagnosis results.

[0102] The fault diagnosis model can be obtained by pre-training. Specifically, the fault diagnosis model can be obtained by the following training method:

[0103] First, the device history fault records and device history operation data of multiple devices of the same device type of the device to be diagnosed are collected; secondly, according to the sample library construction method in the above embodiment, a training sample library of the fault diagnosis model is established, and the training sample library includes a fault sample set and a normal sample set. Thirdly, the initial model is trained according to the training sample library to obtain a fault diagnosis model.

[0104] Based on any of the above embodiments, step 320 includes:

[0105] Obtaining a correction result corresponding to the fault diagnosis result;

[0106] Based on the calibration results and operating data, the fault diagnosis model is updated.

[0107] Specifically, the fault diagnosis result output by the fault diagnosis model is a prediction result, which may have an error with the actual result of the device. In this case, the actual fault record of the device can be used as the correction result.

[0108] When the fault diagnosis result is consistent with the actual fault record of the equipment, it can be considered that the fault diagnosis model has a high prediction accuracy and does not need to be updated.

[0109] When the fault diagnosis result is inconsistent with the actual fault record of the equipment, the actual fault record of the equipment can be used as the correction result and added to the sample library together with the operation data of the equipment as a new training sample to retrain the fault diagnosis model.

[0110] The above method realizes real-time updating of the sample library, improves the training effect of the fault diagnosis model, and improves the accuracy of the fault diagnosis model.

[0111] Based on any of the above embodiments, Figure 4 A schematic diagram of the structure of the sample library construction device provided by the present invention, such as Figure 4 As shown, the device comprises:

[0112] An acquisition unit 410 is used to acquire device historical fault records and device historical operation data corresponding to the target device type;

[0113] The retrieval unit 420 is used to retrieve the basic operation data corresponding to each sample device in the historical operation data of the device based on the fault field in the historical fault record of the device; the basic operation data includes the fault operation basic data and the normal operation basic data;

[0114] The clustering unit 430 is used to extract features from the basic operation data to obtain initial samples, and cluster the initial samples to determine a fault sample set and a normal sample set corresponding to the target device type;

[0115] The analyzing unit 440 is configured to determine a sample library corresponding to the target device type based on the fault sample set and the normal sample set.

[0116] The sample library construction device provided by the embodiment of the present invention retrieves the basic operation data corresponding to each sample device in the historical operation data of the device according to the fault field in the historical fault record of the device; performs feature extraction on the basic operation data to obtain an initial sample, and clusters the initial sample to determine the fault sample set and the normal sample set corresponding to the target device type, thereby obtaining a sample library corresponding to the target device type, and retrieves the historical operation data of the device through the fault field in the historical fault record of the device to obtain the basic operation data with accurate time records and comprehensive data records, thereby avoiding the influence of human subjective factors on the sample library construction, and improving the data reliability and accuracy of the sample library. By clustering the initial samples, the separation of normal samples and fault samples is achieved, and the accuracy of the sample library is further improved. In addition, the sample library is automatically constructed according to the historical fault records of the device and the historical operation data of the device without human intervention.

[0117] Based on any of the above embodiments, the retrieval unit is used to:

[0118] Based on the device number in the fault field in the device historical fault record, retrieve the historical operation data corresponding to each sample device in the device historical operation data;

[0119] Based on the fault recording time in the fault field and the preset time deviation, determining the fault operation basic data corresponding to each sample device in the historical operation data corresponding to each sample device;

[0120] Based on the historical operation data corresponding to each sample device and the fault operation basic data corresponding to each sample device, the normal operation basic data corresponding to each sample device is determined.

[0121] Based on any of the above embodiments, the clustering unit is used for:

[0122] Clustering initial normal samples in the initial samples to obtain multiple first candidate categories, and determining a normal sample set corresponding to the target device type based on the number of initial normal samples in each first candidate category;

[0123] Clustering initial fault samples in the initial samples to obtain multiple second candidate categories;

[0124] If the similarity between any second candidate category and the normal sample set is less than a preset similarity threshold, the second candidate category is determined as the third candidate category;

[0125] Based on the number of initial fault samples in each third candidate category, a fault sample set corresponding to the target device type is determined.

[0126] Based on any of the above embodiments, the analyzing unit is used for:

[0127] Samples are randomly removed from the fault sample set and the normal sample set respectively, so that the number of the fault sample set remains the same as that of the normal sample set.

[0128] Based on any of the above embodiments, the present invention provides a fault diagnosis device, comprising:

[0129] A data acquisition unit, used to acquire operating data of the device to be diagnosed;

[0130] A fault diagnosis unit, used for inputting the operation data into the fault diagnosis model to obtain the fault diagnosis result output by the fault diagnosis model;

[0131] The training sample library of the fault diagnosis model is determined based on the above sample library construction method.

[0132] Based on any of the above embodiments, the fault diagnosis device further includes:

[0133] A model updating unit, used to obtain a correction result corresponding to the fault diagnosis result;

[0134] Based on the calibration results and operating data, the fault diagnosis model is updated.

[0135] Based on any of the above embodiments, Figure 5A schematic diagram of the structure of an electronic device provided by the present invention, such as Figure 5 As shown, the electronic device may include: a processor (Processor) 510, a communication interface (Communications Interface) 520, a memory (Memory) 530 and a communication bus (Communications Bus) 540, wherein the processor 510, the communication interface 520, and the memory 530 communicate with each other through the communication bus 540. The processor 510 may call the logic command in the memory 530 to execute the following method:

[0136] Obtain the equipment historical fault records and equipment historical operation data corresponding to the target equipment type; based on the fault field in the equipment historical fault records, retrieve the operation basic data corresponding to each sample equipment in the equipment historical operation data; the operation basic data includes fault operation basic data and normal operation basic data; extract features from the operation basic data to obtain initial samples, and cluster the initial samples to determine the fault sample set and normal sample set corresponding to the target equipment type; based on the fault sample set and the normal sample set, determine the sample library corresponding to the target equipment type.

[0137] In addition, the logic commands in the above-mentioned memory 530 can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several commands to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., various media that can store program codes.

[0138] The processor in the electronic device provided in the embodiment of the present invention can call the logic instructions in the memory to implement the above method. Its specific implementation method is consistent with the implementation method of the aforementioned method and can achieve the same beneficial effects, which will not be repeated here.

[0139] An embodiment of the present invention further provides a non-transitory computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the method provided in each of the above embodiments is implemented, for example, including:

[0140] Obtain the equipment historical fault records and equipment historical operation data corresponding to the target equipment type; based on the fault field in the equipment historical fault records, retrieve the operation basic data corresponding to each sample equipment in the equipment historical operation data; the operation basic data includes fault operation basic data and normal operation basic data; extract features from the operation basic data to obtain initial samples, and cluster the initial samples to determine the fault sample set and normal sample set corresponding to the target equipment type; based on the fault sample set and the normal sample set, determine the sample library corresponding to the target equipment type.

[0141] When the computer program stored on the non-transitory computer-readable storage medium provided by the embodiment of the present invention is executed, the above method is implemented. Its specific implementation method is consistent with the implementation method of the aforementioned method and can achieve the same beneficial effects, which will not be repeated here.

[0142] An embodiment of the present invention provides a computer program product, including a computer program, and when the computer program is executed by a processor, the steps of the above method are implemented.

[0143] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for constructing a sample library, characterized in that: include: Obtain the equipment historical fault records and equipment historical operation data corresponding to the target equipment type; Based on the fault field in the historical fault record of the device, retrieve and obtain the basic operation data corresponding to each sample device in the historical operation data of the device; the basic operation data includes fault operation basic data and normal operation basic data; Extracting features from the basic operation data to obtain initial samples, and clustering the initial samples to determine a fault sample set and a normal sample set corresponding to the target device type; Determine a sample library corresponding to the target device type based on the fault sample set and the normal sample set; The clustering of the initial samples to determine a fault sample set and a normal sample set corresponding to the target device type includes: Clustering initial normal samples in the initial samples to obtain a plurality of first candidate categories, and determining a normal sample set corresponding to the target device type based on the number of initial normal samples in each first candidate category; Clustering initial fault samples in the initial samples to obtain a plurality of second candidate categories; If the similarity between any second candidate category and the normal sample set is less than a preset similarity threshold, the second candidate category is determined as the third candidate category; Based on the number of initial fault samples in each third candidate category, a fault sample set corresponding to the target device type is determined.

2. The sample library construction method according to claim 1, characterized in that: The retrieving and acquiring the operation basic data corresponding to each sample device in the device historical operation data based on the fault field in the device historical fault record includes: Based on the device number in the fault field in the device historical fault record, retrieve the historical operation data corresponding to each sample device in the device historical operation data; Based on the fault recording time in the fault field and the preset time deviation, determining the fault operation basic data corresponding to each sample device in the historical operation data corresponding to each sample device; Based on the historical operation data corresponding to each sample device and the fault operation basic data corresponding to each sample device, the normal operation basic data corresponding to each sample device is determined.

3. The sample library construction method according to claim 1, characterized in that: The determining, based on the fault sample set and the normal sample set, a sample library corresponding to the target device type includes: Samples are randomly removed from the fault sample set and the normal sample set respectively, so that the number of the fault sample set remains the same as the number of the normal sample set.

4. A fault diagnosis method, characterized in that: include: Obtaining the operating data of the equipment to be diagnosed; Inputting the operating data into a fault diagnosis model to obtain a fault diagnosis result output by the fault diagnosis model; Wherein, the training sample library of the fault diagnosis model is determined based on the sample library construction method according to any one of claims 1 to 3.

5. The fault diagnosis method according to claim 4, characterized in that: After the step of inputting the operating data into the fault diagnosis model to obtain the fault diagnosis result output by the fault diagnosis model, the method further comprises: Obtaining a correction result corresponding to the fault diagnosis result; The fault diagnosis model is updated based on the correction result and the operation data.

6. A sample library construction device, characterized in that: include: An acquisition unit, used to acquire equipment historical fault records and equipment historical operation data corresponding to the target equipment type; A retrieval unit, configured to retrieve and obtain basic operation data corresponding to each sample device from the device historical operation data based on the fault field in the device historical fault record; The basic operation data includes fault operation basic data and normal operation basic data; A clustering unit, configured to extract features from the basic operation data to obtain initial samples, and cluster the initial samples to determine a fault sample set and a normal sample set corresponding to a target device type; An analyzing unit, configured to determine a sample library corresponding to the target device type based on the fault sample set and the normal sample set; The clustering unit is used for: Clustering initial normal samples in the initial samples to obtain a plurality of first candidate categories, and determining a normal sample set corresponding to the target device type based on the number of initial normal samples in each first candidate category; Clustering initial fault samples in the initial samples to obtain a plurality of second candidate categories; If the similarity between any second candidate category and the normal sample set is less than a preset similarity threshold, the second candidate category is determined as the third candidate category; Based on the number of initial fault samples in each third candidate category, a fault sample set corresponding to the target device type is determined.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the steps of the sample library construction method according to any one of claims 1 to 3 or the fault diagnosis method according to claim 4 or 5 are implemented.

8. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the sample library construction method according to any one of claims 1 to 3 or the fault diagnosis method according to claim 4 or 5 are implemented.

9. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the sample library construction method according to any one of claims 1 to 3 or the fault diagnosis method according to claim 4 or 5 are implemented.

Images