A method and system for predicting fault points based on multiple failure analysis

Abstract

The present invention relates to a method and system for predicting a fault point based on multiple failure analyses, which belongs to the technical field of failure analysis and solves the problems of incompleteness and low accuracy of existing failure analyses. The method includes analyzing historical maintenance records to obtain abnormal points in the fault points, deleting the detection steps related to the abnormal points based on the parts associated with the detection steps and the lines to which they belong, and adding the detection steps associated with other parts belonging to the same line to obtain the actual detection steps; performing failure analysis based on the diagnostic data of the assembly to be detected, and if it is the first failure analysis, obtaining the actual detection steps, otherwise updating the actual detection steps based on the fault point of the previous failure analysis; obtaining the fault point based on the actual detection value, and ending the failure analysis if the fault point is correct, otherwise iterating until the maximum number of failure analyses is reached to obtain the final fault point; if the final fault point is incorrect, predicting the fault point based on the trained classification model. This improves the accuracy of failure analysis.

Description

Technical Field

[0001] The present invention relates to the technical field of failure analysis, and in particular to a method and system for predicting a fault point based on multiple failure analyses. Background Art

[0002] Failure analysis generally involves analyzing and verifying failure modes and phenomena, simulating and reproducing failure phenomena, identifying the causes, and uncovering the failure mechanisms. Failure analysis has significant practical significance in improving product quality, technological development and improvement, product repair, and arbitration of failure incidents. As the volume of equipment increases, the amount of industrial failure analysis data stored will grow exponentially.

[0003] Typically, faulty industrial assemblies undergo an initial diagnosis on a test machine to identify the fault symptoms. Then, a failure analysis process is performed using the corresponding standard failure analysis process. The fault point is then identified and repaired. The process is then completed to see if the repair is successful. If the repair is not successful, the initial diagnosis process is repeated to determine the basic fault symptoms. This cycle continues until a certain number of failure analysis and repairs are no longer required.

[0004] The existing standard failure analysis process has many steps and cannot quickly and accurately find the fault point. In addition, the standard process used in multiple failure analyses is the same, and the analysis accuracy is low. Summary of the Invention

[0005] In view of the above analysis, the embodiments of the present invention aim to provide a method and system for predicting fault points based on multiple failure analyses, so as to solve the problems of incompleteness and low accuracy of existing failure analyses.

[0006] In one aspect, an embodiment of the present invention provides a method for predicting a fault point based on multiple failure analysis, comprising the following steps:

[0007] Analyze historical maintenance records to identify anomalies in fault points. Based on the parts and lines associated with all inspection steps in each inspection process, delete the inspection steps related to the anomalies and add the inspection steps associated with other parts belonging to the same line to obtain the actual inspection steps of each inspection process.

[0008] Perform failure analysis based on the diagnostic data of the assembly to be tested. If this is the first failure analysis, obtain the actual test steps of the corresponding test process. Otherwise, update the actual test steps of the corresponding test process based on the fault point of the previous failure analysis. Obtain the fault point based on the obtained actual test value. If the fault point is correct, end the failure analysis. Otherwise, iterate the next failure analysis until the maximum number of failure analyses is reached and the final fault point is obtained.

[0009] If the final fault point is incorrect, the data to be predicted is constructed based on multiple failure analyses and input into the corresponding trained classification model to predict the fault point.

[0010] Based on the further improvement of the above method, the historical maintenance records are analyzed to obtain the abnormal points in the fault points, including:

[0011] Based on historical maintenance records, the yield rate of failure points is calculated periodically;

[0012] Remove the fault points whose yield does not meet the Sigma principle and obtain the fault points to be clustered;

[0013] Using the density clustering algorithm, the fault points to be clustered are clustered according to their yield rate in the same cycle and the preset neighborhood radius to obtain cluster categories. The fault points in the category with a number of fault points less than the threshold are regarded as outliers.

[0014] Outliers whose yield is less than the minimum yield threshold are regarded as abnormal points.

[0015] Based on the further improvement of the above method, based on the parts and their lines associated with all the inspection steps of each inspection process, the inspection steps related to the abnormal points are deleted and the inspection steps associated with other parts belonging to the same line are added to obtain the actual inspection steps of each inspection process, including:

[0016] Based on the failure analysis knowledge graph, the parts associated with each inspection step and the associated line entities are obtained under each inspection process. The name of the abnormal point is used as the part entity name, and the part entity corresponding to the abnormal point and its associated line entity are queried to obtain the line to which the abnormal point belongs.

[0017] In all the inspection steps of each inspection process, delete the inspection steps corresponding to the abnormal point and the inspection steps associated with the parts belonging to the same line as the abnormal point;

[0018] For each inspection process, the inspection steps associated with other parts belonging to the same circuit are added according to the parts associated with the remaining inspection steps and their corresponding circuits, and the relationship between the inspection steps is updated to obtain the actual inspection steps of each inspection process.

[0019] Based on a further improvement of the above method, the diagnostic data of the assembly to be inspected includes inspection item names and inspection values ​​corresponding to multiple fault phenomena respectively, and the corresponding inspection process is obtained according to the first inspection item name.

[0020] Based on the further improvement of the above method, according to the fault point of the previous failure analysis, the actual detection steps of the corresponding detection process are updated, including:

[0021] Based on the fault point of the last failure analysis, obtain the inspection steps associated with all parts on the adjacent lines of the line to which the fault point belongs, put them into the set to be supplemented in order, add the set to be supplemented to the actual inspection steps of the corresponding inspection process, and update the relationship between the inspection steps.

[0022] Based on the further improvement of the above method, the relationship between the updated detection steps includes:

[0023] Obtain the detection step corresponding to the fault point as the first operation step, and obtain the corresponding detection step according to the next operation relationship of the first operation step as the second operation step; associate the next operation relationship of the first operation step to the first detection step in the set to be supplemented, and associate the next operation relationship of the last detection step in the set to be supplemented to the second operation step.

[0024] Based on the further improvement of the above method, the fault point is obtained according to the actual detection value, including:

[0025] Based on the actual detection steps of the corresponding detection process, the detection step with the preset first operation step code is used as the first operation;

[0026] The actual detection value of the first operation is taken as the judgment result. According to the detection result associated with the first operation, it is identified whether the fault point corresponding to the judgment result is empty. If it is not empty, the fault point is obtained. Otherwise, according to the actual detection value, the detection step associated with the next operation relationship of the first operation is obtained, and the actual detection value of the detection step is continued to be obtained until the fault point of the detection step corresponding to the actual detection value is not empty.

[0027] Based on a further improvement of the above method, the classification model is trained based on historical diagnostic data, historical failure analysis records, and historical maintenance records generated by multiple failure analyses. The sample detection process of each failure analysis of each detection process is constructed according to the sample detection process of each failure analysis of each detection process. The sample detection process of each failure analysis of each detection process includes:

[0028] If it is the first failure analysis, according to the parts associated with each inspection step and the circuit to which they belong under the corresponding inspection process, the inspection steps associated with other parts belonging to the same circuit are added to the original inspection steps, and the relationship between the inspection steps is updated to obtain the sample inspection process; otherwise, according to the inspection results associated with each inspection step under the corresponding inspection process, the fault point corresponding to the judgment result is obtained, and the inspection steps associated with all parts on the adjacent circuits of the circuit to which each fault point belongs are added to the original inspection steps in turn, and the relationship between the inspection steps is updated to obtain the sample inspection process.

[0029] Based on the further improvement of the above method, according to the sample detection process of each failure analysis of each detection process, a sample set of each detection process is constructed, including:

[0030] Based on each piece of historical diagnostic data generated by each failure analysis, multiple test item names and their test values ​​are sequentially obtained; the test item names are converted into test codes, the test codes and their test values ​​are combined into a pair of diagnostic information, and multiple pairs of diagnostic information are sequentially spliced ​​to obtain a piece of diagnostic data for each failure analysis;

[0031] Based on historical failure analysis records, obtain the failure analysis record corresponding to each diagnostic data of each failure analysis, match the actual test value with the sample test steps, obtain the fault point of each failure analysis, and obtain the fault point that was successfully repaired based on the corresponding historical maintenance records;

[0032] Each piece of diagnostic data of the same assembly in the same failure analysis, the actual detection value corresponding to the sample detection step, and the fault point are spliced ​​as one-time detection information of the assembly. According to the preset maximum number of failure analyses, multiple detection information are spliced ​​as a sample, and the corresponding successfully repaired fault point is used as the classification result and placed in the sample set of the corresponding detection process.

[0033] On the other hand, an embodiment of the present invention provides a fault point prediction system based on multiple failure analysis, including:

[0034] The inspection step acquisition module is used to analyze historical maintenance records to obtain abnormal points in the fault points. Based on the parts and lines associated with all inspection steps in each inspection process, the inspection steps related to the abnormal points are deleted and the inspection steps associated with other parts belonging to the same line are added to obtain the actual inspection steps of each inspection process.

[0035] The failure analysis module is used to perform failure analysis based on the diagnostic data of the assembly to be tested. If it is the first failure analysis, the actual test steps of the corresponding test process are obtained. Otherwise, the actual test steps of the corresponding test process are updated based on the failure point of the previous failure analysis. The failure point is obtained based on the actual test value obtained. If the failure point is correct, the failure analysis is terminated. Otherwise, the next failure analysis is iterated until the maximum number of failure analyses is reached to obtain the final failure point.

[0036] The fault point prediction module is used to construct the data to be predicted based on multiple failure analyses when the final fault point obtained by multiple failure analyses is incorrect, input it into the corresponding trained classification model, and predict the fault point.

[0037] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects: analyzing abnormal points from historical maintenance records, deleting detection steps related to the abnormal points, and accelerating the location of the fault point; as the number of failure analyses increases, new detection steps are added according to the fault point to avoid incomplete parts being detected, or damage to surrounding parts caused by the repair of the last faulty part, which makes it impossible to accurately locate the fault point, making the detection steps more complete and achieving comprehensive failure analysis; at the same time, the classification model is trained using historical data generated by multiple failure analyses, and after multiple failure analyses, the final repair attempt is provided to the maintenance engineer, maximizing the value of historical data and making failure analysis more automatic and convenient.

[0038] In the present invention, the above-mentioned technical solutions can be combined with each other to achieve more preferred combinations. Other features and advantages of the present invention will be described in the following description, and some advantages will become apparent from the description or be learned through practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained through the contents particularly pointed out in the description and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The accompanying drawings are only for the purpose of illustrating particular embodiments and are not to be considered limiting of the present invention. Like reference symbols denote like parts throughout the drawings.

[0040] Figure 1 This is a flow chart of a fault point prediction method based on multiple failure analysis in Example 1 of the present invention. DETAILED DESCRIPTION

[0041] The preferred embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, and are not used to limit the scope of the present invention.

[0042] Example 1

[0043] A specific embodiment of the present invention discloses a method for predicting fault points based on multiple failure analysis, such as Figure 1 As shown, the following steps are included:

[0044] S11: Analyze historical maintenance records to obtain abnormal points in the fault points. Based on all detection steps of each detection process, delete the detection steps related to the abnormal points to obtain the actual detection steps of each detection process.

[0045] It should be noted that the maintenance engineer's failure analysis of the defective assembly based on diagnostic data and the standard failure analysis process is recorded in the historical failure analysis record. This record details the failure analysis process code, the defective assembly code, the test values ​​at each step, and the final failure point (i.e., component). The historical maintenance record is a record of the maintenance engineer's repairs based on the failure point in the historical failure analysis record, and records whether the failure point was repaired and the repair time.

[0046] Specifically, we analyze historical maintenance records to identify abnormal points in the fault points, including:

[0047] Based on historical maintenance records, the yield of the fault points is statistically calculated by cycle; the fault points whose yield does not meet the Sigma principle are removed to obtain the fault points to be clustered; the density clustering algorithm is used to cluster the fault points according to the yield of the fault points to be clustered in the same cycle and the preset neighborhood radius to obtain the cluster categories; the fault points in the category with a number of fault points less than the number threshold are regarded as outliers; the outliers with a yield less than the minimum yield threshold are regarded as abnormal points.

[0048] It should be noted that the yield rate of a fault point, i.e., the probability of a fault point being repaired, is calculated periodically based on whether the fault point has been repaired in historical maintenance records. This period can be monthly or weekly, and the statistical range is determined based on actual maintenance conditions, such as the number of historical maintenance records and the frequency of repairs. For example, the yield rate of a fault point is calculated for six consecutive months.

[0049] After obtaining the yield of all fault points, we apply the three-sigma principle based on a normal distribution. For example, if the yield of a fault point is less than the mean minus three sigma, it is removed from the fault point list. The remaining fault points are clustered using the DBSCAN algorithm based on their yield over the past month. Based on the clustering results, if the number of fault points in a category is less than a threshold, the fault points in that category are considered outliers.

[0050] The yield of the outlier is compared with the minimum yield threshold again. If it is less than the minimum yield threshold, it is regarded as an outlier.

[0051] It should be noted that all the inspection steps of each inspection process already exist in the failure analysis knowledge graph. When constructing the failure analysis knowledge graph, the inspection steps and inspection results under each inspection process are extracted from the manually written historical failure analysis files. Among them, the inspection object involved in each inspection step is the part entity that has been established in the knowledge graph, and the association relationship established with the corresponding inspection step, and the part entity is associated with the line entity, so that the line to which the part belongs can be obtained. When the inspection step entity is instantiated, the next operation relationship between the inspection step entities has been initialized, including the next operation when the inspection is normal and the next operation when the inspection is abnormal. The fault points corresponding to normal and / or abnormal inspections are given in the inspection results of each inspection step. For the operation step code of the first inspection step of each inspection process, it is set to the preset operation step code, such as A001. The operation step codes of other inspection steps can be set according to preset rules.

[0052] For example, Table 1 shows a standard failure analysis process generated based on the failure analysis knowledge graph. The step numbers and operation types in Table 1 are directly derived from the attribute values ​​of the detection step entity. The operation step description is filled in the operation step description template corresponding to the visual inspection and measurement based on the associated part entity. The next step is the step number of the detection step entity corresponding to the next operation of the normal detection (0) and / or abnormal detection (1). The fault point is the fault part corresponding to the detection result (0) when the detection is normal and / or the detection result (1) when the detection is abnormal based on the associated detection results.

[0053] Table 1 Example of standard failure analysis process

[0054]

[0055]

[0056] In this embodiment, based on the lines to which all detection steps of each detection process belong, the detection steps related to the outliers are deleted, and the detection steps belonging to the same line as the remaining detection steps are added to obtain the actual detection steps of each detection process, including:

[0057] Based on the failure analysis knowledge graph, the parts and associated lines of each inspection step in each inspection process are obtained according to the part entities associated with each inspection step entity and the line entities associated with the part entities. The name of the outlier point is used as the part entity name to query the part entity corresponding to the outlier point and its associated line entities to obtain the line to which the outlier point belongs.

[0058] In all the inspection steps of each inspection process, delete the inspection steps corresponding to the abnormal point and the inspection steps associated with the parts belonging to the same line as the abnormal point;

[0059] For each inspection process, the inspection steps associated with other parts belonging to the same circuit are added according to the parts associated with the remaining inspection steps and the circuits they belong to, and the relationship between the inspection steps is updated to obtain the actual inspection steps.

[0060] It should be noted that updating the relationship between the detection steps is to establish the next operation relationship between the detection step of normal or abnormal detection corresponding to the next detection step based on the detection result associated with the entity of the previous detection step and the abnormal or normal determination result in the detection result, including:

[0061] When the judgment result corresponding to the previous detection step is normal and the fault point is not empty, the previous detection step and the next detection step establish a next operation relationship of abnormal detection; when the judgment result corresponding to the previous detection step is abnormal and the fault point is not empty, the previous detection step and the next detection step establish a next operation relationship of normal detection; if the fault point corresponding to the judgment result corresponding to the current detection step is empty, the previous detection step entity and the next detection step entity simultaneously establish next operation relationships of normal detection and abnormal detection; if the current detection step entity has both normal and abnormal judgment results and both fault points are not empty, the fault point corresponding to the clear judgment result is normal, and the previous detection step and the next detection step establish a next operation relationship of normal detection.

[0062] For example, if part R3300 is an outlier and parts R3301 and R3300 belong to the same circuit, then A002, A003, and A007 are deleted from Table 1. The remaining inspection steps associated with parts C9400, C9401, C9402, and C9403 belong to another circuit. According to the failure analysis knowledge graph, there is also part C9404 on this circuit. Therefore, the inspection steps associated with part C9404 need to be added to Table 1 and placed after step A006. When updating the relationship between inspection steps, since the judgment result corresponding to A001 is abnormal and the fault point is C9400, a next-step operation relationship is established with the subsequent A004, indicating that the inspection is normal. The fault point C9404 corresponding to the normal judgment result of step A006 is cleared, and a next-step operation relationship is established with the added step, indicating that the inspection is normal, thus obtaining the actual inspection steps for the first analysis.

[0063] It should be noted that the yield rate of the abnormal point is very low, that is, when the abnormal point is located as the fault point, the assembly is difficult to repair. Therefore, the steps involving the abnormal point and the detection steps belonging to the same line as the abnormal point are deleted from all the detection steps to reduce the interfering detection steps and streamline the detection process, so that the maintenance engineer can locate the exact fault point according to the remaining detection steps.

[0064] S12: Perform failure analysis based on the diagnostic data of the assembly to be tested. If it is the first failure analysis, obtain the actual detection steps of the corresponding detection process. Otherwise, update the actual detection steps of the corresponding detection process based on the fault point of the previous failure analysis; obtain the fault point based on the actual detection value obtained. If the fault point is correct, end the failure analysis. Otherwise, iterate the next failure analysis until the maximum number of failure analyses is reached and the final fault point is obtained.

[0065] It should be noted that the diagnostic data of the assembly to be tested includes the names of the test items and their test values ​​corresponding to the multiple fault phenomena respectively, and the corresponding test process is obtained according to the first test item name.

[0066] In this embodiment, when performing the first failure analysis, actual detection steps that exclude detection steps related to abnormal points are used. If the fault point obtained in the first failure analysis is verified by the maintenance engineer to be unable to modify the assembly, then when performing the second and subsequent failure analyses, the fault point obtained in the previous failure analysis will be used to supplement the detection steps corresponding to the parts surrounding the fault point, so as to avoid the inability to accurately locate the fault point due to incomplete parts being tested in the actual detection process, or damage to surrounding parts caused by the repair of the previous fault part.

[0067] It should be noted that according to the fault point of the previous failure analysis, the actual detection steps of the corresponding detection process are updated, including:

[0068] Based on the fault point of the last failure analysis, obtain the inspection steps associated with all parts on the adjacent lines of the line to which the fault point belongs, put them into the set to be supplemented in order, add the set to be supplemented to the actual inspection steps of the corresponding inspection process, and update the relationship between the inspection steps.

[0069] It should be noted that the detection steps to be added are related to the fault point of the previous failure analysis. Therefore, this embodiment disconnects the relationship between the detection steps corresponding to the fault point of the previous failure analysis in the actual detection process and the detection steps associated with its next operation, and establishes a relationship with the detection steps in the set to be added, so that during the next failure analysis, after executing the detection steps corresponding to the fault point of the previous failure analysis, the detection steps added to the set to be added are executed, and then the original next operation is returned to.

[0070] For example, in an automotive part failure analysis, the first failure analysis identified the brake as the fault point. A maintenance engineer repaired the brake but was unsuccessful. A knowledge graph association query revealed that the brake drum, friction pad, and drum-to-pad contact were located near the brake. Therefore, additional testing of the brake drum, friction pad, and drum-to-pad contact was added to the first failure analysis, making the testing more comprehensive and enabling a more comprehensive failure analysis.

[0071] When performing failure analysis, the fault point is obtained based on the actual test values, including:

[0072] Based on the failure analysis knowledge graph, the inspection step with the preset first operation step code is used as the first operation;

[0073] The actual detection value of the first operation is taken as the judgment result. According to the detection result associated with the first operation, it is identified whether the fault point corresponding to the judgment result is empty. If it is not empty, the fault point is obtained. Otherwise, according to the actual detection value, the detection step associated with the next operation relationship of the first operation is obtained, and the actual detection value of the detection step is continued to be obtained until the fault point of the detection step corresponding to the actual detection value is not empty.

[0074] For example, in Table 1, the preset first step operation step code, such as A001, is first displayed, and the first step operation step description is displayed. The maintenance engineer visually inspects the part C9400. If there is a problem, the actual detection value 1 is input according to the operation step description. At this time, 1 is used as the abnormal judgment result of the first step, which corresponds to the fault point C9400 in Table 1, and a failure analysis is completed. If there is no problem, the actual detection value 0 is input according to the operation step description. At this time, there is no fault point, and the next step operation corresponding to normal detection is A002, so the operation step description of A002 is displayed. The maintenance engineer then inputs the actual detection value based on the visual inspection situation, and according to the fault point or the next operation relationship corresponding to A002, until the fault point is obtained, and a failure analysis is completed.

[0075] It should be noted that the maximum number of failure analyses is determined by the actual failure analysis area and the quality of the assembly. For example, if the assembly is a motherboard, the maximum number of failures is set to 3.

[0076] S13: If the final fault point is incorrect, the data to be predicted is constructed based on multiple failure analyses and input into the corresponding trained classification model to predict the fault point.

[0077] It should be noted that the classification model is based on historical diagnostic data, historical failure analysis records, and historical maintenance records generated by multiple failure analyses. It is trained by constructing a sample set for each detection process according to the sample detection process of each failure analysis of each detection process. The classification models include: random forest, Xgboost, and graph neural network (GNN).

[0078] It should be noted that in actual failure analysis, the first failure analysis is the step after removing the abnormal point related detection steps in all detection steps, and then the detection steps related to other parts belonging to the same circuit are added. Subsequent failure analysis is the step of adding in the dynamic fault points. Therefore, when training the classification model, all steps of each failure analysis must be considered. The sample detection process of each failure analysis in each detection process includes:

[0079] If it is the first failure analysis, according to the parts associated with each inspection step and the circuit to which they belong under the corresponding inspection process, the inspection steps associated with other parts belonging to the same circuit are added to the original inspection steps, and the relationship between the inspection steps is updated to obtain the sample inspection process; otherwise, according to the inspection results associated with each inspection step under the corresponding inspection process, the fault point corresponding to the judgment result is obtained, and the inspection steps associated with all parts on the adjacent circuits of the circuit to which each fault point belongs are added to the original inspection steps in turn, and the relationship between the inspection steps is updated to obtain the sample inspection process.

[0080] It should be noted that if at least two failure analyses are performed in actual situations, there is no need to train the model for the first failure analysis.

[0081] When constructing the sample set, we select historical data based on diagnostic data, failure analysis records, and repair records. Fault points that have undergone multiple failure analyses and have been successfully repaired are not subject to the maximum number of failure analyses. If the number of failure analyses for certain assemblies in the historical data is less than the maximum number, we can supplement the previous sample data with the current sample data format to increase the sample size and improve the classification accuracy of the model.

[0082] Specifically, according to the sample inspection process of each failure analysis of each inspection process, a sample set of each inspection process is constructed, including:

[0083] ① Based on each piece of historical diagnostic data generated by each failure analysis, obtain the names of multiple test items and their test values ​​in sequence; convert the test item names into test codes, and form a pair of diagnostic information with the test codes and their test values. Multiple pairs of diagnostic information are spliced ​​in sequence to obtain a piece of diagnostic data for each failure analysis.

[0084] For example, a piece of diagnostic data is a-1c-5e10g100w5, where "acegw" is a detection code with a sequence, and "-1-5101005" are detection values ​​corresponding to "acegw" respectively.

[0085] ② Based on historical failure analysis records, obtain the failure analysis record corresponding to each diagnostic data of each failure analysis, correspond the actual detection value to the sample detection step, obtain the fault point of each failure analysis, and obtain the fault point that was successfully repaired based on the corresponding historical maintenance records.

[0086] It should be noted that the actual test values ​​in historical failure analysis records are typically 0 or 1. Depending on the actual test values, different test steps are executed. Actual test values ​​for test steps that were not executed are left blank. Based on the step code, the actual test values ​​are mapped to the sample test steps. Samples with a null value ratio greater than the null value threshold are deleted. For samples with a null value ratio less than the null value threshold, the null values ​​are filled with the mode. The mode is calculated based on the test records of each test process.

[0087] For example, the procedure for step 1 is as follows: Visually inspect the C9400. Enter "1" if there is a problem, and "0" if there is no problem. If you enter 1, proceed to step 2; if you enter 0, proceed to step 3. If the maintenance engineer enters an actual test value of 0, step 3 is executed directly, and the actual test value corresponding to step 2 is left blank. If there are many actual test values ​​of 1 under this inspection process, the actual test value of 1 is added to step 2.

[0088] ③ Each diagnostic data of the same assembly in the same failure analysis, the actual detection value corresponding to the sample detection step, and the fault point are spliced ​​as one-time detection information of the assembly. According to the preset maximum number of failure analyses, multiple detection information are spliced ​​as a sample, and the corresponding successfully repaired fault point is used as the classification result and placed in the sample set of the corresponding detection process.

[0089] The constructed sample set is divided into a training set and a test set. The training set is used to train the classification model, and the test set is used to test the performance and calculate the accuracy of the analysis model to prevent the network from overfitting or undertraining the training dataset. These are common usages and will not be explained in detail.

[0090] If the final fault point obtained in step S12 is incorrect, the data to be predicted is constructed based on multiple failure analyses and input into the corresponding trained classification model to predict the fault point.

[0091] It should be noted that the data to be predicted is constructed in the format of a sample set according to the sample detection steps of each failure analysis of the corresponding detection process, including:

[0092] Obtain diagnostic data, actual test values ​​and failure points of the assembly to be tested obtained from each failure analysis;

[0093] According to the detection step operation code, the actual detection value of each failure analysis input is matched to the sample detection step of the current detection process. For the sample detection step that cannot be matched, it is filled according to the mode of the sample set of the current detection process to obtain the detection data;

[0094] The diagnostic data, test data and fault points of the assembly to be tested obtained from each failure analysis are sequentially spliced ​​together to obtain the data to be predicted.

[0095] Compared with the prior art, the fault point prediction method based on multiple failure analyses in this embodiment analyzes abnormal points from historical maintenance records, deletes detection steps related to the abnormal points, and speeds up the location of the fault points. As the number of failure analyses increases, new detection steps are added according to the fault points to avoid incomplete parts being detected or damage to surrounding parts caused by the repair of the previous faulty part, which may result in the inability to accurately locate the fault point. This makes the detection steps more complete and achieves comprehensive failure analysis. At the same time, the classification model is trained using historical data generated by multiple failure analyses, and after multiple failure analyses, the classification model is provided to the maintenance engineer for the final repair attempt, thereby maximizing the value of the historical data and making failure analysis more automatic and convenient.

[0096] Example 2

[0097] Another embodiment of the present invention discloses a fault point prediction system based on multiple failure analysis, thereby implementing the fault point prediction method of multiple failure analysis in Example 1. The specific implementation of each module refers to the corresponding description in Example 1. The system includes:

[0098] The inspection step acquisition module is used to analyze historical maintenance records to obtain abnormal points in the fault points. Based on the parts and lines associated with all inspection steps in each inspection process, the inspection steps related to the abnormal points are deleted and the inspection steps associated with other parts belonging to the same line are added to obtain the actual inspection steps of each inspection process.

[0099] The failure analysis module is used to perform failure analysis based on the diagnostic data of the assembly to be tested. If it is the first failure analysis, the actual test steps of the corresponding test process are obtained. Otherwise, the actual test steps of the corresponding test process are updated based on the failure point of the previous failure analysis. The failure point is obtained based on the actual test value obtained. If the failure point is correct, the failure analysis is terminated. Otherwise, the next failure analysis is iterated until the maximum number of failure analyses is reached to obtain the final failure point.

[0100] The fault point prediction module is used to construct the data to be predicted based on multiple failure analyses when the final fault point obtained by multiple failure analyses is incorrect, input it into the corresponding trained classification model, and predict the fault point.

[0101] Since the fault point prediction system based on multiple failure analysis in this embodiment can be mutually referenced with the aforementioned generation method, it is redundant and will not be repeated here. Since the principles of this system embodiment and the aforementioned method embodiment are the same, this system embodiment also has the corresponding technical effects of the aforementioned method embodiment.

[0102] Those skilled in the art will appreciate that all or part of the process steps of the above-described embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, such as a magnetic disk, an optical disk, a read-only memory, or a random access memory.

[0103] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by any technician familiar with this technical field within the technical scope disclosed by the present invention should be covered by the scope of protection of the present invention.

Claims

1. A method for predicting fault points based on multiple failure analysis, characterized in that: The steps include: Analyze historical maintenance records to identify anomalies in fault points. Based on the parts and lines associated with all inspection steps in each inspection process, delete the inspection steps related to the anomalies and add the inspection steps associated with other parts belonging to the same line to obtain the actual inspection steps of each inspection process. Perform failure analysis based on the diagnostic data of the assembly to be tested. If this is the first failure analysis, obtain the actual test steps of the corresponding test process. Otherwise, update the actual test steps of the corresponding test process based on the fault point of the previous failure analysis. Obtain the fault point based on the obtained actual test value. If the fault point is correct, end the failure analysis. Otherwise, iterate the next failure analysis until the maximum number of failure analyses is reached and the final fault point is obtained. If the final fault point is incorrect, the data to be predicted is constructed based on multiple failure analyses and input into the corresponding trained classification model to predict the fault point.

2. The method for predicting a fault point based on multiple failure analysis according to claim 1, characterized in that: The analysis of historical maintenance records to obtain abnormal points in fault points includes: Based on historical maintenance records, the yield rate of failure points is calculated periodically; Remove the fault points whose yield does not meet the Sigma principle and obtain the fault points to be clustered; Using the density clustering algorithm, the fault points to be clustered are clustered according to their yield rate in the same cycle and the preset neighborhood radius to obtain cluster categories. The fault points in the category with a number of fault points less than the threshold are regarded as outliers. Outliers whose yield is less than the minimum yield threshold are regarded as abnormal points.

3. The method for predicting fault points based on multiple failure analysis according to claim 2, characterized in that: The parts and their respective lines associated with all the inspection steps of each inspection process are deleted, and the inspection steps associated with the abnormal points are added, so as to obtain the actual inspection steps of each inspection process, including: Based on the failure analysis knowledge graph, the parts associated with each inspection step and the associated line entities are obtained under each inspection process. The name of the abnormal point is used as the part entity name, and the part entity corresponding to the abnormal point and its associated line entity are queried to obtain the line to which the abnormal point belongs. In all the inspection steps of each inspection process, delete the inspection steps corresponding to the abnormal point and the inspection steps associated with the parts belonging to the same line as the abnormal point; For each inspection process, the inspection steps associated with other parts belonging to the same circuit are added according to the parts associated with the remaining inspection steps and their corresponding circuits, and the relationship between the inspection steps is updated to obtain the actual inspection steps of each inspection process.

4. The method for predicting a fault point based on multiple failure analysis according to claim 1, characterized in that: The diagnostic data of the assembly to be inspected includes inspection item names and inspection values ​​corresponding to a plurality of fault phenomena, and the corresponding inspection process is obtained according to the first inspection item name.

5. The method for predicting a fault point based on multiple failure analysis according to claim 1, characterized in that: The actual detection steps of updating the corresponding detection process according to the fault point of the previous failure analysis include: Based on the fault point of the last failure analysis, obtain the inspection steps associated with all parts on the adjacent lines of the line to which the fault point belongs, put them into the set to be supplemented in order, add the set to be supplemented to the actual inspection steps of the corresponding inspection process, and update the relationship between the inspection steps.

6. The method for predicting fault points based on multiple failure analysis according to claim 5, characterized in that: The relationship between the update detection steps includes: Obtain the detection step corresponding to the fault point as the first operation step, and obtain the corresponding detection step according to the next operation relationship of the first operation step as the second operation step; associate the next operation relationship of the first operation step to the first detection step in the set to be supplemented, and associate the next operation relationship of the last detection step in the set to be supplemented to the second operation step.

7. The method for predicting a fault point based on multiple failure analysis according to claim 1, characterized in that: Determining the fault point based on the actual detection value obtained includes: Based on the actual detection steps of the corresponding detection process, the detection step with the preset first operation step code is used as the first operation; The actual detection value of the first operation is taken as the judgment result. According to the detection result associated with the first operation, it is identified whether the fault point corresponding to the judgment result is empty. If it is not empty, the fault point is obtained. Otherwise, according to the actual detection value, the detection step associated with the next operation relationship of the first operation is obtained, and the actual detection value of the detection step is continued to be obtained until the fault point of the detection step corresponding to the actual detection value is not empty.

8. The method for predicting fault points based on multiple failure analysis according to claim 1, characterized in that: The classification model is trained by constructing a sample set for each detection process based on historical diagnostic data, historical failure analysis records, and historical maintenance records generated by multiple failure analyses, according to the sample detection process of each failure analysis of each detection process; The sample testing process for each failure analysis of each testing process includes: If this is the first failure analysis, based on the parts and circuits associated with each inspection step in the corresponding inspection process, the inspection steps associated with other parts belonging to the same circuit are added to the original inspection steps, and the relationship between the inspection steps is updated to obtain a sample inspection process. Otherwise, according to the test results associated with each test step under the corresponding test process, the fault point corresponding to the judgment result is obtained, and the test steps associated with all parts on the adjacent lines of the line to which each fault point belongs are added to the original test steps in turn, and the relationship between the test steps is updated to obtain the sample test process.

9. The method for predicting fault points based on multiple failure analysis according to claim 8, characterized in that: The sample detection process of each failure analysis of each detection process is used to construct a sample set of each detection process, including: Based on each piece of historical diagnostic data generated by each failure analysis, multiple test item names and their test values ​​are sequentially obtained; the test item names are converted into test codes, the test codes and their test values ​​are combined into a pair of diagnostic information, and multiple pairs of diagnostic information are sequentially spliced ​​to obtain a piece of diagnostic data for each failure analysis; Based on historical failure analysis records, obtain the failure analysis record corresponding to each diagnostic data of each failure analysis, match the actual test value with the sample test steps, obtain the fault point of each failure analysis, and obtain the fault point that was successfully repaired based on the corresponding historical maintenance records; Each piece of diagnostic data of the same assembly in the same failure analysis, the actual detection value corresponding to the sample detection step, and the fault point are spliced ​​as one-time detection information of the assembly. According to the preset maximum number of failure analyses, multiple detection information are spliced ​​as a sample, and the corresponding successfully repaired fault point is used as the classification result and placed in the sample set of the corresponding detection process.

10. A fault point prediction system based on multiple failure analysis, characterized in that: include: The inspection step acquisition module is used to analyze historical maintenance records to obtain abnormal points in the fault points. Based on the parts and lines associated with all inspection steps in each inspection process, the inspection steps related to the abnormal points are deleted and the inspection steps associated with other parts belonging to the same line are added to obtain the actual inspection steps of each inspection process. The failure analysis module is used to perform failure analysis based on the diagnostic data of the assembly to be tested. If it is the first failure analysis, the actual test steps of the corresponding test process are obtained. Otherwise, the actual test steps of the corresponding test process are updated based on the failure point of the previous failure analysis. The failure point is obtained based on the actual test value obtained. If the failure point is correct, the failure analysis is terminated. Otherwise, the next failure analysis is iterated until the maximum number of failure analyses is reached to obtain the final failure point. The fault point prediction module is used to construct the data to be predicted based on multiple failure analyses when the final fault point obtained by multiple failure analyses is incorrect, input it into the corresponding trained classification model, and predict the fault point.

Images