Fault early warning key data screening method and device, equipment and medium

Abstract

The invention discloses a fault early warning key data screening method, device and equipment and a medium, and the method comprises the steps: obtaining historical fault data of a plurality of vehicles, carrying out the data integration of the historical fault data of the plurality of vehicles, building a standard fault data set, and storing the standard fault data set in a database; original dichotomy fault tags of the data in the standard fault data set are expanded into trichotomy fault tags, and the trichotomy fault tags comprise normal, early warning and fault three types of fault tags; preliminarily screening data in the standard fault data set, and removing abnormal values and redundant values of data with normal fault labels in the standard fault data set to obtain a preliminarily screened data set; and based on a pre-constructed multi-strategy comprehensive evaluation method, evaluating the data in the preliminary screening data set, and screening out key data from the preliminary screening data set according to an evaluation result. According to the method, the early recognition capability and the cross-vehicle adaptability of the fault early warning system are remarkably improved, and a technical support is provided for the safety guarantee of the new energy automobile.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of vehicle fault early warning, and in particular to a fault early warning key data screening method, device, equipment and medium. BACKGROUND

[0002] With the widespread application of new energy vehicles, their safety and reliability have become the focus of public attention. The vehicle-mounted system of new energy vehicles integrates a large number of sensors and controllers, and these devices can generate a large amount of multi-dimensional signal data, which not only records the running state of the vehicle, but also may contain potential fault information. However, these signal data generally have problems such as high dimension, signal redundancy, high correlation between signals, and poor quality of some signals. If all signals are input into the model as early warning data for training, a large amount of noise will be inevitably introduced, which will interfere with the training of the model and reduce the recognition ability of the model to key early warning signals. Moreover, the existing feature screening methods mostly perform signal dimension reduction based on a single screening standard, which cannot meet the demand of new energy vehicle fault early warning systems for the stability and controllability of key signals, especially in the scene of cross-vehicle data, the performance of the model fails to achieve the expected effect.

[0003] Therefore, how to efficiently and accurately screen out key signals that have significant changes before the occurrence of faults and are highly correlated with faults from the multi-dimensional signals of new energy vehicles has become a technical problem to be solved in the field of vehicle fault early warning. SUMMARY

[0004] In view of the above problems, the present application provides a fault early warning key data screening method, device, equipment and medium, which realizes the complete retention of fault precursor signals by accurately cleaning normal data and introducing and protecting early warning labels in the preliminary screening stage, and efficiently and accurately screens out key signals that are strongly correlated with faults by combining multi-strategy comprehensive evaluation, solves the noise interference and single screening standard limitation problems mentioned in the background art, significantly improves the early identification ability of the fault early warning system and the cross-vehicle adaptability, and provides technical support for the safety guarantee of new energy vehicles.

[0005] In a first aspect, a fault early warning key data screening method is provided, comprising:

[0006] obtaining historical fault data of a plurality of vehicles, and integrating the historical fault data of the plurality of vehicles to establish a standard fault data set, the standard fault data set comprising original binary classification fault labels corresponding to each piece of historical fault data;

[0007] label reconstruction is performed on data in the standard fault data set, and the original binary fault label of the data is expanded to a ternary fault label, the original binary fault label including two types of normal and fault, and the ternary fault label including three types of normal, early warning and fault;

[0008] preliminary screening is performed on data in the standard fault data set, and abnormal values and redundant values of each target data in the standard fault data set are removed to obtain a preliminary screening data set, wherein the target data is data with a normal fault label in the standard fault data set;

[0009] Based on the pre-constructed multi-strategy comprehensive evaluation method, the data in the preliminary screening data set is evaluated, and key data is selected from the preliminary screening data set according to the evaluation result.

[0010] In some implementations, the standard fault data set includes a time stamp and a vehicle identification code corresponding to each piece of historical fault data, and the data integration of the historical fault data of the plurality of vehicles to establish the standard fault data set includes:

[0011] The historical fault data of the plurality of vehicles is converted to unify the data types of the historical fault data of the plurality of vehicles;

[0012] Based on the vehicle identification code and the time stamp, the historical fault data of the plurality of vehicles is time-synchronized to sort and integrate the historical fault data of different vehicles in chronological order into a standard fault data set.

[0013] In some implementations, the label reconstruction is performed on data in the standard fault data set, and the original binary fault label of the data is expanded to a ternary fault label, including:

[0014] For the historical fault data with a time stamp within T time before the fault occurs, the original binary label is reconstructed as an early warning label, and the remaining data remains unchanged.

[0015] In some implementations, the standard fault data set includes measurement values of each signal collected at a corresponding time for a plurality of historical fault data; and the preliminary screening is performed on data in the standard fault data set to remove abnormal values of each target data in the standard fault data set, including:

[0016] It is determined whether there is an outlier measurement value in the sample data, the sample data being any one of the target data in the standard fault data set;

[0017] When there is an outlier measurement value in the sample data, the proportion of the number of outlier measurement values in the sample data is determined.

[0018] when the proportion of the number of outliers in the sample data is less than or equal to a preset outlier proportion threshold, removing the outliers in the sample data;

[0019] when the proportion of the number of outliers in the sample data is greater than a preset outlier proportion threshold, determining the overall deviation degree of all measurement values in the sample data;

[0020] when the overall deviation degree is greater than a preset deviation proportion threshold, removing the sample data.

[0021] In some implementations, the standard fault data set includes measurement values of each signal collected at a corresponding time by a plurality of historical fault data; and the preliminary screening of the data in the standard fault data set to remove redundant values of each target data in the standard fault data set includes:

[0022] determining the variance of the measurement values of each signal in the sample data, and removing the measurement values with a corresponding variance greater than a preset variance threshold, the sample data being any one target data in the standard fault data set;

[0023] performing correlation analysis on the measurement values of the same type of signals in two sample data to determine the correlation between the measurement values of the two same type of signals;

[0024] when the correlation between the measurement values of the two same type of signals is greater than a preset correlation threshold, removing one of the measurement values.

[0025] In some implementations, before evaluating the data in the preliminary screening data set based on the pre-constructed multi-strategy comprehensive evaluation method, the method further includes:

[0026] using a linear interpolation method to numerically fill the data with missing measurement values in the preliminary screening data set;

[0027] using a label forward filling method to fill the fault labels for the data with missing fault labels in the preliminary screening data set.

[0028] In some implementations, based on the pre-constructed multi-strategy comprehensive evaluation method, the data in the preliminary screening data set is evaluated, and key data is selected from the preliminary screening data set according to the evaluation result, including:

[0029] inputting each data in the preliminary screening data set into a preset multi-strategy comprehensive evaluation model, the multi-strategy comprehensive evaluation model being based on a pre-constructed multi-strategy comprehensive evaluation method to determine a supervision score, a stability score and a fault sensitivity score of each data;

[0030] The supervised score, stability score and failure sensitivity score of each data in the preliminary screening data set are weighted and summed to obtain a comprehensive evaluation index value corresponding to each data;

[0031] The data whose comprehensive evaluation index value is greater than a preset comprehensive evaluation threshold value is determined as the key data.

[0032] In a second aspect, a fault early warning key data screening device is provided, comprising:

[0033] A data integration module is configured to obtain historical fault data of a plurality of vehicles and integrate the historical fault data of the plurality of vehicles to establish a standard fault data set, wherein the standard fault data set comprises original binary classification fault labels corresponding to each piece of historical fault data.

[0034] A label reconstruction module is configured to reconstruct labels of data in the standard fault data set, and expand the original binary classification fault labels of the data into ternary classification fault labels, wherein the original binary classification fault labels comprise two types of fault labels, i.e., normal and fault, and the ternary classification fault labels comprise three types of fault labels, i.e., normal, early warning and fault.

[0035] A data preliminary screening module is configured to preliminarily screen data in the standard fault data set, and remove abnormal values and redundant values of each target data in the standard fault data set to obtain a preliminary screening data set, wherein the target data is data with a normal fault label in the standard fault data set.

[0036] A data fine screening module is configured to evaluate data in the preliminary screening data set based on a pre-constructed multi-strategy comprehensive evaluation method, and screen key data from the preliminary screening data set according to an evaluation result.

[0037] In a third aspect, an electronic device is provided, comprising a memory and a processor, which are communicatively connected to each other, and the memory stores computer instructions, and the processor executes the computer instructions to perform the fault early warning key data screening method according to the first aspect.

[0038] In a fourth aspect, a computer readable storage medium is provided, which stores computer instructions for causing the computer to execute the fault early warning key data screening method according to the first aspect.

[0039] The technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages:

[0040] The fault early warning key data screening method, device, equipment and medium provided by the embodiment of the application can more accurately depict the vehicle state evolution process and provide more rich fault development information for subsequent analysis by establishing a standard fault data set through data integration and expanding the original binary classification fault label of the data into a three-classification label containing normal, early warning and fault. The abnormal values and redundant values of each target data with a normal fault label in the standard fault data set are removed through preliminary screening of the data in the standard fault data set, the loss of key fault features caused by directly processing early warning and fault label data is avoided, noise interference in the normal data is effectively reduced, and the data quality is significantly improved; the integrity of the early warning label data is retained, the early features of the potential fault mode are completely retained, the model can learn the signal change rule based on the complete fault development trajectory (normal -> early warning -> fault), and the identification ability of the model to early faults is enhanced. Finally, the preliminary screening data set is evaluated and key data is screened based on a multi-strategy comprehensive evaluation method, the limitations of the existing single screening standard are overcome, the evaluation of the importance of signals considering various factors is comprehensively considered, key signals with significant changes and high correlation before the occurrence of faults are more efficiently and accurately screened out, the demand of the new energy vehicle fault early warning system for the stability and controllability of key signals is met, the problem of insufficient model performance in the cross-vehicle data scene is solved, and the safety and operation stability of the new energy vehicle are improved.

[0041] The above description is only a summary of the technical solutions of the application. In order to enable the technical means of the application to be more clearly understood, the application can be implemented according to the content of the specification, and in order to enable the above and other purposes, characteristics and advantages of the application to be more obvious and easy to understand, the specific embodiments of the application are described below. BRIEF DESCRIPTION OF DRAWINGS

[0042] Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The accompanying drawings are included to provide a description of the preferred embodiments and are not meant to limit the present application. Moreover, the same reference numerals in the attached drawings indicate the same or similar components. In the drawings:

[0043] Figure 1 FIG. 1 is a method flowchart of a fault early warning key data screening method provided by an embodiment of the application;

[0044] Figure 2 FIG. 2 is a method flowchart of step S130;

[0045] Figure 3 FIG. 3 is a structural block diagram of a fault early warning key data screening device provided by an embodiment of the application. DETAILED DESCRIPTION

[0046] For better understanding of the above technical solutions, the above technical solutions will be described in detail below in combination with the drawings and specific embodiments of the specification. It should be understood that the embodiments of the present disclosure and the specific features in the embodiments are detailed descriptions of the technical solutions of the present application, and are not limitations of the technical solutions of the present application. In the case of no conflict, the technical features in the embodiments of the present application and the embodiments can be combined with each other.

[0047] Figure 1 is a method flow chart of a fault early warning key data screening method provided by an embodiment of the present application, as shown in Figure 1 , the method comprises:

[0048] Step S110, obtaining historical fault data of a plurality of vehicles, and integrating the historical fault data of the plurality of vehicles to establish a standard fault data set, the standard fault data set comprising original binary classification fault labels corresponding to each piece of historical fault data.

[0049] Step S120, reconstructing the labels in the standard fault data set, and expanding the original binary classification fault labels of the data into ternary classification fault labels, the original binary classification fault labels comprising two types of fault labels of normal and fault, and the ternary classification fault labels comprising three types of fault labels of normal, early warning and fault.

[0050] Step S130, performing preliminary screening on the data in the standard fault data set, eliminating outliers and redundant values of each target data in the standard fault data set, and obtaining a preliminary screening data set, wherein the target data is data with a normal fault label in the standard fault data set.

[0051] Step S140, based on a pre-constructed multi-strategy comprehensive evaluation method, evaluating the data in the preliminary screening data set, and selecting key data from the preliminary screening data set according to the evaluation result.

[0052] The above method screens the key signals related to the fault precursor from the historical fault data of the vehicle through data integration, label reconstruction, signal rough screening and fine screening, which significantly improves the accuracy of fault early warning and the signal retrieval efficiency.

[0053] In some implementations, the standard fault data set comprises a time stamp and a vehicle identification code corresponding to each piece of historical fault data, and step S110 comprises:

[0054] Converting the historical fault data of the plurality of vehicles to make the data types of the historical fault data of the plurality of vehicles uniform; performing time synchronization processing on the historical fault data of the plurality of vehicles based on the vehicle identification code and the time stamp, and sorting and integrating the historical fault data of different vehicles in time sequence into the standard fault data set.

[0055] In a specific implementation, a standard fault data set table can be constructed to store multi-vehicle historical fault data in a structured manner. The standard fault data set table can adopt a two-dimensional relational design, with each row corresponding to a historical fault data as a minimum data record unit, and with each column corresponding to a field attribute unit to define a data characteristic dimension. Each column in the table can include a vehicle identification code, a timestamp, a fault label, and different signal names, and each historical fault data includes specific values under different fields. At the same time, a two-dimensional index can be constructed based on the vehicle identification code and the timestamp, and a time window sliding alignment algorithm (such as a nearest neighbor interpolation method based on a sliding window) can be used to perform time synchronization processing on the historical fault data of different vehicles. Specifically, the vehicle data is arranged in ascending order of the timestamp, and is discretely sampled according to a preset time granularity to ensure that the cross-vehicle data is comparable in the time dimension, and finally integrated into a standardized time series data set.

[0056] For example, bus signal data can be loaded in batches from data files of multiple vehicles to obtain historical fault data of the multiple vehicles. In view of the differences in data types in the historical fault data of different vehicles, a data conversion engine can be used to map all data fields to a unified data type standard to eliminate analysis bias caused by inconsistent data types.

[0057] In some implementations, step S120 includes:

[0058] For historical fault data with a timestamp within T time before the occurrence of a fault, the original binary label is reconstructed as a pre-warning label, and the remaining data remains unchanged.

[0059] For example, the pre-warning label is set based on a time window T before the occurrence of a fault of the current vehicle, and T can be set according to actual needs, for example, T = 5 min, to realize early perception of potential faults and ensure the accuracy and timeliness of the pre-warning.

[0060] Figure 2 is a method flowchart of step S130, as Figure 2 shown, step S130 includes preliminary screening of data in the standard fault data set to eliminate outliers of each target data in the standard fault data set, including:

[0061] Step S131, determining whether there is an outlier in the sample data, the sample data being any target data in the standard fault data set.

[0062] The standard fault data set includes measurement values of each signal collected at a corresponding time by a plurality of historical fault data.

[0063] Step S132, when there is an outlier measurement value in the sample data, determining a proportion of the number of outlier measurement values in the sample data.

[0064] The proportion of the number of outlier values in the sample data is the proportion of the number of outlier measurement values in the sample data to the number of all measurement values.

[0065] Step S133, when the proportion of the number of outlier measurement values in the sample data is less than or equal to a preset outlier proportion threshold, removing the outlier measurement value in the sample data.

[0066] Step S134, when the proportion of the number of outlier measurement values in the sample data is greater than the preset outlier proportion threshold, determining the overall deviation degree of all measurement values in the sample data.

[0067] Step S135, when the overall deviation degree is greater than a preset deviation proportion threshold, removing the sample data.

[0068] In some implementations, the determination of whether there is an outlier measurement value in the sample data in step S131 includes:

[0069] The Z-score method is used to calculate the Z-score of each signal measurement value in the sample data; when the Z-score of the measurement value is greater than a preset Z-score threshold, the measurement value is determined to be an outlier measurement value.

[0070] The calculation formula of the Z-score method is as follows:

[0071]

[0072] Wherein, Zscore i,j represents the Z-score corresponding to the measurement value of the sample data i on the signal j, and a i,j represents the measurement value of the sample data i on the signal j, and μ j respectively represent the mean and standard deviation of the signal j. The mean and standard deviation of the signal j are the mean and standard deviation of the measurement value of the signal j of all target data in the standard fault data set.

[0073] In some implementations, the determination of the overall deviation degree of all measurement values in the sample data in step S134 includes:

[0074] The MZ-score method is used to calculate the overall deviation degree of all sample data in the sample data, and the calculation formula of the MZ-score method is as follows:

[0075]

[0076] Wherein, MZscore iZscore represents the overall deviation degree of the measurement values of n signals in sample data i, Zscore i,j Zscore represents the Z-score corresponding to the measurement value of sample data i on signal j, n represents the total number of measurement values of all signals in sample data i, 1≤j≤n.

[0077] In some implementations, the method further comprises:

[0078] Before removing outliers of each target data in the standard fault data set, a protection mask is constructed for data with early warning and fault labels in the standard fault data set, and the protection mask is used to protect the data from being deleted.

[0079] After removing outliers of each target data in the standard fault data set, it is determined whether the measurement values in the data with early warning and fault labels in the standard fault data set are mistakenly removed, and if so, the mistakenly removed measurement values are restored.

[0080] The above method realizes fine protection and error recovery of early warning and fault data by introducing a protection mask mechanism and a mistaken removal recovery strategy in the outlier removal process, so as to ensure the integrity of the data with early warning and fault labels and the reliability of the analysis results.

[0081] In some implementations, the standard fault data set includes measurement values of each signal collected at a corresponding time for a plurality of historical fault data. In step S130, the data in the standard fault data set is preliminarily screened to remove redundant values of each target data in the standard fault data set, including:

[0082] The variance of the measurement values of each signal in the sample data is determined, and the measurement value with a variance greater than a preset variance threshold is removed, the sample data being any target data in the standard fault data set; the measurement values of the same type of signals in two sample data are analyzed for correlation to determine the correlation between the measurement values of the two same type of signals; when the correlation between the measurement values of the two same type of signals is greater than a preset correlation threshold, one of the measurement values is removed.

[0083] The variance is an index for measuring signal volatility, and a signal with excessively large variance may cause unstable data due to noise or outliers, and thus the measurement value thereof needs to be removed. Meanwhile, if the measurement values of two same type of signals are highly correlated, there may be information redundancy, and thus one of the signals needs to be retained to reduce the data dimension. The specific calculation of the variance and the correlation is a routine technique in the art, and thus will not be described herein.

[0084] For example, the variance threshold and the correlation threshold can be set according to actual needs, for example, the variance threshold can be set to 0.01, and the correlation threshold can be set to 0.95, and the embodiments of the present application are not limited in this regard.

[0085] In some implementations, after step S130 is performed, before step S140 is performed, the method can further include:

[0086] adopting a linear interpolation method to fill in the missing measurement data in the preliminary screening data set;

[0087] adopting a label forward filling method to fill in the missing fault label data in the preliminary screening data set.

[0088] The linear interpolation method estimates the missing value through the linear relationship of adjacent non-missing data points. By filling in the missing measurement values, the continuity and integrity of each signal in the time series can be ensured. The label forward filling method fills in the current missing value with the last non-missing fault label value. Forward filling of the missing fault label can ensure the continuity and consistency of the label information, ensure the integrity of the fault label of the sample data, and together improve the data quality, providing reliable support for subsequent key data selection.

[0089] In some implementations, step S140 includes:

[0090] inputting each data in the preliminary screening data set into a preset multi-strategy comprehensive evaluation model, determining the supervision score, stability score, and fault sensitivity score of each data based on a pre-constructed multi-strategy comprehensive evaluation method, performing weighted summation on the supervision score, stability score, and fault sensitivity score of each data in the preliminary screening data set to obtain a comprehensive evaluation index value corresponding to each data, and determining data with a comprehensive evaluation index value greater than a preset comprehensive evaluation threshold as key data.

[0091] The supervision score can measure the direct correlation between the signal and the occurrence of the fault, i.e., the effectiveness of the signal in fault early warning. The stability score can measure the consistency of the signal in different vehicles. The fault sensitivity score can evaluate the sensitivity of the signal to the occurrence of the fault, i.e., whether the signal can show significant changes before the occurrence of the fault. By fusing the supervision score (correlation), stability score (generalization), and fault sensitivity score (dynamic response), the comprehensive evaluation index value constructs a "precision-robustness-efficiency" triangular balance screening framework. This method not only improves the accuracy of single early warning, but also constructs an intelligent early warning ecosystem that adapts to complex scenarios and supports continuous optimization, providing core technical support for the safe operation of new energy vehicles.

[0092] In this embodiment, the pre-constructed multi-strategy comprehensive evaluation method includes a supervision score evaluation method, a stability score evaluation method, and a fault sensitivity score evaluation method.

[0093] The supervised score evaluation method comprises: scoring each data in the preliminary screening data set by using ANOVA variance analysis method, mutual information evaluation method and random forest importance evaluation method respectively, calculating the average value of the scores obtained by the three methods, and determining the average value as the supervised score corresponding to each data in the preliminary screening data set.

[0094] For example, the ANOVA variance analysis method is implemented based on the following calculation formula:

[0095]

[0096] wherein, n i represents the number of the i-th sample in the preliminary screening data set, u i represents the within-group mean, u represents the overall mean, N represents the total number of samples, and k represents the number of sample categories; the score F of each data in the preliminary screening data set using the ANOVA variance analysis method can be calculated by the above calculation formula.

[0097] The mutual information evaluation method is implemented based on the following calculation formula:

[0098] I(X,Y)=H(X)+H(X)-H(X,Y);

[0099] wherein, H(X) and H(Y) represent the entropy of the signal and the label, and H(X,Y) represents the joint entropy; the score I(X,Y) of each data in the preliminary screening data set using the mutual information evaluation method can be calculated by the above calculation formula.

[0100] The random forest importance evaluation method is implemented based on the following calculation formula:

[0101]

[0102] wherein, X i represents the i-th signal, F represents the total number of signals, M represents the number of decision trees in the random forest model, t represents the node of the decision tree, l represents the left node of the decision tree, r represents the right node of the decision tree, θ(·) represents the uncertainty of the sample category in the node, n represents the number of samples contained in the node t, n l represents the number of left node samples contained in the node t, and n r represents the number of right node samples contained in the node t. The score S(X i ) of each data in the preliminary screening data set using the random forest importance evaluation method can be calculated by the above calculation formula.

[0103] For example, the stability score evaluation method can be implemented based on the following formula:

[0104]

[0105] ​wherein x represents the evaluated signal, n represents the total number of vehicles, s n represents the comprehensive score of signal x in the nth vehicle data. The stability score sta(x) of the measurement value of each data in the preliminary screening data set between different vehicles can be calculated by the above formula, and the score is taken as the current cross-vehicle stability evaluation index;

[0106] For example, the fault sensitivity scoring method includes: calculating the trend consistency score, the distribution distinction score, the signal change speed difference score and the early warning position ideal degree score of each data in the preliminary screening data set respectively, and weighting and summing each score to obtain the fault sensitivity score corresponding to each data.

[0107] In this embodiment, the sum of the weights corresponding to the trend consistency score, the distribution distinction score, the signal change speed difference score and the early warning position ideal degree score is 1, for example, the weights corresponding to the trend consistency score, the distribution distinction score, the signal change speed difference score and the early warning position ideal degree score can be 0.25, 0.2, 0.35 and 0.2 respectively.

[0108] The trend consistency score can be determined based on the following formula:

[0109]

[0110] If the signal meets the monotonic trend, it is marked as 1.0, otherwise it is marked as 0.3;

[0111] The distribution distinction score can be determined based on the following formula:

[0112]

[0113] wherein [n 25 ,n 75 ] and [w 25 ,w 75 ] represent the quartile range of the normal state and the early warning state respectively, and max(·) and min(·) represent the maximum and minimum calculation functions respectively;

[0114] The signal change speed difference score can be determined based on the following formula:

[0115]

[0116] wherein η(x) represents the ratio of the average rate of change of the early warning state data to the rate of change of the normal state data, mean(·) warning and mean(·) normal represent the change rate of information in the early warning and normal states respectively.

[0117] The early warning position desirability score can be determined based on the following formula:

[0118]

[0119] wherein λ(x) n , λ(x) w and λ(x) f represent the average values of the data in the normal, early warning and failure states, respectively.

[0120] For example, the weighted sum of the supervision score, the stability score and the failure sensitivity score of each data in the preliminary screening data set in step S140 to obtain the comprehensive evaluation index value corresponding to each data can include:

[0121] The comprehensive evaluation index value is determined based on the following formula:

[0122] Comprehensive evaluation index value = w1 x supervision score + w2 x stability score + w3 x failure sensitivity score

[0123] wherein w1, w2 and w3 are the weights of the supervision score, the stability score and the failure sensitivity score, respectively, and satisfy w1 + w2 + w3 = 1. For example, w1, w2 and w3 can be set to 0.5, 0.3 and 0.2, respectively.

[0124] It should be noted that in this embodiment, after the key data is screened out from the preliminary screening data set after step S140 is performed, the Pearson correlation coefficient matrix method can be used to remove the signals with high correlation (the correlation threshold can be set to, for example, 0.85) in the key data, so as to further reduce the redundancy of the key data. Using the Pearson correlation coefficient matrix method to remove redundant data is a conventional technical means in the art, and will not be described here.

[0125] Based on the same inventive concept, the embodiments of the present application also provide a fault early warning key data screening device for screening fault early warning key data. Figure 3 is a structural block diagram of a fault early warning key data screening device provided by the embodiments of the present application, as shown in Figure 3 The device 300 comprises:

[0126] A data integration module 310 is configured to obtain historical fault data of a plurality of vehicles, and integrate the historical fault data of the plurality of vehicles to establish a standard fault data set, wherein the standard fault data set comprises original binary classification fault labels corresponding to each piece of historical fault data.

[0127] The label reconstruction module 320 is configured to reconstruct the labels of the data in the standard fault data set, and expand the original binary fault labels of the data into ternary fault labels. The original binary fault labels include two types of fault labels, i.e., normal and fault. The ternary fault labels include three types of fault labels, i.e., normal, pre-warning and fault.

[0128] The data preliminary screening module 330 is configured to preliminarily screen the data in the standard fault data set, and remove the abnormal values and redundant values of each target data in the standard fault data set to obtain a preliminary screening data set. The target data is the data with a normal fault label in the standard fault data set.

[0129] The data fine screening module 340 is configured to evaluate the data in the preliminary screening data set based on a pre-constructed multi-strategy comprehensive evaluation method, and screen out key data from the preliminary screening data set according to the evaluation result.

[0130] In some implementations, the standard fault data set includes time stamps and vehicle identification codes corresponding to each piece of historical fault data, and the data integration module 310 is configured to:

[0131] perform data conversion on the historical fault data of the plurality of vehicles, so that the data types of the historical fault data of the plurality of vehicles are unified;

[0132] perform time synchronization processing on the historical fault data of the plurality of vehicles based on the vehicle identification codes and the time stamps, and sort and integrate the historical fault data of different vehicles in time sequence into the standard fault data set.

[0133] In some implementations, the label reconstruction module 320 is configured to:

[0134] reconstruct the original binary labels of the historical fault data with a time stamp within T time before the fault occurrence into pre-warning labels, and keep the original normal or fault labels unchanged for the rest of the data.

[0135] In some implementations, the standard fault data set includes measurement values of each signal collected at a corresponding time by a plurality of pieces of historical fault data. The data preliminary screening module 330 is configured to:

[0136] determine whether there is an outlier measurement value in sample data, the sample data being any target data in the standard fault data set; when there is an outlier measurement value in the sample data, determine a proportion of the number of the outlier measurement values in the sample data; when the proportion of the number of the outlier measurement values in the sample data is less than or equal to a preset outlier proportion threshold, remove the outlier measurement values in the sample data; when the proportion of the number of the outlier measurement values in the sample data is greater than the preset outlier proportion threshold, determine an overall deviation degree of all measurement values in the sample data; and when the overall deviation degree is greater than a preset deviation proportion threshold, remove the sample data.

[0137] In some implementations, the data preliminary screening module 330 is further configured to:

[0138] determine the variance of the measurement values of each signal in the sample data, and remove the measurement values with a variance greater than a preset variance threshold, the sample data being any target data in the standard fault data set;

[0139] perform correlation analysis on the measurement values of the same type of signals in the two sample data, and determine the correlation between the measurement values of the two same type of signals;

[0140] when the correlation between the measurement values of the two same type of signals is greater than a preset correlation threshold, remove one of the measurement values.

[0141] In some implementations, the device further comprises a data filling module 350, which is configured to, before evaluating the data in the preliminary screening data set based on the pre-constructed multi-strategy comprehensive evaluation method:

[0142] fill the missing measurement values in the preliminary screening data set by using linear interpolation method; and fill the missing fault labels in the preliminary screening data set by using label forward filling method.

[0143] In some implementations, the data fine screening module 340 is configured to:

[0144] input each data in the preliminary screening data set into a preset multi-strategy comprehensive evaluation model, the multi-strategy comprehensive evaluation model being based on a pre-constructed multi-strategy comprehensive evaluation method, to determine the supervision score, the stability score and the fault sensitivity score of each data;

[0145] perform weighted summation on the supervision score, the stability score and the fault sensitivity score of each data in the preliminary screening data set to obtain the comprehensive evaluation index value corresponding to each data;

[0146] determine the data with a comprehensive evaluation index value greater than a preset comprehensive evaluation threshold as key data.

[0147] The specific details of the fault early warning key data screening method used in the above fault early warning key data screening device can be understood as the corresponding related descriptions and effects in the above-described fault early warning key data screening method embodiments, which will not be repeated here.

[0148] Based on the same inventive concept as the fault early warning key data screening method described above, the present application also provides an electronic device, which can include a processor and a memory, wherein the processor and the memory can be connected to each other in communication through a bus or other means. The processor can be a central processing unit (CPU). The processor can also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. or a combination of the above-mentioned chips. The memory, as a non-transitory computer readable storage medium, can be used to store non-transitory software programs, non-transitory computer executable programs and modules, such as program instructions / modules corresponding to the fault early warning key data screening method in the embodiments of the present application. The processor executes various functional applications and data processing of the processor by running the non-transitory software programs, instructions and modules stored in the memory, that is, implements the fault early warning key data screening method in the above-mentioned method embodiments.

[0149] The memory can include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application program required by a function; the data storage area can store data created by the processor, etc. In addition, the memory can include a high-speed random access memory, and can also include a non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state memory device. The one or more modules are stored in the memory, and when executed by the processor, perform the fault early warning key data screening method in the embodiments shown. Figure 1 The fault early warning key data screening method in the embodiments shown.

[0150] The above-mentioned specific details of the electronic device can be understood by referring to the corresponding related descriptions and effects in the embodiments shown, which will not be described here. Figure 1 The above-mentioned specific details of the electronic device can be understood by referring to the corresponding related descriptions and effects in the embodiments shown, which will not be described here.

[0151] Based on the same inventive concept as the fault early warning key data screening method, the present application also provides a computer readable storage medium, wherein the computer readable storage medium stores computer instructions, and the computer program instructions are used to make a computer execute the fault early warning key data screening method in the above-mentioned embodiments.

[0152] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a computer readable storage medium. When the program is executed, it can include the processes of the above-mentioned embodiment methods. The storage medium can be a read-only memory (ROM), a random access memory (RAM), a flash memory, a hard disk drive (HDD) or a solid-state drive (SSD), etc. The storage medium can also include a combination of the above-mentioned types of memories.

[0153] The technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages:

[0154] The fault early warning key data screening method, device, equipment and medium provided by the embodiment of the present application can more accurately depict the vehicle state evolution process and provide more rich fault development information for subsequent analysis by establishing a standard fault data set through data integration and expanding the original binary classification fault label of the data into a three-class label containing normal, early warning and fault. The abnormal values and redundant values of each target data with a normal fault label in the standard fault data set are eliminated through preliminary screening of the data in the standard fault data set, which avoids the loss of key fault features that may be caused by directly processing early warning and fault label data, effectively reduces the noise interference in the normal data, and significantly improves the data quality; the integrity of the early warning label data is preserved, the early features of potential fault patterns are completely preserved, the model can learn the signal change rule based on the complete fault development trajectory (normal → early warning → fault), and the identification ability of the model for early faults is enhanced. Finally, the preliminary screening data set is evaluated and key data are screened based on a multi-strategy comprehensive evaluation method, which overcomes the limitations of the existing single screening standard, can comprehensively consider the evaluation of various factors on the importance of signals, and thus more efficiently and accurately screens out key signals that have significant changes and are highly related before fault occurrence, meets the demand of the new energy vehicle fault early warning system for the stability and controllability of key signals, solves the problem of insufficient model performance in the cross-vehicle data scenario, and is beneficial to improving the safety and operation stability of new energy vehicles.

[0155] In the specification provided herein, a large number of specific details are described. However, it can be understood that the embodiments of the present application can be practiced without these specific details. In some examples, well-known methods, structures and techniques are not shown in detail in order not to obscure the understanding of the present specification.

[0156] Similarly, it should be understood that in order to streamline the present disclosure and aid in understanding one or more of the various inventive aspects, in the above description of exemplary embodiments of the invention, various features of the invention are sometimes grouped together into a single embodiment, figure, or description thereof. However, this disclosed method should not be interpreted as reflecting an intention that the claimed invention requires more features than are expressly recited in each claim. Rather, as reflected in the claims below, inventive aspects lie in less than all the features of the individual embodiments disclosed above. Accordingly, the claims following the detailed description are hereby expressly incorporated into this detailed description, with each claim standing on its own as a separate embodiment of the invention.

[0157] It should be noted that the above-mentioned embodiments illustrate rather than limit the invention and that those skilled in the art will be able to design alternative embodiments without departing from the scope of the appended claims.

Claims

1. A method for screening key data for fault warning, characterized in that: include: Acquire historical fault data of multiple vehicles, and integrate the historical fault data of the multiple vehicles to establish a standard fault data set, wherein the standard fault data set includes original binary classification fault labels corresponding to each piece of historical fault data; Reconstructing labels for the data in the standard fault dataset, expanding the original two-category fault labels of the data into three-category fault labels, wherein the original two-category fault labels include two types of fault labels, normal and fault, and the three-category fault labels include three types of fault labels, normal, warning, and fault; Performing a preliminary screening on the data in the standard fault data set, removing abnormal values ​​and redundant values ​​of each target data in the standard fault data set, and obtaining a preliminary screening data set, wherein the target data is the data in the standard fault data set with a normal fault label; Based on a pre-constructed multi-strategy comprehensive evaluation method, the data in the primary screening data set is evaluated, and key data is screened out from the primary screening data set according to the evaluation results.

2. The method according to claim 1, characterized in that The standard fault data set includes a timestamp and a vehicle identification code corresponding to each piece of historical fault data. The historical fault data of the plurality of vehicles are integrated to establish a standard fault data set, including: Performing data conversion on historical fault data of a plurality of vehicles so that the data types of the historical fault data of the plurality of vehicles are unified; The historical fault data of the plurality of vehicles are subjected to time synchronization processing based on the vehicle identification codes and the timestamps, and the historical fault data of different vehicles are uniformly sorted in time series and integrated into a standard fault data set.

3. The method according to claim 1, characterized in that Label reconstruction is performed on the data in the standard fault dataset to expand the original two-category fault labels of the data into three-category fault labels, including: For the historical fault data whose timestamps are within T time before the fault occurs, their original binary classification labels are reconstructed into warning labels, and the remaining data retains the original normal or fault labels unchanged.

4. The method according to claim 1, wherein The standard fault data set includes measurement values ​​of various signals collected at corresponding moments of a plurality of historical fault data; the preliminary screening of the data in the standard fault data set to remove abnormal values ​​of each target data in the standard fault data set includes: Determining whether there is an outlier measurement value in sample data, wherein the sample data is any one of the target data in the standard fault data set; When there are outlier measurement values ​​in the sample data, determining a proportion of the number of outlier measurement values ​​in the sample data; When the proportion of outlier measurement values ​​in the sample data is less than or equal to a preset outlier proportion threshold, the outlier measurement values ​​in the sample data are eliminated; When the proportion of the number of outlier measurement values ​​in the sample data is greater than a preset outlier proportion threshold, determining the overall deviation degree of all measurement values ​​in the sample data; When the overall deviation degree is greater than a preset deviation ratio threshold, the sample data is eliminated.

5. The method according to claim 1, wherein The standard fault data set includes measurement values ​​of various signals collected at corresponding moments of a plurality of historical fault data; the preliminary screening of the data in the standard fault data set to remove redundant values ​​of each target data in the standard fault data set includes: Determining the variance of the measurement values ​​of each signal in the sample data, and eliminating the measurement values ​​whose corresponding variances are greater than a preset variance threshold, wherein the sample data is any target data in the standard fault data set; Performing a correlation analysis on the measurement values ​​of the same type of signals in the two sample data to determine the correlation between the measurement values ​​of the two same type of signals; When the correlation between the measurement values ​​of two signals of the same type is greater than a preset correlation threshold, one of the measurement values ​​is removed.

6. The method according to claim 1, characterized in that Before evaluating the data in the primary screening dataset based on the pre-built multi-strategy comprehensive evaluation method, the method further includes: The missing measurement values ​​in the primary screening data set are numerically filled using linear interpolation; The label forward filling method is used to fill the fault labels of the data with missing fault labels in the initial screening data set.

7. The method according to claim 1, characterized in that Based on a pre-built multi-strategy comprehensive evaluation method, the data in the preliminary screening dataset is evaluated, and key data is screened from the preliminary screening dataset based on the evaluation results, including: Inputting each data in the primary screening data set into a preset multi-strategy comprehensive evaluation model, wherein the multi-strategy comprehensive evaluation model determines the supervision score, stability score, and fault sensitivity score of each data based on a pre-built multi-strategy comprehensive evaluation method; Performing a weighted summation of the supervision score, stability score, and fault sensitivity score of each data in the preliminary screening data set to obtain a comprehensive evaluation index value corresponding to each data; Determine that the data whose comprehensive evaluation index value is greater than a preset comprehensive evaluation threshold value is the key data.

8. A device for screening key data of fault warning, characterized in that: include: A data integration module is used to establish and obtain historical fault data of multiple vehicles, and integrate the historical fault data of the multiple vehicles to establish a standard fault data set, wherein the standard fault data set includes the original binary classification fault labels corresponding to each piece of historical fault data; a label reconstruction module, configured to reconstruct labels for the data in the standard fault dataset, and expand the original two-category fault labels of the data into three-category fault labels, wherein the original two-category fault labels include two types of fault labels, normal and fault, and the three-category fault labels include three types of fault labels, normal, warning, and fault; a data preliminary screening module, configured to perform preliminary screening on the data in the standard fault data set, remove abnormal values ​​and redundant values ​​of each target data in the standard fault data set, and obtain a preliminary screening data set, wherein the target data is the data in the standard fault data set with a normal fault label; The data fine screening module is used to evaluate the data in the primary screening data set based on a pre-built multi-strategy comprehensive evaluation method, and to screen out key data from the primary screening data set according to the evaluation results.

9. An electronic device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the fault warning key data screening method according to any one of claims 1 to 7 by executing the computer instructions.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute the fault warning key data screening method according to any one of claims 1 to 7.

Images