A data-driven adaptive fault diagnosis method based on multiple evaluation indicators

Abstract

The present invention belongs to the field of fault diagnosis technology, and discloses a data-driven adaptive fault diagnosis method based on multiple evaluation indicators, comprising the following steps: step S1, inputting system data defined in a standard format; step S2, establishing a data processing algorithm model and a fault diagnosis algorithm model; step S3, comprehensive evaluation indicator APR; step S4, based on the adaptive fault diagnosis method, obtaining the optimal algorithm combination and the final fault diagnosis result, and reverse verification. The present invention adopts the above-mentioned data-driven adaptive fault diagnosis method based on multiple evaluation indicators, based on the comprehensive performance evaluation indicator APR, comprehensively weighing the needs of different indicators, reversely evaluating the adaptability of different system data to the current model library, eliminating unnecessary time-consuming manual research, and supporting the directionally expanded model library.

Description

Technical Field

[0001] The present invention relates to the technical field of fault diagnosis, and in particular to a data-driven adaptive fault diagnosis method based on multi-evaluation index traction. Background Art

[0002] Currently, machine learning is the mainstream data-driven fault diagnosis method. The diagnostic process generally includes feature extraction, data dimensionality reduction, noise reduction, balancing, normalization, and the selection and optimization of different algorithm models. Matching fault diagnosis methods to different systems and data types typically requires research and comparative analysis of different algorithm models. Based on existing popular algorithms, specialized fault diagnosis models are developed through parameter optimization and data processing to achieve models that meet performance evaluation requirements. This results in relatively long development times and limited versatility.

[0003] Many existing studies, such as those using a single algorithm to optimize training models, are prone to overfitting and are only suitable for relatively optimal fault diagnosis in simple systems and for specialized data. Ensemble learning methods, integrating multiple homogeneous or heterogeneous base classifiers, can reduce performance if there are too many weak classifiers, requiring considerable research and time to achieve the optimal base classifier combination, making it relatively complex. Adaptive learning methods such as genetic algorithms and particle swarm optimization fail to consider alternative solutions from other perspectives, such as data processing. The integration of data processing and fault diagnosis algorithms, as well as the fusion of different algorithms, can lead to even worse performance if mismatched, requiring significant time and effort. Therefore, using a single accuracy rate as an evaluation metric is difficult to meet demand.

[0004] To sum up, the existing fault diagnosis methods lack systematic resource integration and have "technical island barriers". Different technicians who want to obtain the optimal fault diagnosis method for different system data basically have to spend a long time to conduct research on data processing, fault diagnosis methods and their relationship. Moreover, the evaluation indicators are relatively single and insufficient for comprehensive evaluation. There is a lack of an adaptive fault diagnosis method based on the integration of existing system resources. Summary of the Invention

[0005] The purpose of the present invention is to provide a data-driven adaptive fault diagnosis method based on multiple evaluation indicators. Based on the comprehensive performance evaluation indicator APR of fault diagnosis accuracy A, fault detection precision rate P, and fault detection completeness rate R, the requirements of different indicators are comprehensively weighed. For different system data, by adaptively evaluating whether data processing such as dimensionality reduction / noise reduction is required and matching different fault classification diagnosis models, the optimal algorithm model or combination and fusion model is output to obtain the optimal fault diagnosis result based on the APR indicator threshold evaluation. At the same time, based on the obtained optimal APR indicator, the adaptability of different system data to the current model library is reversely evaluated, eliminating unnecessary time-consuming manual research, and supporting the directional expansion of the model library.

[0006] To achieve the above objectives, the present invention provides a data-driven adaptive fault diagnosis method based on multiple evaluation indicators, comprising the following steps:

[0007] Step S1, inputting system data defined in a standard format;

[0008] Step S2: establishing a data processing algorithm model and a fault diagnosis algorithm model;

[0009] Step S3, comprehensive evaluation index APR;

[0010] Step S4: Based on the adaptive fault diagnosis method, obtain the optimal algorithm combination and the final fault diagnosis result, and perform reverse verification.

[0011] Preferably, in step S1, the input system data is specified in a unified standard format, and each row is specified to store a type of normal / failure mode feature data, the first column is specified as a classification label, and the other columns are different feature data;

[0012] Among them, the data format supports .xlsx / .xls / .txt / .csv, and can be expanded to recognize different data types as needed.

[0013] Preferably, in step S1, the input system data is discrete feature data stored according to the format requirements, and the next step of adaptive fault diagnosis is performed directly; for time-continuous data, typical time domain features and frequency domain features are adaptively extracted, and the time domain features and frequency domain feature data are automatically saved locally according to the discrete data format requirements, and then the next step of adaptive fault diagnosis is performed.

[0014] Preferably, in step S2, for different feature data, a library of classic or improved data processing and fault classification algorithms is integrated to support adaptive random fusion calling and matching of different data processing and fault classification algorithm combinations to evaluate performance indicators.

[0015] Preferably, in step S3, the comprehensive evaluation index APR includes accuracy A, precision P and recall R;

[0016] (1) Accuracy A represents the ratio of the number of predicted classification labels that are consistent with the true classification labels to the total number of labels;

[0017] (2) Precision P: Indicates how many of the test samples predicted to be other classes are actually other class samples, as shown below:

[0018] (1);

[0019] in, TN X Indicates the number of predicted classes that are actually other classes. FN X Indicates that the prediction is other classes and the actual X the number of classes;

[0020] (3) Recall R: Indicates how many other types of faults in the test sample are correctly predicted, as shown below:

[0021] (2);

[0022] in, FP X The prediction is X The class is actually the number of other classes;

[0023] Since precision and recall are often a set of contradictory indicators, F1 is used as the evaluation indicator; F1 X It is defined as the weighted harmonic mean of precision and recall, assuming that the weight of each indicator is 1, as shown below:

[0024] (3);

[0025] Among them, F1 X There are also contradictions in the indicators, so the overall F1 indicator value is defined as the F1 of different categories of data X The weighted harmonic mean of the indicator values ​​is as follows:

[0026] (4);

[0027] Therefore, the APR metric is defined as the weighted harmonic mean of F1 and accuracy as follows:

[0028] (5).

[0029] Preferably, in step S4, based on the adaptive fault diagnosis method, the final fault diagnosis result is obtained and reversely verified. The specific steps are as follows:

[0030] Step S41: Select a genetic algorithm to implement adaptive fault diagnosis, and determine the highest algorithm combination number dim based on existing data processing and model algorithm library;

[0031] Step S42: Generate a dim-bit binary chromosome by a genetic algorithm, perform random encoding, generate a given population number num, and give a termination genetic generation number gen;

[0032] Step S43: Decode the population according to the pre-defined meaning of each bit of chromosome code corresponding to different algorithm enabled functions;

[0033] Step S44: Calculate the evaluation index APR for the algorithm combination corresponding to the chromosomes in different populations, the original algorithm combination, and each single classification algorithm therein. Compare the APR index obtained by the original algorithm combination with the algorithm combination after eliminating the weak single algorithm with substandard APR index. Finally, optimize and retain the algorithm combination with the highest APR index.

[0034] Step S45: Based on the diagnosis results of the combination of multiple algorithms and the idea of ​​ensemble learning, a voting decision-making method is adopted;

[0035] Step S46: The parent population is screened according to a predetermined evaluation index APR threshold to retain the excellent population; if there is no chromosome in the population that meets the APR threshold, the chromosome that obtains the optimal APR index threshold in the population is retained as the excellent parent chromosome;

[0036] Step S47: Generate offspring populations based on crossover and mutation of parent populations, fuse excellent parent populations with offspring populations, and then generate a new generation population based on the evaluation index APR competition and selection, and iterate repeatedly until the given evaluation index APR threshold is reached, and then terminate the iteration or the given genetic generation condition is met;

[0037] Step S48: Finally, the output algorithm combination is decoded from the chromosome of the output optimal APR index to obtain the final fault diagnosis result, and the adaptability of the current system input data is judged based on whether the optimal APR index meets the given APR threshold.

[0038] Therefore, the present invention adopts the above-mentioned data-driven adaptive fault diagnosis method based on multiple evaluation indicators, and the beneficial effects are as follows:

[0039] (1) The comprehensive evaluation performance index APR method proposed in this invention is suitable for comprehensive consideration of multi-objective optimization problems. Other different evaluation indicators can be constructed as optimization objectives in the same way.

[0040] (2) The present invention can reversely evaluate whether the input system data is suitable for the current data processing and algorithm library by obtaining the comprehensive evaluation performance index APR to see whether it meets the threshold requirements. It can directionally expand the existing different types of feature extraction, data processing and machine learning algorithm databases according to demand, which is conducive to the integration of technical resources and avoidance of "technical island barriers".

[0041] (3) The present invention aims to construct a universal system data-driven adaptive fault diagnosis method to reduce the time-consuming cost of repeated cross-research, comparison and development of dedicated models by different technicians on different system data. It can lay a diagnostic foundation for auxiliary decision optimization and health management evaluation of different systems, and help to construct an integrated operation and maintenance platform for different systems.

[0042] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 This is a schematic diagram of a data-driven adaptive fault diagnosis method based on multiple evaluation indicators in the present invention;

[0044] Figure 2 This is a workflow diagram of a data-driven adaptive fault diagnosis method based on multiple evaluation indicators. DETAILED DESCRIPTION

[0045] The technical solution of the present invention is further described below with reference to the accompanying drawings and embodiments.

[0046] like Figure 1 As shown, the present invention presents a data-driven adaptive fault diagnosis system based on multiple evaluation indicators. This system uses the comprehensive performance evaluation indicator APR (fault diagnosis accuracy rate A), fault detection precision rate P, and fault detection recovery rate R) as the optimization target. It employs an adaptive learning method to encode the combined data processing algorithm model and the fault diagnosis algorithm model. Simply inputting discrete feature data or time-continuous data defined in a standard format, the encoding and decoding system can adaptively evaluate and filter a single or multiple combinations of algorithm models to obtain the optimal algorithm model combination that meets the comprehensive evaluation indicator threshold, as well as the diagnostic results for multi-point, multi-classification fault data. Furthermore, the adaptability of the input system data can be reversely evaluated based on the APR indicator threshold.

[0047] In this invention, we will not elaborate too much on the connotation of feature extraction, data processing, and machine learning algorithms. The present invention is a data-driven adaptive fault diagnosis method based on multiple evaluation indicators, focusing on the integration and utilization of various algorithm model resources that have been optimized and improved, and adaptive evaluation selection. Its workflow is as follows: Figure 2 shown.

[0048] Step S1: Input system data defined in a standard format.

[0049] To meet the universality of automatic identification of system data, a unified standard format is specified for input system data. Each row is required to store a type of normal / fault mode characteristic data. The first column is specified as the classification label, and the other columns are different characteristic data. The data format supports .xlsx / .xls / .txt / .csv, and can be expanded to identify different data types as needed.

[0050] The input system data needs to select discrete feature data or time-continuous data type; for discrete feature data stored according to the format requirements, the next step of adaptive fault diagnosis can be carried out directly; for time-continuous data, typical time domain features or frequency domain features will be adaptively extracted, and the time domain features and frequency domain feature data will be automatically saved locally according to the discrete data format requirements, and then the next step of adaptive fault diagnosis will be carried out.

[0051] Step S2: Establish a data processing algorithm model and a fault diagnosis algorithm model.

[0052] For different feature data, it integrates classic or improved data processing and fault classification algorithm libraries, supports adaptive random fusion calls to match different data processing and fault classification algorithm combinations, and evaluates performance indicators to obtain the optimal algorithm combination.

[0053] Through comparative analysis, it is found that the performance indicators can be improved by integrating and calling certain fault classification algorithms after dimensionality reduction processing. This is because the principal component analysis method forms independent principal components after transforming the characteristic variables of the original data indicators. If the correlation between the features is strong, it will have a greater impact on the diagnosis and prediction of certain algorithms. For example, research has found that the performance indicators will be significantly improved by calling the KNN algorithm after the features have been reduced in dimension.

[0054] Step S3: Comprehensive evaluation index APR.

[0055] The evaluation of fault diagnosis results generally relies on accuracy, precision, and recall. However, fault diagnosis often focuses on "fault-type" samples. For a certain category X, the classification test results of different algorithm models have the following concentrated categories:

[0056] TP X Indicates the number of predicted X categories that are actually X categories; FP X Indicates the number of predicted X categories that are actually other categories, TN X Indicates the number of predicted classes that are actually other classes, FN X Indicates the number of cases predicted to be other classes but actually belonging to class X.

[0057] Assuming that the total number of fault categories (including normal categories) is n, we can get:

[0058] Accuracy A: It indicates the ratio of the number of predicted classification labels that are consistent with the true classification labels to the total number of labels.

[0059] The precision rate P of a certain type of fault: indicates how many of the test samples predicted to be other classes are actually other class samples, as shown below:

[0060] (1);

[0061] The recovery rate R of a certain type of fault: indicates how many other types of faults in the test sample are correctly predicted, as shown below:

[0062] (2);

[0063] Since precision and recall are often a set of contradictory indicators, when it is uncertain which indicator is more important, or when both indicators are considered equally important, F1 is generally used as the evaluation indicator. X It is defined as the weighted harmonic mean of precision and recall, taking the weight of each indicator as 1 as an example, as shown below:

[0064] (3);

[0065] When evaluating and comparing the fitness of different algorithm combinations, the comprehensive index APR is used as the optimization target. APR is a comprehensive evaluation index that considers the multi-objective optimization of fault classification accuracy A, fault detection precision P, and fault detection rate R. X Indicators are often contradictory, so the overall F1 index value is defined as the F1 of different categories of data. X The weighted harmonic mean of the indicator values ​​is as follows:

[0066] (4);

[0067] Therefore, the APR metric is defined as the weighted harmonic mean of F1 and accuracy as follows:

[0068] (5).

[0069] Step S4: Based on the adaptive fault diagnosis method, obtain the final fault diagnosis result and perform reverse verification.

[0070] Combining the different characteristics and requirements of data processing and fault classification algorithm model combination or fusion, the adaptive algorithm model combination optimization and the algorithm model fusion properly selected by machine learning further improve the performance indicators.

[0071] Adaptive combination optimization based on evaluation indicators, with the integration of existing resources, eliminates the need for repeated and time-consuming cross-research by different technical personnel. Instead, it can adaptively evaluate and select the model combination or the optimal combination that meets the indicator threshold. The specific implementation process is as follows:

[0072] Step S41: Select a genetic algorithm to implement adaptive fault diagnosis in the test verification, and determine the highest algorithm combination number dim based on existing data processing and model algorithm library;

[0073] Step S42: Generate a dim binary chromosome by a genetic algorithm, perform random encoding, generate a given population number num, and give a termination genetic generation number gen;

[0074] Step S43: Decode the population according to the pre-defined meaning of each bit of chromosome code corresponding to different algorithm enabled functions;

[0075] Step S44: Calculate the evaluation index APR for the algorithm combination corresponding to the chromosomes in different populations, the original algorithm combination, and each single classification algorithm therein. Compare the APR index obtained by the original algorithm combination with the algorithm combination after eliminating the weak single algorithm with substandard APR index. Finally, optimize and retain the algorithm combination with the highest APR index.

[0076] Step S45: Based on the diagnosis results of the combination of multiple algorithms and the idea of ​​ensemble learning, a voting decision-making method is adopted;

[0077] Step S46: The parent population is screened according to a predetermined evaluation index APR threshold to retain the excellent population; if there is no chromosome in the population that meets the APR threshold, the chromosome that obtains the optimal APR index threshold in the population is retained as the excellent parent chromosome;

[0078] Step S47: Generate offspring populations based on crossover and mutation of parent populations, fuse excellent parent populations with offspring populations, and then generate a new generation population based on the evaluation index APR competition and selection, and iterate repeatedly until the given evaluation index APR threshold is reached, and then terminate the iteration or the given genetic generation condition is met;

[0079] Step S48: Finally, the output algorithm combination is decoded from the chromosome of the output optimal APR index to obtain the final fault diagnosis result, and the adaptability of the current system input data is judged based on whether the optimal APR index meets the given APR threshold.

[0080] Example

[0081] The test verification of the data-driven adaptive fault diagnosis method based on multiple evaluation indicators in the present invention is as follows:

[0082] a) Discrete feature data and time-continuous data.

[0083] The test and verification of this embodiment is divided into inputting discrete feature data and time-continuous data, starting with inputting standard format system data.

[0084] Discrete feature data is extracted from a certain CAN network and is classified into four categories: normal (category 1), single-point fault (category 2), excessive impedance (category 3), and insufficient impedance (category 4). Training samples total 1502 groups: 700 for category 1, 160 for category 2, 284 for category 3, and 358 for category 4. Testing samples total 926 groups: 88 for category 1, 222 for category 2, 156 for category 3, and 460 for category 4. Each sample contains 15-dimensional network feature values, including maximum amplitude, minimum amplitude, and overshoot. The data is stored in an .xlsx file format. The first column contains the data classification label, the first row contains the feature name corresponding to each column, and the remaining columns contain the corresponding 15-dimensional features.

[0085] This embodiment uses the rolling bearing standard fault dataset (CWRU) published by Case Western Reserve University in the United States as an example for time-continuous data testing. The extracted data includes normal data and single-point fault data under four different loads (speeds). The single-point fault data, exemplified by single-point drive-end fault data, includes faults of the bearing roller, inner race, and outer race. Roller faults, inner race faults, and outer race faults are classified as minor (fault diameter 0.007 mils), moderate (fault diameter 0.014 mils), and severe (fault diameter 0.021 mils), respectively. This means that each load data type contains data for 10 fault categories (including normal data).

[0086] In order to keep the extracted features uniform, two sensor measurement points are retained for each fault category data, including the time-continuous data of the DE drive end acceleration data and the FE fan end acceleration data; the data sampling frequency is 12kHz.

[0087] For the time-continuous data tested in this embodiment, the data format is stored in .xlsx format when input into the system test evaluation. The first row contains the classification label for each column of time-continuous data. The DE drive-end acceleration data and FE fan-end acceleration data collected from the same category are placed in two adjacent columns.

[0088] Taking the training data of this embodiment as an example, each category of data includes 2 columns of adjacent time-continuous data, and the 10 types of fault category data (including normal data) have a total of 20 columns of time-continuous data.

[0089] b) Feature extraction of time-continuous data.

[0090] For time-continuous data input into the system according to standard format requirements, by entering the number of measurement points contained in each type of data (each column of time-continuous data is one measurement point data), the sampling frequency (12kHz in this case), and the sampling period (5s in this case, that is, feature extraction is performed every 60,000 points), the system can automatically extract 10-dimensional typical time-domain features such as maximum value, minimum value, peak-to-peak value, mean value, and kurtosis, as well as 12-dimensional typical frequency-domain features such as mean frequency and frequency center for each column of time-continuous data, for a total of 22 dimensions of features. It can also automatically fuse the data features of multiple measurement points that represent the same type of data.

[0091] Each type of data in the test data of this embodiment contains acceleration data from two measuring points, and a 44-dimensional fusion feature of each type of data can be obtained. This embodiment extracts 70 groups of training feature data containing 10 types of data (10 groups of each type of data) under a certain load and 70 groups of test feature data containing 10 types of data (10 groups of each type of data) under the other three loads. The extracted feature data is saved in .xlsx format according to the standard format of discrete feature data.

[0092] c) Establish data processing and fault diagnosis algorithm models.

[0093] During testing and verification of this embodiment, the data processing models selected included the commonly used kernel principal component analysis (KPCA) and normalized data processing methods. The machine learning fault diagnosis algorithm models selected included seven machine learning methods: support vector machine (SVM), K-nearest neighbor (KNN), BP neural network algorithm, random forest (RF), naive Bayes (NB), improved logistic regression (LR), and improved K-means clustering.

[0094] d) Test and verify the results.

[0095] Through testing in this embodiment, for discrete system data input, the adaptive genetic algorithm set the maximum possible number of algorithm combinations corresponding to the number of chromosome individual bits to 9, the number of generated populations to 18, the number of genetic generations for iteration termination to 3, the APR input threshold to 0.9, and the APR iteration termination index to 0.99. Multiple algorithm combinations with an APR index of 0.9 were obtained. In the first generation of populations, the algorithm model combination with the optimal APR index of 0.99875 (A=0.99676, P=0.99839, R=0.9991, F1=0.99875) and the final fault diagnosis results were obtained. The adaptive evaluation output of the optimal algorithm model combination was "001111101." Based on the algorithm combination decoding, no data processing was required, and the SVM-based classifier, KNN-based classifier, BP neural network-based classifier, RF random forest-based classifier, LR logistic regression-based classifier, and K-means-based classifier were selected for voting decision fault classification. Finally, based on the fact that the optimal APR index of 0.99875 meets the 0.9 threshold requirement, the adaptability of the input discrete system data to the current system solution is reversely verified.

[0096] For the time-continuous input data in this example, the adaptive genetic algorithm set the maximum possible number of combinations corresponding to the number of chromosome individual bits to 9, the number of generated populations to 18, the number of genetic generations for iteration termination to 3, the APR input threshold to 0.9, and the APR iteration termination index to 0.99. Still, within the first generation, an algorithm model combination with an optimal APR index of 1 (A=1, P=1, R=1, F1=1) and the final fault diagnosis results were obtained. The adaptive evaluation output of the optimal algorithm model combination was "100000100." Based on the algorithm combination decoding, the KPCA algorithm was used to reduce the data dimensionality, and the LR logistic regression base classifier was selected for voting fault diagnosis. Finally, the optimal APR index of 1, which met the 0.9 threshold requirement, reversely verified the adaptability of this time-continuous input data to the current system solution.

[0097] Therefore, the present invention adopts the above-mentioned data-driven adaptive fault diagnosis method based on multiple evaluation indicators, and comprehensively measures the needs of different indicators based on the comprehensive performance evaluation indicator APR of fault diagnosis accuracy A, fault detection precision rate P, and fault detection rate R; for different system data, through adaptive evaluation of whether data processing such as dimensionality reduction / noise reduction is needed and matching different fault classification diagnosis models, the optimal algorithm model or combination and fusion model is output to obtain the optimal fault diagnosis result based on the APR indicator threshold evaluation; at the same time, based on the obtained optimal APR indicator, the adaptability of different system data to the current model library is reversely evaluated, eliminating unnecessary time-consuming manual research, and supporting the directional expansion of the model library.

[0098] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit the same. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solutions of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A data-driven adaptive fault diagnosis method based on multiple evaluation indicators, characterized in that: The following steps are involved: Step S1, inputting system data defined in a standard format; Step S2: establishing a data processing algorithm model and a fault diagnosis algorithm model; Step S3, the comprehensive evaluation index APR includes accuracy A, precision P and recall R; (1) Accuracy A represents the ratio of the number of predicted classification labels that are consistent with the true classification labels to the total number of labels; (2) Precision P: It indicates how many of the test samples predicted as other classes are actually other class samples, as shown below: Among them, TN X Indicates the number of predicted classes that are actually other classes, FN X Indicates the number of X categories that are actually predicted to be other categories; (3) Recall R: It indicates how many other types of faults in the test sample are correctly predicted, as shown below: Among them, FP X Indicates the number of categories predicted to be X but actually other categories; Since precision and recall are often a set of contradictory indicators, F1 is used as the evaluation indicator; F1 X It is defined as the weighted harmonic mean of precision and recall, assuming that the weight of each indicator is 1, as shown below: Among them, F1 X There are also contradictions in the indicators, so the overall F1 indicator value is defined as the F1 of different categories of data X The weighted harmonic mean of the indicator values ​​is as follows: Therefore, the APR metric is defined as the weighted harmonic mean of F1 and accuracy as follows: Step S4: Based on the adaptive fault diagnosis method, the optimal algorithm combination and the final fault diagnosis result are obtained and reversely verified. The specific steps are as follows: Step S41: Select a genetic algorithm to implement adaptive fault diagnosis, and determine the highest algorithm combination number dim based on existing data processing and model algorithm library; Step S42: Generate a dim-bit binary chromosome by a genetic algorithm, perform random encoding, generate a given population number num, and give a termination genetic generation number gen; Step S43: Decode the population according to the pre-defined meaning of each bit of chromosome code corresponding to different algorithm enabled functions; Step S44: Calculate the evaluation index APR for the algorithm combination corresponding to the chromosomes in different populations, the original algorithm combination, and each single classification algorithm therein. Compare the APR index obtained by the original algorithm combination with the algorithm combination after eliminating the weak single algorithm with substandard APR index. Finally, optimize and retain the algorithm combination with the highest APR index. Step S45: Based on the diagnosis results of the combination of multiple algorithms and the idea of ​​ensemble learning, a voting decision-making method is adopted; Step S46: The parent population is screened according to a predetermined evaluation index APR threshold to retain the excellent population; if there is no chromosome in the population that meets the APR threshold, the chromosome that obtains the optimal APR index threshold in the population is retained as the excellent parent chromosome; Step S47: Generate offspring populations based on crossover and mutation of parent populations, fuse excellent parent populations with offspring populations, and then generate a new generation population based on the evaluation index APR competition and selection, and iterate repeatedly until the given evaluation index APR threshold is reached, and then terminate the iteration or the given genetic generation condition is met; Step S48: Finally, the output algorithm combination is decoded from the chromosome of the output optimal APR index to obtain the final fault diagnosis result, and the adaptability of the current system input data is judged based on whether the optimal APR index meets the given APR threshold.

2. The data-driven adaptive fault diagnosis method based on multiple evaluation indicators according to claim 1 is characterized by: In step S1, the input system data is specified in a unified standard format, and each row is specified to store a type of normal / failure mode feature data, the first column is specified as the classification label, and the other columns are different feature data; Among them, the data format supports .xlsx / .xls / .txt / .csv, and can be expanded to recognize different data types as needed.

3. The data-driven adaptive fault diagnosis method based on multiple evaluation indicators according to claim 2 is characterized by: In step S1, the input system data is discrete feature data stored according to the format requirements, and the next step of adaptive fault diagnosis is directly performed; for time-continuous data, typical time domain features and frequency domain features will be adaptively extracted, and the time domain features and frequency domain feature data will be automatically saved locally according to the discrete data format requirements, and then the next step of adaptive fault diagnosis will be performed.

4. The data-driven adaptive fault diagnosis method based on multiple evaluation indicators according to claim 1 is characterized in that: In step S2, for different feature data, a library of classic or improved data processing and fault classification algorithms is integrated to support adaptive random fusion calls and matching of different data processing and fault classification algorithm combinations to evaluate performance indicators.

Images