Hydraulic pump multi-working-condition fault identification method and device

Abstract

The invention provides a hydraulic pump multi-working-condition fault identification method and device, and belongs to the field of hydraulic pump fault detection. The method comprises the following steps: acquiring monitoring data of the hydraulic pump in different operation states; extracting features of the monitoring data under each working condition to obtain feature distribution of the monitoring signals under each working condition; determining sensitive features based on the feature distribution; determining an irrelevant first feature set and a relevant second feature set based on the sensitive features; constructing a first fault diagnosis model according to the first feature set, and identifying a corresponding fault; determining a current working condition according to the second feature set, matching a corresponding second fault diagnosis model based on the determined working condition, and inputting the second feature set into the second fault diagnosis model for fault recognition; and fusing the identification result of the first feature set and the identification result of the second feature set to obtain a comprehensive fault diagnosis result. The hydraulic pump multi-working-condition fault identification method and device can adapt to hydraulic pump fault diagnosis scenes under working condition changes, and the diagnosis precision is improved.

Description

Technical Field

[0001] The present application relates to the technical field of hydraulic pump fault detection, and in particular to a method and device for identifying multi-operating condition faults of a hydraulic pump. Background Art

[0002] With the rapid development of aviation technology, aircraft electromechanical systems play a vital role in the overall operation of aircraft. At present, the structure of aircraft electromechanical systems is becoming increasingly complex, covering many components, such as hydraulic pumps, motors, transmission devices, sensors, etc., which cooperate with each other and have significant dynamic characteristics. It is known that in modern air transportation, the safety and reliability of aircraft have attracted much attention, and the stable operation of electromechanical products has become a key factor. However, the current frequent failure of electromechanical products has brought severe challenges to the aviation field. In particular, the onboard hydraulic pump, as one of the core power sources of the aircraft electromechanical system, its operating condition directly affects the flight safety and performance of the aircraft.

[0003] In order to ensure the normal operation of electromechanical products, the industry has carried out a series of research and application on condition monitoring and fault diagnosis technology. For airborne hydraulic pumps, the current main method based on monitoring parameters is used for condition assessment. By installing various sensors on the hydraulic pump, monitoring parameters such as pressure, flow, temperature, vibration, etc. are collected in real time, and the working status of the hydraulic pump is judged based on the changes in these parameters. At the same time, there are also some diagnostic methods based on empirical thresholds, that is, the normal working range of certain parameters is pre-set. When the parameters are monitored to be outside the range, it is considered that there may be a fault.

[0004] However, in actual use, the operation of airborne hydraulic pumps is complex, affected by external input and system coupling, and the working conditions are changeable. Real-time changes in working conditions will cause changes in monitoring parameters. At this time, the characteristics of the changes still belong to the scope of normal product operation, and will overlap with the characteristics of fault changes. It is impossible to use a simple feature threshold to determine whether the parameter is caused by a change in working conditions or a fault. This can easily lead to misdiagnosis and is difficult to adapt to the high-precision fault diagnosis needs of complex nonlinear time-varying mission scenarios such as airborne. Summary of the invention

[0005] In view of this, the present application provides a hydraulic pump multi-operating condition fault identification method and device, which can adapt to the hydraulic pump fault diagnosis scenario under changing operating conditions and improve the diagnosis accuracy.

[0006] Specifically, the present application is implemented through the following technical solutions: A first aspect of the present application provides a method for identifying multi-condition faults of a hydraulic pump, the method comprising: Obtain monitoring data of the hydraulic pump under different operating conditions and record the working condition information corresponding to the monitoring data; Extracting the characteristics of the monitoring data under each working condition to obtain the characteristic distribution of the monitoring signal under each working condition; Perform sensitivity analysis based on the feature distribution to determine sensitive features; Based on the sensitive features, determining the correlation between each sensitive feature and the working condition, and determining an irrelevant first feature set and a relevant second feature set; Constructing a first fault diagnosis model according to the first feature set to identify the corresponding fault, wherein the first fault diagnosis model is used to decouple the first feature set, extract the decoupled features, and identify the corresponding fault based on the extracted features; determining a current operating condition according to the second feature set, matching a corresponding second fault diagnosis model based on the determined operating condition, and inputting the second feature set into the second fault diagnosis model for fault identification; The recognition results of the first feature set and the second feature set are integrated to obtain a comprehensive fault diagnosis result.

[0007] A second aspect of the present application provides a hydraulic pump multi-operating condition fault identification device, the device comprising an acquisition module, an extraction module, an analysis module, a determination module and an identification module; The acquisition module is used to acquire monitoring data of the hydraulic pump under different operating conditions and record the working condition information corresponding to the monitoring data; The extraction module is used to extract the characteristics of the monitoring data under each working condition to obtain the characteristic distribution of the monitoring signal under each working condition; The analysis module is used to perform sensitivity analysis based on the feature distribution to determine sensitive features; The determination module is used to determine the correlation between each sensitive feature and the working condition based on the sensitive features, and determine the irrelevant first feature set and the relevant second feature set; The identification module is used to construct a first fault diagnosis model according to the first feature set to identify the corresponding fault, wherein the first fault diagnosis model is used to decouple the first feature set, extract the decoupled features, and identify the corresponding fault based on the extracted features; The identification module is further used to determine the current operating condition according to the second feature set, match the corresponding second fault diagnosis model based on the determined operating condition, and input the second feature set into the second fault diagnosis model for fault identification; The identification module is also used to fuse the identification result of the first feature set and the identification result of the second feature set to obtain a comprehensive fault diagnosis result.

[0008] The multi-condition fault identification method and device of the hydraulic pump provided in the present application can cover various working conditions that may be encountered in the actual operation of the hydraulic pump by acquiring the monitoring data of the hydraulic pump under different operating conditions and recording the working condition information, providing a rich information basis for the subsequent accurate identification of faults, helping to adapt to the characteristics of the hydraulic pump's variable working conditions under complex working environments, and avoiding the limitations of fault diagnosis caused by a single working condition. The characteristics of the monitoring data under each working condition are extracted to obtain a characteristic distribution. The characteristic distribution under different working conditions can reflect the performance of the hydraulic pump under different working conditions, which is helpful to discover potential fault modes, and then perform sensitivity analysis based on the characteristic distribution to determine sensitive features, which can focus on features that react significantly to changes in fault states, effectively reduce the interference of irrelevant information, and improve the accuracy and efficiency of fault diagnosis.

[0009] In addition, the sensitive features are divided into the first feature set that is not related to the working condition and the second feature set that is related to the working condition, and the corresponding fault diagnosis models are constructed respectively. For the features that are not related to the working condition, the constructed first fault diagnosis model can be directly used for fault identification, avoiding the influence of the working condition changes on these features and improving the stability of diagnosis; for the features that are related to the working condition, by determining the current working condition to match the corresponding second fault diagnosis model, targeted diagnosis according to the specific working condition is achieved, and the accuracy of diagnosis is improved. Finally, the comprehensive fault diagnosis result is obtained by fusing the recognition results of the first feature set and the second feature set. Under the condition of variable working conditions, the contribution of different types of features to fault diagnosis is fully considered, and the recognition method that best matches the monitoring signal characteristics, is the fastest, and is the most accurate is directly selected, so that the final diagnosis result is more comprehensive and reliable, and can more accurately reflect the actual health status of the hydraulic pump, which helps to take effective maintenance and guarantee measures in time, reduce the equipment failure rate, and improve the reliability and safety of equipment operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 A flowchart of a first embodiment of a method for identifying multi-operating condition faults of a hydraulic pump provided in the present application; Figure 2 This is a schematic diagram of the normal state and fault state characteristic distribution shown in this application; Figure 3 This is a structural schematic diagram of Example 2 of the hydraulic pump multi-condition fault identification device provided in this application. DETAILED DESCRIPTION

[0011] Here, exemplary embodiments are described in detail, and examples thereof are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application.

[0012] The terms used in this application are only for the purpose of describing specific embodiments and are not intended to limit this application. The singular forms of "a", "said" and "the" used in this application are also intended to include plural forms, unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used in this article refers to and includes any or all possible combinations of one or more associated listed items.

[0013] It should be understood that although the terms first, second, third, etc. may be used in the present application to describe various information, these information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0014] Specific embodiments are given below to introduce the technical solution of the present application in detail.

[0015] Figure 1 This is a flow chart of the first embodiment of the method for identifying multi-condition faults of hydraulic pumps provided in this application. Figure 1 , the method provided in this embodiment may include: S101. Acquire monitoring data of the hydraulic pump under different operating conditions, and record working condition information corresponding to the monitoring data.

[0016] It should be noted that the monitoring signals of the hydraulic pump in different health states can be obtained through simulation or experimental methods to form monitoring data. Specifically, the monitoring content, that is, the monitoring signals can include parameters such as vibration signals, pressure pulsation, flow signals, temperature signals, etc., to ensure that the key operating states of the hydraulic pump are covered, and the monitoring data obtained includes monitoring signals under different working conditions, different failure modes and different failure degrees.

[0017] Among them, hydraulic pumps will face a variety of different operating states in actual work, and their performance and potential fault characteristics may be different in each state. By obtaining monitoring data under different operating states, the working conditions of hydraulic pumps under various working conditions can be fully captured. Different working condition information corresponds to monitoring data, which provides important background and reference basis for subsequent fault analysis. For example, the pressure data monitored under high load conditions has different meanings and analytical values ​​from the pressure data under low load conditions. Combined with working condition information, it can more accurately determine whether the change in monitoring data is due to normal fluctuations in working conditions or abnormal changes caused by faults. This helps to avoid misdiagnosis caused by changes in working conditions and improve the accuracy and reliability of fault diagnosis. Recording working condition information enables the subsequent fault diagnosis process to be targeted according to different working conditions. For complex and changeable hydraulic pump working scenarios, such as aircraft onboard hydraulic pumps, real-time changes in working conditions are common. Only by accurately recording working condition information can the appropriate fault diagnosis model or method be selected according to the characteristics of the working conditions in the subsequent steps to achieve effective identification of faults under different working conditions.

[0018] It should be noted that ensuring the diversity and representativeness of the operating conditions of the monitoring signal and covering as much as possible all typical operating conditions within the actual working envelope of the monitored hydraulic pump can provide data support for establishing a more accurate fault characteristic model.

[0019] In specific implementation, obtaining monitoring data of the hydraulic pump under different operating conditions may include: (1) Determine the first monitoring data set.

[0020] It should be noted that the first monitoring data set contains various monitoring signals collected initially under the operating status of the hydraulic pump. These signals may come from multiple sensors and cover information in multiple dimensions such as vibration, pressure, flow, temperature, etc. These raw data provide basic materials for subsequent analysis and processing, but may contain some information that is not critical to the identification of working conditions or is redundant.

[0021] (2) Identifying a monitoring signal of a single dimension in the first monitoring data set, and removing the monitoring signal of the single dimension to obtain a second monitoring data set.

[0022] The monitoring signals of a single dimension are selected and removed from the first monitoring data set. The purpose of this is to systematically explore the impact of each signal dimension on the working condition identification. For example, the vibration signal dimension may be removed first, leaving the pressure, flow, and temperature signals to form the second monitoring data set. Alternatively, the pressure signal dimension may be removed first, leaving the vibration, flow, and temperature signals to form the second monitoring data set. Alternatively, the vibration and pressure signal dimensions may be removed, leaving the flow and temperature signals to form the second monitoring data set. In this way, the role of each signal in the working condition identification can be evaluated individually, laying the foundation for the subsequent determination of key signals.

[0023] (3) Inputting the second monitoring data set into a neural network for operating condition identification.

[0024] The processed second monitoring data set is input into the neural network, and the powerful learning and classification capabilities of the neural network are used to determine the working condition of the hydraulic pump. The neural network can learn and identify based on the association between the signal features in the data set and the known working condition patterns. If the neural network can accurately identify the working condition, it means that the current remaining signal combination can reflect the working condition information of the hydraulic pump to a certain extent; conversely, if the recognition result is invalid, the signal combination needs to be further adjusted.

[0025] (4) If the operating condition identification result is valid, the importance of the removed monitoring signals is sorted according to their influence on the operating condition identification, and the process returns to the step of identifying the monitoring signals of a single dimension in the first monitoring data set.

[0026] If the result of the working condition identification is valid, the importance of the removed monitoring signals is ranked according to their influence on the working condition identification. This means that the importance of the signal to the working condition identification is judged by comparing the change in the accuracy of the working condition identification before and after the removal of a certain signal. For example, if the accuracy of the working condition identification drops significantly after the vibration signal is removed, then the importance of the vibration signal is high. Then return to the step of removing the single-dimensional signal and continue to explore other signal combinations to find a better combination of monitoring signals to further improve the accuracy and efficiency of working condition identification.

[0027] Specifically, by sorting by importance, we can filter out signals that play a key role in condition identification and remove those that are relatively unimportant or redundant. For example, in a monitoring data set containing signals such as vibration, pressure, flow, and temperature, after sorting by importance, it may be found that the vibration signal contributes the most to condition identification, while the temperature signal is relatively small. In this way, we can focus on the vibration signal in subsequent monitoring, while considering whether the monitoring system can be appropriately simplified and the collection or processing of temperature signals can be reduced, thereby optimizing the combination of monitoring signals and reducing the cost and complexity of data processing without affecting the accuracy of condition identification.

[0028] In addition, the importance of each signal may change under different working conditions of the hydraulic pump. For example, in the startup phase, the current signal may be more critical to determine whether the startup process is normal; while in the stable operation phase, the importance of pressure and flow signals may be more prominent. By sorting the importance, the most appropriate combination of monitoring signals can be determined for different working conditions.

[0029] (5) If the working condition identification result is invalid, retain the monitoring data set before the single-dimensional signal is removed from the second monitoring data set as the optimal set of monitoring signals for working condition identification, and obtain monitoring data of the hydraulic pump under different operating conditions based on the optimal set of monitoring signals.

[0030] Specifically, if the working condition identification result is invalid, the monitoring data set before the last removal of the single-dimensional signal from the second monitoring data set is retained as the optimal set of monitoring signals for working condition identification, wherein, when there is only one monitoring data set before the last removal of the single-dimensional signal, the monitoring data set is directly selected as the optimal set of monitoring signals; when there are multiple monitoring data sets before the last removal of the single-dimensional signal, the following operations are performed: (i) evaluating the accuracy of the output recognition result of the neural network for operating condition recognition and the complexity of the neural network for operating condition recognition, and determining the optimal set of monitoring signals based on the accuracy and complexity.

[0031] (ii) If both the accuracy and the complexity do not meet the preset requirements, a comprehensive ranking is performed according to the recognition results of the accuracy and complexity, and the one with the highest ranking is selected as the optimal set of monitoring signals, wherein the higher the ranking, the more the accuracy and complexity meet the preset requirements.

[0032] It should be noted that the accuracy and complexity here do not meet the preset requirements, which means that when the working conditions can be identified (that is, the working conditions can be accurately identified, but we have made higher requirements on the accuracy according to needs), the accuracy and complexity are further limited, among which the preset requirements are set according to actual needs.

[0033] It should be noted that when the working condition identification result is invalid, the current signal combination is not simply abandoned, but the performance of the neural network is deeply evaluated. The accuracy of the output recognition result reflects the accuracy of the neural network's judgment of the working condition under the current signal combination, while the complexity of the neural network is related to the computing resource requirements and efficiency in practical applications. For example, if a signal combination can achieve working condition identification, but requires a very complex neural network structure, it may result in long computing time and high resource consumption, which may not be ideal in practical applications.

[0034] The optimal set of monitoring signals is determined based on accuracy and complexity. If the current accuracy and complexity meet the preset requirements, the current second monitoring data set is used as the optimal set for subsequent monitoring data acquisition of the hydraulic pump under different operating conditions. If there are multiple second monitoring data sets that meet the preset requirements, the second monitoring data set with the highest accuracy and the lowest complexity is selected as the optimal set of monitoring signals. If none of them meet the preset requirements, they are comprehensively sorted according to the recognition results of accuracy and complexity, and the top-ranked ones are selected as the optimal set of monitoring signals. In this way, a relatively better choice can be found among multiple undesirable signal combinations. The higher the ranking, the more the accuracy and complexity meet the preset requirements, which provides a clear standard for selection. For example, if the accuracy of several signal combinations is low, but the neural network complexity of one of the combinations is relatively low, then it may be ranked in front in the comprehensive sorting and be preferentially selected as the optimal set of monitoring signals, thereby balancing the relationship between accuracy and efficiency to a certain extent, avoiding the problem of only pursuing high accuracy and ignoring resource consumption, and avoiding the situation where only complexity is considered and the diagnosis is inaccurate.

[0035] S102: extracting the characteristics of the monitoring data under each working condition to obtain characteristic distribution of the monitoring signal under each working condition.

[0036] It should be noted that the monitoring data itself may contain a large amount of complex information, and it is often difficult to accurately judge the operating status and potential faults of the hydraulic pump by directly analyzing these raw data. By extracting features, the original monitoring data can be converted into a more representative and analyzable form. For example, features such as vibration kurtosis can be extracted from vibration signals and pressure pulsation peak-to-peak values ​​can be extracted from pressure signals. These features can reflect the working status of the hydraulic pump from different angles, such as the degree of wear of mechanical parts and the stability of the hydraulic system. Specifically, the monitoring signal features can be extracted according to the working principle and failure mechanism of the hydraulic pump.

[0037] The characteristic distribution of the monitoring signal under each working condition is obtained, which enables effective comparison and analysis between different working conditions and between healthy and faulty states. Under different working conditions, the operating parameters and performance of the hydraulic pump will be different, and its characteristic distribution will also show corresponding differences. By comparing the characteristic distribution under different working conditions, we can understand the normal operating characteristic range of the hydraulic pump under various working conditions, so as to better judge whether the characteristics of the monitoring data under a certain working condition exceed the normal range, and then find potential signs of failure.

[0038] It should also be noted that the feature distribution provides a quantitative reference standard for fault diagnosis. By extracting and analyzing the features of a large amount of monitoring data, a mapping relationship between the feature value and the health of the hydraulic pump can be established. For example, as the degree of fault increases, a certain feature value may show a regular change trend, such as gradually increasing or decreasing. This quantitative relationship makes it possible to accurately evaluate the health status of the hydraulic pump and judge the severity of the fault based on the feature value of the monitoring data in the actual fault diagnosis process, rather than just qualitatively judging whether there is a fault. The accuracy and effectiveness of the feature distribution directly affect the judgment of sensitive features, the analysis of working condition correlation, and the performance of the fault diagnosis model in the subsequent steps, thereby affecting the accuracy and reliability of the entire hydraulic pump fault identification method. If the feature extraction is inaccurate or the feature distribution analysis is not comprehensive, it may lead to the incorrect selection of sensitive features in the subsequent steps, and the inability to accurately judge the relationship between features and working conditions, which ultimately affects the results of fault diagnosis.

[0039] S103: Perform sensitivity analysis based on the feature distribution to determine sensitive features.

[0040] In addition, by comparing the feature distribution in the healthy state (normal state) and the fault state, we can find out those features that change significantly in the fault state. These features are fault-sensitive features and are the key basis for distinguishing between healthy and fault states. Figure 2 Please refer to the normal state and fault state characteristic distribution diagram shown in this application. Figure 2 Specifically, by identifying the characteristic distribution patterns of the healthy state and the fault state, we can find out the sensitive features that can obviously distinguish the fault state from the healthy state. For example, as the fault increases, the degree of distinction between a candidate feature and the feature in the normal state becomes greater and greater, which is a sensitive feature. On the contrary, if a feature shows an indistinguishable trend between the healthy state and the fault state, it is a non-sensitive feature and needs to be eliminated.

[0041] In a hydraulic pump system operating under multiple working conditions, changes in working conditions will cause changes in the characteristics of the monitoring data. However, the advantage of sensitive features is that their ability to indicate faults under different working conditions is relatively stable, and they can overcome the impact of changes in working conditions to a certain extent. Even if the working conditions change, the sensitive features can still highlight the difference between the fault state and the healthy state. This enables the fault diagnosis method to accurately identify faults when facing complex and changeable working conditions, such as the working conditions of aircraft onboard hydraulic pumps under different flight phases and mission loads, enhancing the adaptability and robustness of the diagnosis method and meeting the needs of high-precision fault diagnosis.

[0042] S104. Determine the correlation between each sensitive feature and the working condition based on the sensitive features, and determine an irrelevant first feature set and a relevant second feature set.

[0043] It should be noted that the correlation between each sensitive feature and the working condition is determined based on the sensitive features, and the irrelevant first feature set and the relevant second feature set are determined, including: (1) Determine the operating envelope of the hydraulic pump.

[0044] It should be noted that the operating envelope describes the performance limit and operating range of the hydraulic pump under various working conditions. It covers the working state of the hydraulic pump under different working conditions (such as different speeds, loads, temperatures, etc.). The operating envelope defines the range of normal operation and possible failure of the hydraulic pump. All fault analysis and feature judgment will be carried out within this range. By determining the operating envelope, it is helpful to fully understand the performance of the hydraulic pump under different working conditions and provide a background framework for accurately determining the target fault.

[0045] (2) Determine a plurality of target faults based on the working envelope.

[0046] Based on the known working envelope, the typical fault types that may occur in hydraulic pumps under different working conditions can be clearly identified. These are the target faults. For example, wear faults may be more likely to occur under high load conditions, while seal failure and other faults may occur under high temperature conditions. Determining the target faults makes the subsequent analysis of sensitive features more targeted, and can focus on the feature changes related to these specific faults, providing a clear direction for establishing an accurate fault diagnosis model, avoiding non-targeted feature analysis, and improving the efficiency of fault diagnosis.

[0047] (3) According to the different degrees of each target fault, the characteristic value variation law of the sensitive feature is fitted to obtain the mapping relationship between the characteristic value and the health degree under each target fault.

[0048] It should be noted that the mapping relationship between the characteristic value and the health degree under each target fault is obtained by fitting the characteristic value change law of the sensitive feature according to the different degrees of each target fault, which provides an accurate quantitative basis for the health status assessment of the hydraulic pump. Through this mapping relationship, the monitored sensitive characteristic value can be converted into an intuitive judgment of the health degree of the hydraulic pump, which is no longer limited to a simple judgment of whether there is a fault, but can understand the degree of development of the fault in more detail. For example, for a certain sensitive feature, its characteristic value is in a specific range under normal conditions. As the degree of the fault deepens, the characteristic value will change according to a certain law (such as gradually increasing or decreasing). Based on this established mapping relationship, in actual monitoring, as long as the characteristic value is obtained, the current health status of the hydraulic pump can be accurately assessed, thereby providing an accurate timing judgment for timely maintenance measures.

[0049] This quantitative mapping relationship can effectively reduce the errors caused by subjective judgment and improve the accuracy and reliability of fault diagnosis. In a complex hydraulic pump system, the characteristic performance under different fault modes and degrees may be more complex, and it is difficult to accurately judge the fault situation based on experience or qualitative analysis alone. Through the mapping relationship between eigenvalues ​​and health, the diagnosis process is more scientific and objective, which helps to avoid misdiagnosis and missed diagnosis, ensures a more accurate assessment of the health status of the hydraulic pump, and thus improves the safety and stability of the entire system operation. In specific implementation, a variety of regression models (such as linear regression, polynomial fitting, or Gaussian process regression) can be used to fit the changing rules of sensitive eigenvalues ​​to form a mapping relationship between the degree of fault and the eigenvalue.

[0050] (4) Filter the features that do not change with the working conditions as the first feature set, and filter the features that change with the working conditions as the second feature set.

[0051] It should be noted that the division of sensitive features into a first feature set that does not change with the working conditions and a second feature set that changes with the working conditions provides a reasonable basis for constructing different types of fault diagnosis models. Since the first feature set is not affected by changes in the working conditions, it reflects the essential characteristics of the internal structure and performance of the hydraulic pump. The first fault diagnosis model constructed based on these features can identify faults more stably and avoid the interference of working condition fluctuations on the diagnosis results. For example, the natural vibration frequency characteristics of certain mechanical components belong to the first feature set. No matter what working condition the hydraulic pump is in, as long as these characteristics change abnormally, they are likely to indicate a specific type of fault. The second feature set is related to the working conditions, and its changes reflect the impact of the working conditions on the operation of the hydraulic pump. By constructing a second fault diagnosis model that matches the working conditions, faults can be diagnosed more accurately according to the changing laws of these features under different working conditions. For example, under different flow conditions, pressure pulsation-related features belong to the second feature set. Establishing corresponding diagnosis models for different flow conditions can improve the accuracy of fault diagnosis under different working conditions.

[0052] By distinguishing the first feature set from the second feature set, it is possible to better adapt to the actual situation of the hydraulic pump's multi-operating conditions and improve diagnostic efficiency. Specifically, if the correlation between the features and the operating conditions is not distinguished, the complex changes of all features under various operating conditions need to be considered when performing fault diagnosis, which is computationally intensive and prone to errors. In specific implementation, statistical methods (such as correlation coefficients and mutual information) can be used to analyze the relationship between feature values ​​and operating parameters (such as speed and load). Highly correlated features are classified into the second feature set, and low-correlation features are classified into the first feature set.

[0053] S105. Construct a first fault diagnosis model according to the first feature set to identify corresponding faults, wherein the first fault diagnosis model is used to decouple the first feature set, extract decoupled features, and identify corresponding faults based on the extracted features.

[0054] It should be noted that the method for constructing the first fault diagnosis model includes: (1) Based on the features of the monitoring data extracted under each working condition, the correlation between the features is calculated.

[0055] It should be noted that the contribution of each feature to the fault mode can be evaluated by calculating the Pearson correlation coefficient, mutual information, covariance, etc. between the features, and then for the features monitored under different working conditions, it is possible to identify which features have a high degree of consistency in changes under healthy and faulty states, so as to determine which features can provide independent and effective information for fault diagnosis and which features may be redundant.

[0056] It should be noted that the operating state of hydraulic pumps is complex, and multiple features may have mutual influence and coordinated changes. For example, vibration features and pressure features may change simultaneously under certain fault conditions. By calculating the correlation between them, we can find the laws hidden behind the data, which helps to more comprehensively understand the operating mechanism of hydraulic pumps and comprehensively evaluate the health status of equipment from multiple dimensions, rather than just being limited to the analysis of a single feature.

[0057] (2) Perform feature decoupling processing based on the calculated feature correlation relationship.

[0058] It should be noted that dimensionality reduction methods such as principal component analysis can be used to transform highly correlated features into a set of unrelated principal components, thereby removing redundant information. Feature decoupling processing can transform complex, interrelated features into relatively independent features. For example, some feature fluctuations caused by changes in operating conditions may mask feature changes that are actually caused by faults. Decoupling can separate features that are directly related to the fault and are not affected by the operating conditions, reduce the impact of redundant information on fault diagnosis, and improve the sensitivity and recognition ability of the diagnostic model to fault features.

[0059] (3) The decoupled features are input into the neural network model to identify the corresponding faults and obtain the first fault diagnosis model.

[0060] It should be noted that the decoupled features are more suitable as the input of the neural network model. When processing relatively independent and concise features, the neural network model can better learn the mapping relationship between features and faults, and avoid problems such as model training difficulties, overfitting or underfitting due to complex correlations between input features. Therefore, inputting the decoupled features into the neural network model helps improve the training efficiency and accuracy of the model.

[0061] S106. Determine the current operating condition according to the second feature set, match the corresponding second fault diagnosis model based on the determined operating condition, and input the second feature set into the second fault diagnosis model for fault identification.

[0062] It should be noted that determining the current operating condition according to the second feature set and matching the corresponding second fault diagnosis model based on the determined operating condition includes: (1) Based on the second feature set, a working condition identification algorithm is used to classify the working condition of the current operating state of the hydraulic pump to obtain the working condition category of the current hydraulic pump.

[0063] It should be noted that, based on the monitoring data of the current operation of the hydraulic pump, the features related to the working conditions (such as pressure, flow, vibration, temperature, etc.) are selected as the second feature set, and then the appropriate working condition classification algorithm (such as K-means clustering, support vector machine (SVM), decision tree, neural network, etc.) is selected for working condition identification, and the working condition category of the current hydraulic pump operation status can be output. Specifically, the second feature set contains sensitive features that change with the working conditions. These features can accurately reflect the differences in the operating conditions of the hydraulic pump under different working conditions. By analyzing these features through the working condition identification algorithm, it is possible to fully explore the working condition information contained in the second feature set and accurately classify the current operating state of the hydraulic pump into the corresponding working condition category. In this way, the changes in the working conditions can be tracked in real time, ensuring that the fault diagnosis method always matches the current working conditions, thereby effectively avoiding misdiagnosis caused by changes in working conditions.

[0064] (2) Determine the feature distribution characteristics of the second feature set.

[0065] It should be noted that for each feature in the second feature set, its mean, variance, standard deviation, maximum value, minimum value and other statistics can be calculated. These statistics can reflect the basic information such as the central tendency and degree of dispersion of the features under different working conditions. Therefore, by comparing the differences in these statistics under different working conditions, we can preliminarily understand the distribution characteristics of the features under different working conditions. Furthermore, the feature distribution characteristics of the second feature set can be understood by drawing feature histograms or probability density graphs, calculating the correlation coefficients between features, and constructing feature correlation matrices or association graphs. In addition, if the monitoring data contains time series information, the feature distribution characteristics of the second feature set can also be understood by calculating the time series characteristics and drawing a curve of the feature changes over time.

[0066] (3) Based on the current operating condition category and the characteristic distribution features, a second fault diagnosis model corresponding to the current operating condition category is selected from a plurality of pre-established second fault diagnosis models.

[0067] It should be noted that, under different working conditions, multiple independent fault diagnosis models are constructed, and each model is established for specific working conditions and characteristic distribution characteristics. Specifically, the training process of the pre-established multiple second fault diagnosis models includes: (i) Acquire monitoring data of the hydraulic pump under various working conditions, and extract sensitive features of the monitoring data as training samples.

[0068] Under different operating conditions of the hydraulic pump (such as different speeds, loads, etc.), real-time monitoring data of the hydraulic pump is collected. These data usually include speed, vibration, pressure, flow signals, etc. By obtaining the monitoring data of the hydraulic pump under various working conditions, it is possible to comprehensively cover various working conditions that may be encountered in the actual operation of the hydraulic pump, providing a rich and diverse source of samples for the training model. The sensitive features in these monitoring data are extracted as training samples because the sensitive features can more accurately reflect the health status and fault information of the hydraulic pump under different working conditions.

[0069] (ii) Based on the training samples under different working conditions, multiple independent fault diagnosis datasets are constructed, where each dataset corresponds to one working condition.

[0070] It should be noted that all the collected training samples are divided into different working condition categories, and each working condition corresponds to an independent data set. The samples in each data set should contain the monitoring data and characteristics of the hydraulic pump under various health and fault conditions under the working condition. Among them, the fault data under different working conditions have their own characteristics and distribution laws. By constructing data sets separately, the neural network algorithm can better adapt to the fault modes under different working conditions, thereby deriving the intrinsic relationship between the sensitive features and faults under each working condition, and improving the diagnostic accuracy of the model under various working conditions.

[0071] In addition, multiple independent data sets help the trained model better distinguish different working conditions. In actual operation, accurately judging the current working condition of the hydraulic pump is a prerequisite for effective fault diagnosis. By training on different working condition data sets, the model can learn the difference characteristics between each working condition. Therefore, in actual applications, when new monitoring data is input, it can more accurately judge the working condition it belongs to, providing a reliable basis for the subsequent selection of appropriate fault diagnosis models.

[0072] (iii) For each data set, a neural network algorithm is used to train the data to obtain a second fault diagnosis model for the corresponding working condition.

[0073] According to the characteristics of the data set and the complexity of the problem, select a suitable neural network algorithm for training. It should be noted that the data set corresponding to each working condition is trained using a neural network. The goal of the training is to enable the neural network to learn the complex relationship between features and fault states by optimizing the loss function (such as cross entropy loss, mean square error loss, etc.). For example, use the back propagation algorithm and gradient descent method to adjust the network weights and train the network until the ideal convergence effect is achieved.

[0074] (iiii) The trained multiple second fault diagnosis models are stored as a model set, and the corresponding operating condition category information is labeled for each model.

[0075] Each trained fault diagnosis model will be labeled with the corresponding working condition category information. In this way, in actual application, when the working condition of the hydraulic pump is identified, the model matching the current working condition can be selected from the stored model set for fault diagnosis. Specifically, when the model to be used is determined according to the current working condition, the corresponding model can be quickly located through the labeled information, avoiding blind search and selection among many models and improving the efficiency of fault diagnosis.

[0076] It should also be noted that the training process of the pre-established multiple second fault diagnosis models includes: identifying the working condition of the collected monitoring data, obtaining the working condition label, and forming a hydraulic pump working condition data set based on the monitoring signal and the working condition label; inputting the hydraulic pump working condition data set into the neural network for training; and constructing a mapping relationship between the characteristics of the monitoring signal and the working condition based on the trained neural network model to form a working condition and feature mapping model. It should be noted that the working condition identification and obtaining the working condition label of the collected monitoring data can give the monitoring data clear working condition attribution information, which makes the subsequent hydraulic pump working condition data set constructed based on these data more accurate and targeted, and each data sample is closely related to the corresponding working condition. The hydraulic pump working condition data set is formed based on the monitoring signal with the working condition label. Compared with the simple monitoring data, the data set contains more information about the working condition. This rich information content provides a more comprehensive input for the neural network training, so that the model can learn the relationship between the monitoring signal and the working condition from multiple angles. During the training process, the model can not only learn the characteristic changes of the monitoring signal itself, but also understand the relationship between these changes and different working conditions, thereby improving the model's sensitivity and recognition ability to working condition changes.

[0077] It should also be noted that the mapping relationship between the characteristics of the monitoring signal and the working condition is constructed based on the trained neural network model to form a working condition and feature mapping model, which provides important technical support for subsequent fault diagnosis. Through this mapping model, the working condition decoupling can be achieved in the fault diagnosis process, that is, the working condition related information and the fault related information in the monitoring signal are separated. When facing the actual monitoring data, the working condition and feature mapping model is first used to determine the current working condition, and then the corresponding second fault diagnosis model is selected for fault identification. This can avoid the interference of working condition changes on fault diagnosis and improve the accuracy of diagnosis. Specifically, the mapping relationship between the characteristics and the working condition can be based on the forward propagation of the neural network, the monitoring signal is input and the output result is obtained, which represents the predicted category of the current working condition of the hydraulic pump, and then the mapping relationship between the characteristics of the monitoring signal and the working condition is constructed through the output of the model.

[0078] In addition, the second feature set is input into the second fault diagnosis model for fault identification, including: inputting the second feature set into the second fault diagnosis model for operating condition identification; judging whether the operating condition identification result is valid; if invalid, selecting an extended monitoring point with an importance higher than a threshold from outside the hydraulic pump, and obtaining an extended monitoring signal to supplement the second feature set.

[0079] It should be noted that the second feature set containing sensitive features that change with the working condition is input into the pre-trained second fault diagnosis model, and the model is used to identify the working condition of the current hydraulic pump. After the working condition identification result is obtained, its effectiveness needs to be judged. The effectiveness judgment can be based on a variety of factors, such as the confidence of the model output, the consistency with the historical working condition data, and whether it conforms to the expected working condition change law. If the confidence of the working condition identification result is high and consistent with the known hydraulic pump operation mode and working condition change logic, then the identification result can be considered valid; conversely, if the confidence is low, or there is an obvious contradiction with the actual situation (such as suddenly identifying an extreme fault condition during the normal operation of the equipment without a reasonable reason), it is judged to be invalid. In this way, the working condition identification results that may have problems can be screened out to avoid subsequent fault diagnosis errors due to incorrect working condition judgments.

[0080] When the working condition identification result is invalid, extended monitoring points with an importance higher than the threshold are selected from outside the hydraulic pump. The selection of these extended monitoring points is based on a deep understanding of the hydraulic pump system and an analysis of the fault mode. For example, in some cases, although the oil temperature of the hydraulic pump is not part of the second feature set for routine monitoring, it may change significantly under specific faults (such as cooling system failure) and has an important indicative role in fault diagnosis. In this case, the oil temperature monitoring point may be selected as an extended monitoring point. The setting of the importance threshold can be determined through data analysis, expert experience or experimental verification to ensure that the selected extended monitoring points can provide valuable supplementary information and improve the accuracy of fault diagnosis.

[0081] Furthermore, the monitoring signals of the extended monitoring points are obtained and added to the second feature set. This is equivalent to increasing the information dimension of fault diagnosis, so that the model can have a more comprehensive understanding of the operating status of the hydraulic pump. The newly added signal may change the distribution characteristics of the feature set, thereby affecting the working condition identification and fault diagnosis results. For example, after adding the oil temperature signal, the originally fuzzy working condition boundary may become clear, or reveal previously undiscovered fault characteristics, which will help the model to more accurately identify the type and degree of faults, improve the reliability of fault diagnosis, ensure accurate judgments in complex working conditions and fault conditions, and ensure the safe and stable operation of the hydraulic pump.

[0082] S107 , integrating the recognition result of the first feature set and the recognition result of the second feature set to obtain a comprehensive fault diagnosis result.

[0083] It should be noted that the first feature set mainly contains sensitive features that are unrelated to the working conditions. These features can stably reflect the essential state of the internal structure and performance of the hydraulic pump, are not affected by changes in working conditions, and have high accuracy in identifying certain specific types of faults. The second feature set focuses on sensitive features that change with working conditions. Its advantage is that it can accurately capture fault information according to changes in working conditions, and is particularly suitable for the diagnosis of working condition-related faults. Fusion of the recognition results of the two can give full play to their respective advantages, complement each other's information, avoid the limitations of single feature set diagnosis, and thus significantly improve the overall accuracy of fault diagnosis.

[0084] The multi-condition fault identification method of the hydraulic pump provided in this embodiment can cover various actual working conditions of the hydraulic pump by acquiring monitoring data under different operating conditions and recording working condition information, providing a comprehensive information basis for subsequent accurate diagnosis, adapting to complex and changeable working conditions, and avoiding diagnostic limitations caused by a single working condition; by extracting the monitoring data features under each working condition to obtain feature distribution, and then determining sensitive features through sensitivity analysis, the key features sensitive to fault reactions can be effectively screened out, reducing irrelevant information interference, and improving the accuracy and efficiency of fault diagnosis; the sensitive features are divided into feature sets related to and unrelated to working conditions, and corresponding diagnostic models are constructed respectively, so as to realize targeted diagnosis according to working condition characteristics. The multi-condition fault identification method of the hydraulic pump proposed in this application has good adaptability and robustness. The method can adaptively extract fault sensitive features according to different working conditions of the hydraulic pump and the characteristics of the monitoring signal. At the same time, when the working state of the hydraulic pump changes, the sensitive features unrelated to the working condition are preferentially used for fault diagnosis; the features related to the working condition are decoupled by constructing two types of working condition mapping models, and the fault diagnosis is performed by selecting the optimal non-interference method, so as to improve the diagnostic accuracy and prevent false alarms caused by changes in working conditions.

[0085] Specifically, by determining the working envelope of the hydraulic pump, multiple target faults are determined based on this, and then the characteristic value change law of the sensitive features is fitted according to the different degrees of each target fault, and the features that do not change with the working conditions are screened out as the sensitive feature set that is independent of the working conditions. These features reflect the essential characteristics of the internal structure and performance of the hydraulic pump. Regardless of the working condition of the hydraulic pump, the changes in these features are mainly related to the health of the internal structure and are not affected by external working condition fluctuations. When the working state of the hydraulic pump changes, these sensitive features that are independent of the working condition are used preferentially for fault diagnosis. Due to their stability, they can provide a relatively constant reference benchmark for fault diagnosis, so they can accurately indicate the existence of faults when the working conditions fluctuate, avoiding the interference caused by changes in working conditions, and improving the initial accuracy and reliability of fault diagnosis.

[0086] In addition, based on the features of the extracted monitoring data under each working condition, the correlation between the features is calculated, and according to the calculated feature correlation, the principal component analysis and other dimensionality reduction methods are used to perform feature decoupling. The highly correlated features are converted into a set of unrelated principal components, and redundant information such as feature fluctuations caused by changes in working conditions are removed. The features that are directly related to the fault and are not affected by the working conditions are separated. Through decoupling processing, independent vibration feature principal components can be obtained, so that the fault diagnosis model can more accurately identify the vibration changes caused by the fault, rather than the fluctuations caused by the working condition changes, thereby improving the sensitivity and recognition ability of the diagnosis model to the fault features. Finally, a working condition and feature mapping model is constructed to prevent false alarms caused by working condition changes.

[0087] Corresponding to the aforementioned embodiment of a hydraulic pump multi-operating condition fault identification method, the present application also provides an embodiment of a hydraulic pump multi-operating condition fault identification device.

[0088] Figure 3 This is a structural diagram of the second embodiment of the multi-condition fault identification device for hydraulic pumps provided in this application. Figure 3 , the device provided in this embodiment includes an acquisition module 310, an extraction module 320, an analysis module 330, a determination module 340 and an identification module 350; The acquisition module 310 is used to acquire monitoring data of the hydraulic pump under different operating conditions and record the operating condition information corresponding to the monitoring data; The extraction module 320 is used to extract the characteristics of the monitoring data under each working condition to obtain the characteristic distribution of the monitoring signal under each working condition; The analysis module 330 is used to perform sensitivity analysis based on the feature distribution to determine sensitive features; The determination module 340 is used to determine the correlation between each sensitive feature and the working condition based on the sensitive features, and determine the irrelevant first feature set and the relevant second feature set; The identification module 350 is used to construct a first fault diagnosis model according to the first feature set to identify the corresponding fault, wherein the first fault diagnosis model is used to decouple the first feature set, extract the decoupled features, and identify the corresponding fault based on the extracted features; The identification module 350 is further used to determine the current operating condition according to the second feature set, match the corresponding second fault diagnosis model based on the determined operating condition, and input the second feature set into the second fault diagnosis model for fault identification; The identification module 350 is further used to fuse the identification result of the first feature set and the identification result of the second feature set to obtain a comprehensive fault diagnosis result.

[0089] The device of this embodiment can be used to perform Figure 1 The steps, specific implementation principles and implementation processes of the method embodiment shown are similar and will not be repeated here.

[0090] The implementation process of the functions and effects of each unit in the above-mentioned device is specifically described in the implementation process of the corresponding steps in the above-mentioned method, and will not be repeated here.

[0091] For the device embodiment, since it basically corresponds to the method embodiment, the relevant parts can refer to the partial description of the method embodiment. The device embodiment described above is only schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present application scheme. A person of ordinary skill in the art can understand and implement it without paying any creative work.

[0092] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A method for identifying multi-condition faults of a hydraulic pump, characterized in that: The method comprises: Obtain monitoring data of the hydraulic pump under different operating conditions and record the working condition information corresponding to the monitoring data; Extracting the characteristics of the monitoring data under each working condition to obtain the characteristic distribution of the monitoring signal under each working condition; Perform sensitivity analysis based on the feature distribution to determine sensitive features; Based on the sensitive features, determining the correlation between each sensitive feature and the working condition, and determining an irrelevant first feature set and a relevant second feature set; Constructing a first fault diagnosis model according to the first feature set to identify the corresponding fault, wherein the first fault diagnosis model is used to decouple the first feature set, extract the decoupled features, and identify the corresponding fault based on the extracted features; determining a current operating condition according to the second feature set, matching a corresponding second fault diagnosis model based on the determined operating condition, and inputting the second feature set into the second fault diagnosis model for fault identification; The recognition results of the first feature set and the second feature set are integrated to obtain a comprehensive fault diagnosis result.

2. The method according to claim 1, characterized in that The method for constructing the first fault diagnosis model includes: Based on the features of the monitoring data extracted under each working condition, the correlation between the features is calculated; Perform feature decoupling processing based on the calculated feature association relationship; The decoupled features are input into a neural network model to identify corresponding faults, thereby obtaining a first fault diagnosis model.

3. The method according to claim 1, characterized in that The step of judging the correlation between each sensitive feature and the working condition based on the sensitive features, and determining an irrelevant first feature set and a relevant second feature set, comprises: Determine the operating envelope of the hydraulic pump; determining a plurality of target faults based on the operating envelope; According to the different degrees of each target fault, the characteristic value variation law of the sensitive feature is fitted to obtain the mapping relationship between the characteristic value and the health degree under each target fault; The features that do not change with the working conditions are selected as the first feature set, and the features that change with the working conditions are selected as the second feature set.

4. The method according to claim 1, characterized in that: The determining the current operating condition according to the second feature set, and matching the corresponding second fault diagnosis model based on the determined operating condition, comprises: Based on the second feature set, a working condition identification algorithm is used to classify the working condition of the current operating state of the hydraulic pump to obtain the working condition category of the current hydraulic pump; determining a feature distribution characteristic of the second feature set; Based on the current operating condition category and the characteristic distribution features, a second fault diagnosis model corresponding to the current operating condition category is selected from a plurality of pre-established second fault diagnosis models.

5. The method according to claim 4, characterized in that The training process of the pre-established plurality of second fault diagnosis models includes: Acquire monitoring data of the hydraulic pump under various working conditions, and extract sensitive features of the monitoring data as training samples; Based on the training samples under different working conditions, multiple independent fault diagnosis data sets are constructed, where each data set corresponds to one working condition; For each data set, a neural network algorithm is used to train the data to obtain a second fault diagnosis model for the corresponding working condition; The trained multiple second fault diagnosis models are stored as a model set, and the corresponding operating condition category information is labeled for each model.

6. The method according to claim 4, characterized in that The training process of the pre-established plurality of second fault diagnosis models includes: Perform working condition identification on the collected monitoring data, obtain working condition labels, and form a hydraulic pump working condition data set based on the monitoring signals and working condition labels; Inputting the hydraulic pump operating condition data set into a neural network for training; Based on the trained neural network model, the mapping relationship between the characteristics of the monitoring signal and the working conditions is constructed to form a working condition and feature mapping model.

7. The method according to claim 1, characterized in that The obtaining of monitoring data of the hydraulic pump under different operating conditions includes: determining a first monitoring data set; Identifying a monitoring signal of a single dimension in the first monitoring data set, and removing the monitoring signal of the single dimension to obtain a second monitoring data set; Inputting the second monitoring data set into a neural network for operating condition identification; If the result of the working condition identification is valid, the importance of the removed monitoring signals is sorted according to their influence on the working condition identification, and the process of identifying the monitoring signals of a single dimension in the first monitoring data set is returned; If the working condition identification result is invalid, the monitoring data set before the last removal of the single dimensional signal from the second monitoring data set is retained as the optimal set of monitoring signals for working condition identification, and the monitoring data of the hydraulic pump under different operating conditions is obtained based on the optimal set of monitoring signals.

8. The method according to claim 7, characterized in that If the working condition identification result is invalid, retaining the monitoring data set before the last removal of the single-dimensional signal from the second monitoring data set as the optimal set of monitoring signals for working condition identification, including: Evaluate the accuracy of the output recognition result of the neural network for operating condition recognition and the complexity of the neural network for operating condition recognition, and determine the optimal set of monitoring signals based on the accuracy and complexity; If both the accuracy and the complexity do not meet the preset requirements, a comprehensive ranking is performed according to the recognition results of the accuracy and complexity, and the top ranked ones are selected as the optimal set of monitoring signals, wherein the higher the ranking, the more the accuracy and complexity meet the preset requirements.

9. The method according to claim 6, characterized in that The step of inputting the second feature set into the second fault diagnosis model for fault identification includes: Inputting the second feature set into the second fault diagnosis model to perform operating condition identification; Determine whether the working condition identification result is valid; If invalid, an extended monitoring point with an importance higher than a threshold is selected from outside the hydraulic pump, and an extended monitoring signal is obtained to supplement the second feature set.

10. A hydraulic pump multi-operating condition fault identification device, characterized in that: The device includes an acquisition module, an extraction module, an analysis module, a determination module and an identification module; The acquisition module is used to acquire monitoring data of the hydraulic pump under different operating conditions and record the working condition information corresponding to the monitoring data; The extraction module is used to extract the characteristics of the monitoring data under each working condition to obtain the characteristic distribution of the monitoring signal under each working condition; The analysis module is used to perform sensitivity analysis based on the feature distribution to determine sensitive features; The determination module is used to determine the correlation between each sensitive feature and the working condition based on the sensitive features, and determine the irrelevant first feature set and the relevant second feature set; The identification module is used to construct a first fault diagnosis model according to the first feature set to identify the corresponding fault; The identification module is further used to determine the current operating condition according to the second feature set, match the corresponding second fault diagnosis model based on the determined operating condition, and input the second feature set into the second fault diagnosis model for fault identification; The identification module is also used to fuse the identification result of the first feature set and the identification result of the second feature set to obtain a comprehensive fault diagnosis result.

Images