Pneumatic actuating mechanism fault diagnosis method based on ICA
By integrating multi-source heterogeneous data using an ICA-based approach and employing whitening processing and a multi-level fault classification model, the problem of accuracy in fault location and classification of pneumatic actuators was solved, achieving efficient fault diagnosis.
Patent Information
- Application Number
- CN202511098402.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-08-06
- Publication Date
- 2025-12-16
AI Technical Summary
Existing technologies struggle to accurately locate and classify faults in pneumatic actuators. Traditional methods rely on single or limited sensor data, failing to capture subtle features in the early stages of faults, and have shortcomings in fault mode identification.
By employing an ICA-based approach, multi-source heterogeneous data is integrated, and through whitening processing, fault feature signal extraction, and a multi-level fault classification model, fault feature signal extraction and accurate classification are achieved.
By integrating multi-source heterogeneous data and intelligent feature extraction, real-time location and accurate classification of pneumatic actuator faults were achieved, improving the accuracy and efficiency of fault diagnosis.
Smart Images
Figure CN121144993A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The embodiment of the present application relates to the technical field of pneumatic actuator state monitoring, in particular to a pneumatic actuator fault diagnosis method based on ICA. BACKGROUND
[0002] In the field of industrial automation, as the core component of controlling the action of key devices such as valves and gates, the running state of the pneumatic actuator is directly related to the stability and safety of the production system. For a long time, the fault diagnosis of such devices mainly relies on traditional methods centered on sensor data analysis and artificial experience judgment, but these methods have certain defects in practical application.
[0003] From the perspective of sensor data analysis, the traditional scheme mainly relies on the collection and monitoring of a few key parameters such as pressure, flow and displacement. Due to the strong nonlinearity and time-varying characteristics of airflow movement in the pneumatic system, single or limited sensor data can only reflect local working conditions and is difficult to capture subtle features in the early stages of failure. In addition, the traditional method has obvious shortcomings in fault mode identification. The fault types of the pneumatic actuator are diverse, including mechanical jamming, gas source pollution, electromagnetic reversing valve failure, etc., and the performance characteristics of different faults overlap. Due to the lack of systematic fault feature library and intelligent analysis model, the traditional method is difficult to accurately locate and classify complex faults. SUMMARY
[0004] Based on the above situation of the prior art, the purpose of the embodiment of the present application is to provide a pneumatic actuator fault diagnosis method based on ICA, which can extract independent fault feature signals from complex multi-source heterogeneous data by integrating multi-source data processing, intelligent feature extraction and hierarchical classification model, and realize real-time positioning and identification of faults through dynamic analysis and accurate classification.
[0005] To achieve the above purpose, according to one aspect of the present application, a pneumatic actuator fault diagnosis method based on ICA is provided, comprising the steps of: collecting multi-source heterogeneous data of the pneumatic actuator; extracting fault feature signals from the multi-source heterogeneous data; locating the fault source based on the extracted fault feature signals and the fault dynamic contribution matrix; performing fault diagnosis on the pneumatic actuator based on the fault source positioning result and the preset multi-level fault classification model.
[0006] Further, the step of extracting fault feature signals from the multi-source heterogeneous data comprises the steps of: whitening the multi-source heterogeneous data to obtain whitened data; Based on the whitened data, an objective function is established to maximize non-Gaussianity, and the objective function includes a separation matrix; Solve the objective function to obtain the separation matrix; Based on the separation matrix, a fault feature signal matrix is obtained, which includes the fault components extracted from the fault feature signals.
[0007] Furthermore, the multi-source heterogeneous data is whitened to obtain whitened data, including the following steps: Obtain the covariance matrix of multi-source heterogeneous data; The covariance matrix is decomposed into eigenvalues to obtain the eigenvector matrix; Based on the eigenvector matrix and the results of eigenvalue decomposition, the whitening matrix is obtained; Based on the whitening matrix, whitening processing is performed on multi-source heterogeneous data to obtain whitened data.
[0008] Furthermore, based on the extracted fault feature signals and the fault dynamic contribution matrix, the fault source is located, including the following steps: Calculate based on the current status data of the pneumatic actuator Statistic and SPE squared prediction error; when When the statistic or SPE squared prediction error exceeds a preset threshold, calculate the contribution rate of the feature at the current time. The fault contribution rate matrix is obtained based on the contribution rates of each type of data calculated. The components corresponding to the high values in the fault contribution rate matrix are identified as fault sources.
[0009] Furthermore, the multi-source heterogeneous data includes the inlet pressure value of the pneumatic actuator, the outlet pressure value of the pneumatic actuator, the linear displacement of the valve stem, the rod speed of the valve stem, the cylinder temperature, the valve body temperature, the pipeline temperature, and the actuator end temperature.
[0010] Furthermore, the preset multi-level fault classification model is constructed based on the following steps: ICA feature extraction for multi-source heterogeneous data; A multi-level fault classification model is constructed, which includes a first-level SVM classifier and a second-level decision tree classifier. The first-layer SVM classifier takes the extracted ICA feature vector as input and outputs a probability vector for the first-level fault classification. The second-layer decision tree classifier takes the concatenated vector of the probability vector and the ICA feature vector as input and outputs a second-level fault classification label.
[0011] Furthermore, the extracted ICA feature vectors include kurtosis features, frequency domain features, gradient features, maximum contribution rate, and contribution entropy.
[0012] Furthermore, the first-level fault classification includes abnormal air pressure, mechanical failure, and seal failure; the second-level fault classification includes sudden drop in intake pressure, fluctuation in exhaust pressure, piston rod jamming, guide rail deformation, cylinder leakage, and valve body leakage.
[0013] Furthermore, based on the fault source location results and a preset multi-level fault classification model, fault diagnosis is performed on the pneumatic actuator, including the following steps: Collect multi-source heterogeneous data during the period of pneumatic actuator failure; ICA feature extraction is performed on the multi-source heterogeneous data during the fault period; The extracted ICA feature vectors are input into the multi-level fault classification model to obtain the probability vector of the first-level fault classification and the second-level fault classification label; The consistency of the fault source location result, the probability vector of the first-level fault classification, and the second-level fault classification label is judged. If the consistency criterion is met, the fault diagnosis result is output.
[0014] Furthermore, a consistency judgment is made on the fault source location result, the probability vector of the primary fault classification, and the secondary fault classification label, including the following steps: Based on the faulty components in the fault source location results, map them to their respective fault categories; Determine whether the probability corresponding to the fault category is the highest in the probability vector of the first-level fault classification and exceeds the first threshold. Determine whether the faulty component corresponding to the secondary fault classification label is among the faulty components in the fault source location result; The fault diagnosis meets the consistency criterion when it simultaneously satisfies both of the above judgment conditions.
[0015] In summary, this invention provides a method for fault diagnosis of pneumatic actuators based on ICA, comprising the steps of: collecting multi-source heterogeneous data of the pneumatic actuator; extracting fault feature signals from the multi-source heterogeneous data; locating the fault source based on the extracted fault feature signals and a fault dynamic contribution matrix; and diagnosing the pneumatic actuator based on the fault source location result and a preset multi-level fault classification model. The technical solution provided by this invention, by integrating multi-source heterogeneous data, intelligent feature extraction, and a hierarchical classification model, can extract independent fault feature signals from complex multi-source heterogeneous data. Combined with the fault dynamic contribution matrix and the multi-level fault classification model, it diagnoses the pneumatic actuator. The fault dynamic contribution matrix quantifies the parameter influence weights, while the multi-level classification model achieves accurate fault classification from macro to micro levels. Through data-driven intelligent analysis, it reduces manual intervention and significantly improves the accuracy and efficiency of pneumatic actuator fault diagnosis. Attached Figure Description
[0016] Figure 1 This is a flowchart of the ICA-based pneumatic actuator fault diagnosis method provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of the pneumatic actuator provided in an embodiment of the present invention.
[0017] Explanation of reference numerals in the attached figures: 21-Positioner (Actuator); 22-Inlet; 23-Exhaust; 24-Cylinder; 241-Stroke pointer; 242-Stroke scale; 243-Upper chamber of diaphragm; 244-Diaphragm; 245-Lower chamber of diaphragm; 246-Spring; 247-Push rod; 248-Valve stem; 25-Valve body; 251-Valve core; 252-Valve seat. Detailed Implementation
[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments and the accompanying drawings. It should be understood that these descriptions are merely exemplary and not intended to limit the scope of the invention. Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concept of the invention.
[0019] It should be noted that, unless otherwise defined, the technical or scientific terms used in one or more embodiments of the present invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in one or more embodiments of the present invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the element or object listed following the word and its equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect.
[0020] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings. An embodiment of the present invention provides a method for diagnosing faults in pneumatic actuators based on ICA (Inductively Coupled Aspect). Figure 1 The flowchart of the ICA-based pneumatic actuator fault diagnosis method provided by an embodiment of the present invention is shown below. Figure 1 As shown, the method includes the following steps: S202. Acquire multi-source heterogeneous data from pneumatic actuators. This can be achieved by installing sensors at key locations on the pneumatic actuators. Figure 2 The diagram shows the structure of a pneumatic actuator, which mainly consists of two parts: a cylinder 24 and a valve body 25. Figure 2 Pressure sensors (e.g., ±0.5 kPa accuracy at the inlet and ±1 kPa accuracy at the outlet) can be installed at the inlet end 22 and outlet end 23 of the pneumatic actuator to collect the inlet pressure value. and exhaust pressure value A displacement sensor is installed at position 248 of the valve stem to collect the linear displacement of the valve stem 248. A speed sensor (which can be calculated based on displacement difference) can also be set to collect the rod movement speed of valve stem 248. ; in cylinder 24, valve body 25, pipeline ( Figure 2 Temperature sensors are installed at the actuator 21 and the actuator 21 respectively to collect the temperatures of the above-mentioned parts, including the cylinder temperature. Valve body temperature Pipeline temperature and execution end temperature Based on the collected multi-source heterogeneous data, the original data matrix of the multi-source heterogeneous data is constructed, which can be represented as: .
[0021] According to certain optional embodiments, the data can be preprocessed, including wavelet denoising and normalization, to obtain a normalized matrix of the multi-source heterogeneous data. Standardized matrix It can be represented as:
[0022] Here, the superscript indicates the index of the sampling time. The preprocessing step can be performed using existing methods, and no specific limitations are made here.
[0023] S204. Extract fault feature signals from multi-source heterogeneous data. In this embodiment of the invention, fault feature signal extraction can be performed using the following steps: S2041. Obtain the covariance matrix of multi-source heterogeneous data. The covariance matrix can reflect the correlation between multi-source heterogeneous data.
[0024] S2042. Perform eigenvalue decomposition on the covariance matrix to obtain the eigenvector matrix. eigenvector matrix The column vectors are unit orthogonal vectors:
[0025] in, Represents the covariance matrix.
[0026] S2043. Based on the results of eigenvalue decomposition and the number of principal components, a whitening matrix is constructed, which is represented as follows:
[0027] in, Indicates the number of principal components. , Indicates the preceding There are eigenvectors. The number of principal components can be determined by calculating the variance contribution rate of each eigenvalue, and then calculating the cumulative contribution rate of the first k principal components, finding the smallest integer that satisfies the threshold condition. This allows us to determine the number of principal components.
[0028] S2044. Based on the whitening matrix, whitening processing is performed on multi-source heterogeneous data to obtain whitened data, which is represented as follows:
[0029] The dimension of the whitened data matrix Z is q*m, where m represents the number of samples.
[0030] S2045. Based on whitened data, establish an objective function that maximizes non-Gaussianity. This objective function includes a separation matrix. By solving the objective function, the separation matrix can be obtained. The separation matrix can be solved using the ICA iterative optimization method. Through the separation matrix, multi-source heterogeneous data can be decomposed into statistically independent components, such as fault characteristic signals (high-frequency impacts, abnormal fluctuations, etc.), normal operating condition signals (low-frequency background noise), and interference signals (sensor noise). The objective function is expressed as:
[0031] in, Represents the row vectors of the separation matrix. The row vector representing the whitened data. Representing a nonlinear Gaussian function: , Separated independent component matrix That is, the separation matrix can be represented as:
[0032] Among them, the independent component matrix The dimension is r*m, and each row represents the value of an independent component (source signal) at all sample points. This represents the ICA separation matrix, with dimensions r*q, where r represents the number of independent components (typically r=q). The independent component matrix... It contains all the independent components.
[0033] S2046. Based on the separation matrix, a fault feature signal matrix is obtained. This matrix contains the extracted fault components and has a dimension of m*r, where m represents the number of sample points in the time series and r represents the number of fault components. The fault feature signal matrix can be output by determining whether each independent component vector in the separation matrix satisfies a fault criterion. The fault criterion may include, for example, a kurtosis criterion and a frequency domain criterion. The kurtosis criterion is expressed as:
[0034] The frequency domain criterion is expressed as:
[0035] Where Kurtosis() represents the kurtosis calculation function, This represents the i-th fault component. This represents the kurtosis threshold, which is taken in this embodiment of the invention. ; PSD() represents the power spectral density function, Indicates the sampling frequency of multi-source heterogeneous data. This represents the frequency domain determination threshold, which is taken in this embodiment of the invention. The independent component vectors that satisfy the kurtosis criterion and / or frequency domain criterion are output to form a fault feature matrix. In this embodiment of the invention, Independent Component Analysis (ICA) is used for fault feature extraction. This method can effectively separate statistically independent fault feature signals from complex multi-source heterogeneous mixed data. By maximizing non-Gaussianity to construct the objective function and solve the separation matrix, it has good adaptability to the strongly nonlinear and time-varying characteristics of airflow motion inside aerodynamic systems, improves the separation accuracy of fault features, and helps to accurately identify early faults.
[0036] S206. Based on the extracted fault feature signals and the fault dynamic contribution matrix, locate the fault source.
[0037] S2061. Establish a fault dynamic contribution matrix. In this embodiment of the invention, based on... A fault dynamic contribution matrix is constructed using statistics and SPE squared prediction error. The statistics are used to measure the deviation between the current state of the pneumatic actuator and the state predicted by the principal component model, while the SPE (Squared Prediction Error) reflects the prediction residuals of the principal component model. The principal component model can be expressed as:
[0038] The statistic and the SPE squared prediction error are expressed as follows:
[0039]
[0040] in, Represents eigenvalues. Represents the i-th principal component. This represents the squared residual value of the data collected by the j-th sensor at the k-th sampling time.
[0041] S2062, Calculate based on the current status data of the pneumatic actuator. Statistic and SPE squared prediction error, when When the statistic or SPE squared prediction error exceeds a preset threshold, calculate the contribution rate of the feature at the current time:
[0042] Contribution rate This represents the percentage contribution of the data collected by the j-th sensor to the total anomaly.
[0043] S2063. Based on the calculated contribution rates of various data types, a fault contribution rate matrix is obtained. Components corresponding to data with higher values can be identified as fault sources. This selection can be based on the amount of multi-source heterogeneous data and practical engineering experience. In this embodiment of the invention, components corresponding to the top 3 values are identified as fault sources. The fault contribution rate matrix can be represented as:
[0044] In the fault contribution rate matrix, the subscript of the variable represents the index of the sensor data. In this embodiment of the invention, a total of 8 variable data are collected, and the superscript represents the index of the sampling time.
[0045] S208. Based on the fault source location results and a pre-defined multi-level fault classification model, perform fault diagnosis on the pneumatic actuator. The pre-defined multi-level fault classification model can be constructed through the following steps: S2081. Perform ICA feature extraction on multi-source heterogeneous data to obtain kurtosis features. Frequency domain characteristics Gradient features Maximum contribution rate and contribution entropy :
[0046]
[0047]
[0048]
[0049]
[0050] Where r represents the number of fault components, Let i represent the i-th fault component, Kurtosis() represent the kurtosis calculation function, PSD() represent the power spectral density function, and max() represent the peak value of the spectrum. This represents the linear displacement of the valve stem at time k. This represents the valve stem velocity at time k. This represents the difference operator, i.e., the sampling difference between consecutive time intervals. This represents the average contribution rate of component j. The feature vector extracted based on ICA in this step contains rich information such as kurtosis and frequency domain features, which can provide high-quality input for subsequent multi-level fault classification models to achieve more accurate fault classification.
[0051] S2082. Construct a multi-level fault classification model, which includes a first-layer SVM classifier and a second-layer decision tree classifier. The input to the SVM classifier is the feature vector v obtained in step S2081, where the elements of v are the degree features. Frequency domain characteristics Gradient features Maximum contribution rate and contribution entropy The output is a probability vector P for the first-level fault classification. The probability vector P and the feature vector v are concatenated as the output of the decision tree classifier, which outputs the second-level fault classification label. Historical data is used to train the multi-level fault classification model, resulting in a trained multi-level fault classification model. In this embodiment, the first-level fault classification includes pneumatic system anomalies, mechanical structure faults, and sealing failures; the second-level fault classification includes intake pressure drop and exhaust pressure fluctuations (corresponding to pneumatic system anomalies), piston rod jamming and guide rail deformation (corresponding to mechanical structure faults), and cylinder leakage and valve body leakage (corresponding to sealing failures).
[0052] Based on the fault source location results and a pre-set multi-level fault classification model, fault diagnosis of pneumatic actuators is performed, following these steps: S2083. Collect multi-source heterogeneous data during the period of pneumatic actuator failure. During the period of failure, collect the aforementioned multi-source heterogeneous data of the pneumatic actuator.
[0053] S2084. Perform ICA feature extraction on the multi-source heterogeneous data during the fault period. The ICA feature extraction method is the same as that involved in step S2081 above.
[0054] S2085. Input the extracted ICA feature vector into the multi-level fault classification model to obtain the probability vector of the first-level fault classification and the second-level fault classification label.
[0055] S2086. Perform a consistency check on the fault source location result, the probability vector of the primary fault classification, and the secondary fault classification label. If the consistency criteria are met, output the fault diagnosis result; if the consistency criteria are not met, the fault diagnosis can be re-performed through confidence arbitration. The consistency check on the fault source location result, the probability vector of the primary fault classification, and the secondary fault classification label can be performed according to the following steps: S20861. Based on the faulty component in the fault source location result, map it to the corresponding fault category.
[0056] S20862. Determine whether the probability corresponding to the fault category in the probability vector of the first-level fault classification is the highest and exceeds the first threshold. The first threshold is usually greater than 50%.
[0057] S20863. Determine whether the faulty component corresponding to the secondary fault classification label is among the faulty components in the above fault source location results.
[0058] Mapping and judgment can be performed according to a preset mapping relationship table. In this embodiment of the invention, the mapping relationship table is shown in Table 1 below, and Table 2 shows the mapping relationship between multi-source heterogeneous data and fault source components.
[0059] Table 1
[0060] Table 2
[0061] S20864. If the consistency judgment result satisfies both judgment conditions S20862 and S20863 above, then the consistency criterion is met, and the fault diagnosis result is output.
[0062] During the confidence arbitration process, the confidence scores of the fault source localization result and the prediction result of the multi-level fault classification model can be calculated separately. These two confidence scores are compared, and the result with the higher confidence score is selected as the primary basis. The other result is then readjusted. The confidence score of the fault source localization result can be measured using the contribution rate of the main faulty component (e.g., the value of the highest contribution rate). The confidence score of the multi-level fault classification model prediction result can be measured using the highest probability value of the first-level fault classification and the confidence score of the second-level fault classification. Existing calculation methods can be used to calculate the confidence score; this embodiment does not impose specific limitations.
[0063] In summary, this invention relates to a fault diagnosis method for pneumatic actuators based on ICA, comprising the following steps: collecting multi-source heterogeneous data of the pneumatic actuator; extracting fault feature signals from the multi-source heterogeneous data; locating the fault source based on the extracted fault feature signals and a fault dynamic contribution matrix; and diagnosing the fault of the pneumatic actuator based on the fault source location result and a preset multi-level fault classification model. The technical solution provided by this invention, by integrating multi-source heterogeneous data, intelligent feature extraction, and a hierarchical classification model, can extract independent fault feature signals from complex multi-source heterogeneous data. Combined with the fault dynamic contribution matrix and the multi-level fault classification model, it diagnoses the fault of the pneumatic actuator. The fault dynamic contribution matrix quantifies the influence weight of parameters, while the multi-level classification model achieves accurate fault classification from macro to micro levels. Through data-driven intelligent analysis, it reduces manual intervention and significantly improves the accuracy and efficiency of pneumatic actuator fault diagnosis.
[0064] It should be understood that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of the invention (including the claims) is limited to these examples. Within the framework of this invention, technical features of the above embodiments or different embodiments can also be combined, steps can be implemented in any order, and many other variations exist regarding different aspects of one or more embodiments of the invention as described above; for the sake of brevity, they are not provided in the details. The specific embodiments described above are merely illustrative or explanatory of the principles of the invention and do not constitute a limitation thereof. Therefore, any modifications, equivalent substitutions, improvements, etc., made without departing from the spirit and scope of the invention should be included within the protection scope of the invention. Furthermore, the appended claims are intended to cover all variations and modifications falling within the scope and boundaries of the appended claims, or equivalent forms of such scope and boundaries.
Claims
1. A fault diagnosis method for pneumatic actuators based on ICA, characterized in that, Including the following steps: Collect multi-source heterogeneous data from pneumatic actuators; Fault feature signals are extracted from the multi-source heterogeneous data; Based on the extracted fault feature signals, the fault source is located using the fault dynamic contribution matrix. The pneumatic actuator is diagnosed based on the fault source location results and a preset multi-level fault classification model.
2. The method according to claim 1, characterized in that, The process of extracting fault feature signals from the multi-source heterogeneous data includes the following steps: The multi-source heterogeneous data is whitened to obtain whitened data; Based on the whitened data, an objective function is established to maximize non-Gaussianity, and the objective function includes a separation matrix; Solve the objective function to obtain the separation matrix; Based on the separation matrix, a fault feature signal matrix is obtained, which includes the fault components extracted from the fault feature signals.
3. The method according to claim 2, characterized in that, The process of whitening the multi-source heterogeneous data to obtain whitened data includes the following steps: Obtain the covariance matrix of multi-source heterogeneous data; The covariance matrix is decomposed into eigenvalues to obtain the eigenvector matrix; Based on the eigenvector matrix and the results of eigenvalue decomposition, the whitening matrix is obtained; Based on the whitening matrix, whitening processing is performed on multi-source heterogeneous data to obtain whitened data.
4. The method according to claim 3, characterized in that, Based on the extracted fault feature signals and the fault dynamic contribution matrix, the fault source is located, including the following steps: Calculate based on the current status data of the pneumatic actuator Statistic and SPE squared prediction error; when When the statistic or SPE squared prediction error exceeds a preset threshold, calculate the contribution rate of the feature at the current time. The fault contribution rate matrix is obtained based on the contribution rates of each type of data calculated. The components corresponding to the high values in the fault contribution rate matrix are identified as fault sources.
5. The method according to any one of claims 1-4, characterized in that, The multi-source heterogeneous data includes the inlet pressure value of the pneumatic actuator, the outlet pressure value of the pneumatic actuator, the linear displacement of the valve stem, the rod speed of the valve stem, the cylinder temperature, the valve body temperature, the pipeline temperature, and the actuator end temperature.
6. The method according to claim 1, characterized in that, The preset multi-level fault classification model is constructed based on the following steps: ICA feature extraction for multi-source heterogeneous data; A multi-level fault classification model is constructed, which includes a first-level SVM classifier and a second-level decision tree classifier. The first-layer SVM classifier takes the extracted ICA feature vector as input and outputs a probability vector for the first-level fault classification. The second-layer decision tree classifier takes the concatenated vector of the probability vector and the ICA feature vector as input and outputs a second-level fault classification label.
7. The method according to claim 6, characterized in that, The extracted ICA feature vectors include kurtosis, frequency domain features, gradient features, maximum contribution rate, and contribution entropy.
8. The method according to claim 7, characterized in that, The first-level fault classification includes abnormal air pressure, mechanical failure, and seal failure; the second-level fault classification includes sudden drop in intake pressure, fluctuation in exhaust pressure, piston rod jamming, guide rail deformation, cylinder leakage, and valve body leakage.
9. The method according to claim 8, characterized in that, Based on the fault source location results and a preset multi-level fault classification model, fault diagnosis is performed on the pneumatic actuator, including the following steps: Collect multi-source heterogeneous data during the period of pneumatic actuator failure; ICA feature extraction is performed on the multi-source heterogeneous data during the fault period; The extracted ICA feature vectors are input into the multi-level fault classification model to obtain the probability vector of the first-level fault classification and the second-level fault classification label; The consistency of the fault source location result, the probability vector of the first-level fault classification, and the second-level fault classification label is judged. If the consistency criterion is met, the fault diagnosis result is output.
10. The method according to claim 9, characterized in that, The consistency judgment of the fault source location result, the probability vector of the primary fault classification, and the secondary fault classification label includes the following steps: Based on the faulty components in the fault source location results, map them to their respective fault categories; Determine whether the probability corresponding to the fault category is the highest in the probability vector of the first-level fault classification and exceeds the first threshold. Determine whether the faulty component corresponding to the secondary fault classification label is among the faulty components in the fault source location result; The fault diagnosis meets the consistency criterion when it simultaneously satisfies both of the above judgment conditions.