A method, device, medium and product for fault diagnosis of electro-hydraulic proportional servo valve

Abstract

This application discloses a method, device, medium, and product for diagnosing electro-hydraulic proportional servo valve faults, relating to the technical field of electromechanical hydraulic system fault diagnosis. The method comprises: obtaining the original signal and decision weight of each sensor in the electro-hydraulic proportional servo valve to be diagnosed; determining the fault type diagnosed by the corresponding sensor based on the original signal of each sensor in the electro-hydraulic proportional servo valve to be diagnosed and the corresponding fault classification sub-model; each sensor corresponds to a fault classification sub-model, and each fault classification sub-model is obtained by training a support vector machine optimized using the Grasshopper Optimization algorithm using the training set of the corresponding sensor. The fault type of the electro-hydraulic proportional servo valve to be diagnosed is determined based on the fault type diagnosed by each sensor and the decision weight of the corresponding sensor. This application improves the diagnostic accuracy of the electro-hydraulic proportional servo valve.

Description

Technical Field

[0001] The present application relates to the technical field of electromechanical hydraulic system fault diagnosis, and in particular to a method, device, medium and product for diagnosing a fault of an electro-hydraulic proportional servo valve. Background Art

[0002] High-performance electro-hydraulic proportional servo valves are core control components in the hydraulic systems of high-end equipment such as intelligent engineering machinery, manufacturing equipment, and robotics. Rapid and accurate fault diagnosis of proportional servo valves is essential for ensuring the safe operation of these hydraulic systems. When an electro-hydraulic proportional servo valve fails, each of its internal sensor signals contains corresponding fault information. Therefore, analyzing the sensitive sensor signals corresponding to each fault and identifying the relationship between the fault characteristics and the fault mode is an effective fault diagnosis method.

[0003] Traditional hydraulic valve fault diagnosis methods only target single signals such as current and pressure signals. However, proportional servo valve faults can manifest in multiple forms, and when key information is lost in the signal collected by a single sensor, the fault diagnosis effect will be greatly affected.

[0004] Therefore, there is an urgent need for a fault diagnosis method for electro-hydraulic proportional servo valve with higher diagnostic accuracy. Summary of the Invention

[0005] The purpose of this application is to provide a method, device, medium and product for diagnosing faults of an electro-hydraulic proportional servo valve, thereby improving the diagnostic accuracy of the electro-hydraulic proportional servo valve.

[0006] To achieve the above objectives, this application provides the following solutions:

[0007] In a first aspect, the present application provides a method for diagnosing a fault of an electro-hydraulic proportional servo valve, comprising:

[0008] Obtaining the original signals and decision weights of each sensor in the electro-hydraulic proportional servo valve to be diagnosed; the sensors include: a coil current sensor, an LVDT position sensor, a spool valve A port pressure sensor, and a spool valve B port pressure sensor;

[0009] Based on the original signal of each sensor in the electro-hydraulic proportional servo valve to be diagnosed and the corresponding fault classification sub-model, the fault type obtained by diagnosis of the corresponding sensor is determined; each sensor corresponds to a fault classification sub-model, and each fault classification sub-model is obtained by training a support vector machine optimized by the Grasshopper optimization algorithm using the training set of the corresponding sensor. The fault types are: normal valve, LVDT offset, LVDT gain, valve core stuck, valve core wear, coil turn short circuit, or multiple faults;

[0010] The fault type of the electro-hydraulic proportional servo valve to be diagnosed is determined based on the fault type obtained through diagnosis of each sensor and the decision weight of the corresponding sensor.

[0011] Optionally, the training process of each fault classification sub-model includes:

[0012] The grasshopper optimization algorithm is used to optimize the support vector machine to obtain the optimized support vector machine;

[0013] Acquire a raw data set of the training electro-hydraulic proportional servo valve; the raw data set includes: a true value of the fault type of the training electro-hydraulic proportional servo valve and raw signals of each sensor in the training electro-hydraulic proportional servo valve;

[0014] The wavelet soft threshold noise reduction method is used to reduce the noise of the original signal of each sensor in the training electro-hydraulic proportional servo valve, and the denoised signal of the corresponding sensor is obtained.

[0015] Extract fault features in the time domain, frequency domain, and time-frequency domain for the denoised signal of each sensor to obtain an initial feature set for the corresponding sensor; the initial feature set includes a time domain feature parameter set, a frequency domain feature parameter set, and a time-frequency domain fault feature parameter set;

[0016] The mRMR feature selection algorithm is used to screen the initial feature set of each sensor to obtain the target feature set of the corresponding sensor;

[0017] Based on the target feature set of each sensor and the actual value of the fault type of the training electro-hydraulic proportional servo valve, a training set of the corresponding sensor is constructed;

[0018] The optimized support vector machine is trained using the training set of each sensor to obtain the fault classification sub-model corresponding to the corresponding sensor.

[0019] Optionally, determining the fault type obtained through diagnosis of the corresponding sensor based on the original signal of each sensor in the electro-hydraulic proportional servo valve to be diagnosed and the corresponding fault classification sub-model includes:

[0020] Determine any sensor in the electro-hydraulic proportional servo valve to be diagnosed as the current sensor;

[0021] The wavelet soft threshold noise reduction method is used to reduce the noise of the original signal of the current sensor to obtain the denoised signal of the current sensor;

[0022] For the denoised signal of the current sensor, fault features are extracted in the time domain, frequency domain, and time-frequency domain respectively to obtain the initial feature set of the current sensor, thereby determining the target feature set of the current sensor;

[0023] The target feature set of the current sensor is input into the fault classification sub-model corresponding to the current sensor to obtain the fault type diagnosed by the current sensor.

[0024] Optionally, the optimized support vector machine is trained using the training set of each sensor to obtain a fault classification sub-model corresponding to the corresponding sensor, including:

[0025] Taking the target feature set of each sensor as input and the true value of the fault type of the training electro-hydraulic proportional servo valve as output, the optimized support vector machine is trained to obtain the fault classification sub-model corresponding to the corresponding sensor.

[0026] Optionally, a wavelet soft threshold noise reduction method is used to reduce the noise of the original signal to obtain a denoised signal, including:

[0027] Perform wavelet decomposition on the original signal to obtain the wavelet transform of the original signal;

[0028] Performing threshold processing on the wavelet coefficients in the wavelet transform of the original signal to obtain the wavelet coefficients after threshold processing;

[0029] The signal is reconstructed based on the wavelet coefficients after threshold processing to obtain the denoised signal.

[0030] Optionally, the time domain characteristic parameter set includes: maximum value, minimum value, peak-to-peak value, average value, effective value, variance, root mean square amplitude, kurtosis, skewness, shape factor and crest factor;

[0031] The frequency domain characteristic parameter set includes: domain amplitude mean, centroid frequency, root mean square frequency, mean square frequency, frequency variance, frequency domain amplitude variance, frequency domain amplitude skewness and frequency standard deviation;

[0032] The time-frequency domain fault characteristic parameter set includes: energy values ​​of 8 nodes extracted based on a wavelet packet decomposition method.

[0033] Optionally, an mRMR feature selection algorithm is used to perform feature screening on the initial feature set of each sensor to obtain a target feature set of the corresponding sensor, including:

[0034] Determine any sensor as a target sensor;

[0035] Calculate the mutual information between each feature in the initial feature set of the target sensor and the true value of the fault type of the training electro-hydraulic proportional servo valve;

[0036] The features in the initial feature set are screened based on the mutual information between each feature and the true value of the fault type of the training electro-hydraulic proportional servo valve to obtain a filtered feature set of the target sensor;

[0037] Calculate the mutual information between any two features in the feature set after a screening;

[0038] The feature set after the primary screening is screened based on the mutual information between any two features to obtain the feature set after the secondary screening;

[0039] Calculate the mRMR value of each feature in the feature set after secondary screening to determine multiple candidate features;

[0040] An incremental search strategy is used to determine the target feature set of the target sensor based on the initial feature set of the target sensor and multiple candidate features.

[0041] In a second aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement any of the above-described electro-hydraulic proportional servo valve fault diagnosis methods.

[0042] In a third aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-mentioned electro-hydraulic proportional servo valve fault diagnosis methods.

[0043] In a fourth aspect, the present application provides a computer program product, comprising a computer program, which, when executed by a processor, implements any of the above-mentioned electro-hydraulic proportional servo valve fault diagnosis methods.

[0044] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0045] The present application discloses a fault diagnosis method, device, medium and product for an electro-hydraulic proportional servo valve, which utilizes signals from multiple sensors and corresponding fault classification sub-models to perform fault diagnosis on the electro-hydraulic proportional servo valve. On the one hand, the data from multiple sensors can more comprehensively reflect the actual working conditions of the electro-hydraulic proportional servo valve, making the fault characteristics more obvious and easier to identify; on the other hand, even if a sensor fails or data is lost, the data from other sensors can still provide the necessary information, thereby avoiding stagnation of the entire monitoring system due to problems with a single data source, and improving the diagnostic accuracy of the electro-hydraulic proportional servo valve. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0047] Figure 1 A schematic flow chart of a method for diagnosing a fault in an electro-hydraulic proportional servo valve according to an embodiment of the present application;

[0048] Figure 2 This is the flow chart of the multi-sensor information fusion fault diagnosis algorithm based on enhanced weighted voting;

[0049] Figure 3 is the original signal curve of the sensor;

[0050] Figure 4 It is the denoised signal curve of the sensor;

[0051] Figure 5 Optimizing support vector machine fault diagnosis algorithm diagram for GOA;

[0052] Figure 6 This is the GOA optimization support vector machine training curve corresponding to the pressure sensor at port A of the slide valve;

[0053] Figure 7 GOA optimization support vector machine training curve corresponding to the pressure sensor at port B of the spool valve;

[0054] Figure 8 GOA optimization support vector machine training curve corresponding to LVDT position sensor;

[0055] Figure 9 GOA optimization support vector machine training curve corresponding to the coil current sensor;

[0056] Figure 10 This is the confusion matrix diagram for the fault diagnosis of the pressure sensor at port A of the spool valve;

[0057] Figure 11 This is the confusion matrix diagram for the fault diagnosis of the pressure sensor at port B of the spool valve;

[0058] Figure 12 Confusion matrix diagram for LVDT position sensor fault diagnosis;

[0059] Figure 13 Confusion matrix diagram for coil current sensor fault diagnosis;

[0060] Figure 14 Confusion matrix diagram for multi-sensor information fusion fault diagnosis;

[0061] Figure 15 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0062] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0063] The purpose of this application is to provide a method, device, medium and product for diagnosing faults of an electro-hydraulic proportional servo valve, aiming to improve the diagnostic accuracy of the electro-hydraulic proportional servo valve.

[0064] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0065] In an exemplary embodiment, Figure 1 and Figure 2 As shown, the electro-hydraulic proportional servo valve fault diagnosis method in this embodiment includes:

[0066] Step 1: Obtain the original signal and decision weight of each sensor in the electro-hydraulic proportional servo valve to be diagnosed; the sensors include: coil current sensor, LVDT position sensor, spool valve A port pressure sensor and spool valve B port pressure sensor.

[0067] Step 2: Based on the original signal of each sensor in the electro-hydraulic proportional servo valve to be diagnosed and the corresponding fault classification sub-model, determine the fault type obtained by the corresponding sensor diagnosis.

[0068] Among them, each sensor corresponds to a fault classification sub-model. Each fault classification sub-model is obtained by training a support vector machine optimized by the Grasshopper optimization algorithm using the training set of the corresponding sensor. The fault types are: normal valve, LVDT bias, LVDT gain, valve core stuck, valve core wear, coil inter-turn short circuit or multiple faults.

[0069] Specifically, as shown in Table 1, the identification codes of various faults and related fault parameters are listed in detail.

[0070] Table 1 Identification codes of various faults and related fault parameters

[0071] Label Fault type Training set\test set Fault parameters 1 Valve normal 4725\2025 - 2 LVDT bias 4725\2025 The displacement sensor output percentage is 120% 3 LVDT gain 4725\2025 Displacement sensor constant bias +2.0V 4 Valve core stuck 4725\2025 30% stuck 5 Valve core wear 4725\2025 B valve port bears the high pressure end wear 6 Coil inter-turn short circuit 4725\2025 Electromagnet current 80% 7 Multiple faults 4725\2025 Valve core wear + stuck 30% + solenoid current 80%

[0072] In practice, 688,500 data points were collected for each fault type, serving as the sample set for this embodiment. The hydraulic valve's operating mode was set to a sinusoidal waveform, with a duty cycle of 1 second. During the model training phase, one sample was extracted from the data for every five hydraulic valve duty cycles, with each sample containing 500 data points. Each fault state resulted in 6,750 samples. Of these samples, 70% were used to construct the training set, while the remaining 30% were used to form the test set.

[0073] As an optional implementation, the training process of each fault classification sub-model includes:

[0074] Step 201: Optimize the support vector machine using the grasshopper optimization algorithm to obtain an optimized support vector machine.

[0075] Step 202: Acquire a raw data set of the training electro-hydraulic proportional servo valve; the raw data set includes: the true value of the fault type of the training electro-hydraulic proportional servo valve and the raw signals of each sensor in the training electro-hydraulic proportional servo valve.

[0076] Step 203: Using a wavelet soft threshold noise reduction method, perform noise reduction processing on the original signal of each sensor in the training electro-hydraulic proportional servo valve to obtain a denoised signal of the corresponding sensor.

[0077] Step 204: Perform fault feature extraction in the time domain, frequency domain, and time-frequency domain on the denoised signal of each sensor to obtain an initial feature set of the corresponding sensor; the initial feature set includes: a time domain feature parameter set, a frequency domain feature parameter set, and a time-frequency domain fault feature parameter set.

[0078] Step 205: Using the mRMR feature selection algorithm, perform feature screening on the initial feature set of each sensor to obtain the target feature set of the corresponding sensor.

[0079] Step 206: Construct a training set for the corresponding sensor based on the target feature set of each sensor and the true value of the fault type of the training electro-hydraulic proportional servo valve.

[0080] Step 207: Use the training set of each sensor to train the optimized support vector machine to obtain a fault classification sub-model corresponding to the corresponding sensor.

[0081] As an optional implementation, step 206 includes:

[0082] Taking the target feature set of each sensor as input and the true value of the fault type of the training electro-hydraulic proportional servo valve as output, the optimized support vector machine is trained to obtain the fault classification sub-model corresponding to the corresponding sensor.

[0083] Specifically, such as Figure 5 As shown in FIG, the Grasshopper Optimization Algorithm (GOA) is used to optimize the key parameters of the Support Vector Machine (SVM), including the penalty coefficient C and the RBF kernel function parameter σ, including:

[0084] (1) Initialization: Initialize the grasshopper population size, the maximum number of iterations, and the upper and lower limits of the space. Use grasshopper individuals to represent the key parameters of the SVM and initialize the search target individual [C, σ].

[0085] (2) Fitness calculation: Calculate the fitness of the SVM parameter combination represented by each grasshopper, and use the diagnostic accuracy of SVM as the fitness evaluation criterion:

[0086]

[0087] Among them, n1 is the number of samples that are accurately identified; N1 is the total number of samples identified.

[0088] (3) Grasshopper behavior simulation: GOA simulates the exploration (global search) and exploitation (local search) behaviors of grasshoppers. The new position of each grasshopper is determined by a combination of its current position, social influence, gravity, and wind. The update formula is as follows:

[0089]

[0090] in, is the position of grasshopper i at time step t+1; is the position of grasshopper i at time step t; N2 is the total number of grasshoppers in the population; S(d ij ) is the social force function, which represents the interaction force between grasshopper i and grasshopper j; is the position of grasshopper j at time step t; d ij is the distance between grasshopper i and grasshopper j; G i is the gravity acting on grasshopper i; A i is the wind force on grasshopper i; D is the attraction intensity; L is the attraction length scale.

[0091] (4) Parameter update: Repeat the fitness calculation and grasshopper behavior simulation, update the individual position, re-evaluate the fitness at the new position, and update the position of the optimal individual until the preset termination condition is met.

[0092] (5) Single sensor information fault diagnosis based on optimized SVM model: According to the fitness evaluation, the SVM parameters represented by the grasshopper with the highest fitness in the population are selected as the optimal solution.

[0093] In the training set, the GOA optimization algorithm is used to optimize the support vector machine penalty coefficient C and RBF kernel function parameters. The parameter settings are as follows: the initial population size is 20, the maximum number of iterations is 100, and the fitness function is selected as the average of five cross-checks of the SVM. By training the GOA-SVM model in the training set, the iterative results of the support vector machine model of each sensor are obtained, as shown in Figure 2. Figure 6-Figure 9 shown.

[0094] The optimized support vector machine model is used to diagnose the faults of the validation set data, and the confusion matrix of the four sensor diagnosis results is as follows: Figure 10-13 By summarizing the fault diagnosis results of the confusion matrix, we can obtain the diagnostic accuracy of each sensor for different types of faults, as shown in Table 2.

[0095] Table 2 Diagnostic accuracy of each sensor for different faults

[0096]

[0097] As an optional implementation, step 2 includes:

[0098] Step 211: Determine any sensor in the electro-hydraulic proportional servo valve to be diagnosed as the current sensor.

[0099] Step 212: Using a wavelet soft threshold noise reduction method, perform noise reduction processing on the original signal of the current sensor to obtain a denoised signal of the current sensor.

[0100] Step 213: Perform fault feature extraction in the time domain, frequency domain, and time-frequency domain on the denoised signal of the current sensor to obtain an initial feature set of the current sensor, thereby determining a target feature set of the current sensor.

[0101] Step 214: Input the target feature set of the current sensor into the fault classification sub-model corresponding to the current sensor to obtain the fault type diagnosed by the current sensor.

[0102] As an optional implementation, in step 203 and step 212, the process of using the wavelet soft threshold noise reduction method to reduce the noise of the original signal to obtain the denoised signal includes:

[0103] Perform wavelet decomposition on the original signal to obtain the wavelet transform of the original signal.

[0104] Threshold processing is performed on the wavelet coefficients in the wavelet transform of the original signal to obtain the wavelet coefficients after threshold processing.

[0105] The signal is reconstructed based on the wavelet coefficients after threshold processing to obtain the denoised signal.

[0106] Specifically, the sensor signal is denoised using a soft wavelet threshold denoising method. The specific process includes:

[0107] (1) Wavelet decomposition: Perform wavelet decomposition on the original signal x(t). Wavelet decomposition decomposes the original signal into components at different scales. These components contain the low-frequency information and high-frequency information of the original signal. This process can be expressed by the following formula:

[0108] X=T+D.

[0109] Among them, X is the wavelet transform of the original signal; T is the low-frequency approximation part; D is the high-frequency detail part.

[0110] (2) Threshold processing: By processing the original signal through wavelet decomposition, it can be found that the wavelet coefficients of the effective signal are usually small in number but large in value, while the wavelet coefficients of the noise signal are the opposite, that is, large in number but small in value. Based on this observation, by setting an appropriate threshold, the coefficients obtained by decomposition are screened and processed, thereby achieving signal noise reduction.

[0111] Choosing a suitable threshold λ is the key to the effectiveness of noise reduction. The threshold can be determined based on the characteristics of the signal and the noise level. The wavelet threshold used here is:

[0112]

[0113] Where N is the length of the original signal.

[0114] For the wavelet coefficient w a Apply the soft threshold function S(w a ,λ) to suppress noise. The soft threshold function is defined as:

[0115] S(w a ,λ)=sign(w a )·max(|w a |-λ,0).

[0116] Among them, w a is the ath original wavelet coefficient; λ wavelet threshold; sign(w a ) is w a The sign function of .

[0117] (3) Signal reconstruction: Use the wavelet coefficients after threshold processing to reconstruct the signal and obtain the denoised signal x'(t):

[0118] x'(t)=T+H(D,λ).

[0119] Among them, H(D,λ) is the detail part with the threshold function applied.

[0120] Take the proportional solenoid coil sensor signal as an example, Figure 3 and Figure 4 As shown in the figure, it can be seen that the soft threshold denoising method can effectively reduce the noise of the original signal, and the noise of the curve is reduced after denoising.

[0121] As an optional implementation, the time domain feature parameter set includes: maximum value, minimum value, peak-to-peak value, average value, effective value, variance, root mean square amplitude, kurtosis, skewness, shape factor, and crest factor. The time domain feature parameter set has a total of 11 features.

[0122] The frequency domain feature parameter set includes: frequency domain amplitude mean, centroid frequency, root mean square frequency, mean square frequency, frequency variance, frequency domain amplitude variance, frequency domain amplitude skewness, and frequency standard deviation. The frequency domain feature parameter set contains 7 features in total.

[0123] The time-frequency domain fault feature parameter set includes: energy values ​​of 8 nodes extracted based on the wavelet packet decomposition method.

[0124] Specifically, a method based on wavelet packet decomposition is used to extract the energy characteristics of the sensor signal as the time-frequency domain characteristics. The specific calculation formula is:

[0125]

[0126] Among them, e p represents the total energy of each frequency band in the pth layer after decomposition; dq,p(o) represents the oth wavelet packet coefficient at node (p,q); and M is the number of wavelet packet coefficients at each node. Here, the original signal is decomposed into three layers of wavelet packets. The third layer has eight nodes. The energy value of each node in the third layer can be calculated and used as the time-frequency domain features E1 to E8 of the original signal.

[0127] Based on the above work, the fault signal of each sensor after noise reduction is analyzed in the time domain, frequency domain, and time-frequency domain. Each feature is normalized to obtain a fault feature set consisting of 26 features, namely the initial feature set S.

[0128] As an optional implementation, step 205 includes:

[0129] Identify any sensor as the target sensor.

[0130] The mutual information between each feature in the initial feature set of the target sensor and the true value of the fault type of the training electro-hydraulic proportional servo valve is calculated respectively.

[0131] The features in the initial feature set are screened based on the mutual information between each feature and the true value of the fault type of the training electro-hydraulic proportional servo valve, and a filtered feature set of the target sensor is obtained.

[0132] Calculate the mutual information between any two features in the filtered feature set.

[0133] The feature set after the primary screening is screened based on the mutual information between any two features to obtain the feature set after the secondary screening.

[0134] The mRMR value of each feature in the feature set after secondary screening is calculated respectively to determine multiple candidate features.

[0135] An incremental search strategy is used to determine the target feature set of the target sensor based on the initial feature set of the target sensor and multiple candidate features.

[0136] Specifically, step 205 includes:

[0137] First, calculate the mutual information between each feature and the fault type. Given a feature set S = (S1, S2, ... S N ), the mutual information between each feature and the fault type c can be calculated to determine the maximum correlation of the features, which is expressed as:

[0138] maxD(S,c).

[0139]

[0140] Where D(S,c) is the mutual information between each feature and fault type c; I(S b ,c) represents feature S b The mutual information between and label c.

[0141] Then, the mutual information between the feature pairs in the selected feature set is calculated to determine the degree of redundancy between them. The goal of the minimum redundancy condition is to select a feature subset with the least information overlap between the features. By minimizing the mutual information between the features, the selected feature set is guaranteed to contain no duplicate information. The objective function can be expressed as:

[0142] minR(S b ,S c ).

[0143]

[0144] Among them, R(S b ,S c ) is the mutual information between features; I(S b ,S c ) is the characteristic S b With feature S c The mutual information between them.

[0145] Finally, the mRMR value of each candidate feature is calculated based on mutual information and redundancy, and the feature with the highest mRMR value is selected and added to the selected feature set. b , mRMR(S b ) is calculated as follows:

[0146]

[0147] Step 3: Based on the fault type obtained through diagnosis of each sensor and the decision weight of the corresponding sensor, the fault type of the electro-hydraulic proportional servo valve to be diagnosed is determined.

[0148] Specifically, step 3 includes:

[0149] Calculate the decision weight of each sensor:

[0150]

[0151] in, is the voting weight of the kth sensor when diagnosing the rth fault; K is the total number of sensors participating in the fault diagnosis; is the diagnostic accuracy of the k-th sensor when diagnosing the r-th fault.

[0152] The fusion result based on the weighted voting method can be calculated according to the following formula:

[0153]

[0154] Among them, VOTE_x is the final diagnosis result obtained by voting fusion, that is, the fault type of the electro-hydraulic proportional servo valve to be diagnosed; C is the number of fault types; g k (x1) The fault classification result of the kth sensor for sample x1; V[g k (x1),r] is the judgment result of the kth sensor on the rth fault; is the total number of votes obtained for the rth fault, and the fault with the most votes, VOTE_x, is the final diagnosis result obtained by vote fusion.

[0155] The weighted voting algorithm is used to integrate the fault diagnosis results of all sensors at the decision layer. The fault diagnosis confusion matrix results are as follows: Figure 14 By analyzing the fault diagnosis results of the confusion matrix, we can summarize the diagnostic accuracy of the multi-sensor information fusion fault diagnosis method for various faults, as shown in Table 3.

[0156] Table 3 Accuracy of weighted voting multi-sensor information capacity fault diagnosis algorithm

[0157] Fault type Diagnostic accuracy Valve normal 93.06％ LVDT constant bias 96.93％ LVDT constant gain 94.01％ Valve core stuck 95.53％ Valve core wear 92.17％ Proportional solenoid degradation 89.54％ Multiple faults 94.20％ Average accuracy 93.64％

[0158] By comparing and analyzing the data in Table 2 and Table 3, it can be found that after adopting the multi-sensor information fusion fault diagnosis algorithm with a weighted voting mechanism, its average diagnostic accuracy has been significantly improved compared with the diagnostic algorithm using only a single sensor. This result further confirms the superior performance of the method of the present application in fault diagnosis.

[0159] In an exemplary embodiment, a computer device is provided, including: a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the computer program to implement the electro-hydraulic proportional servo valve fault diagnosis method in Example 1.

[0160] In an exemplary embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the electro-hydraulic proportional servo valve fault diagnosis method in Example 1 is implemented.

[0161] In an exemplary embodiment, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the electro-hydraulic proportional servo valve fault diagnosis method in embodiment 1 is implemented.

[0162] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 15 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O) and a communication interface. The processor, memory and input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for diagnosing faults of an electro-hydraulic proportional servo valve is implemented.

[0163] Those skilled in the art will understand that Figure 15 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0164] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0165] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, database or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory may include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0166] The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processors involved in the various embodiments provided herein may include, but are not limited to, general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic units, data processing logic units based on quantum computing, and the like.

[0167] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0168] This document uses specific examples to illustrate the principles and implementation methods of this application. The description of the above examples is only intended to help understand the method and core concept of this application. At the same time, for those skilled in the art, based on the concept of this application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A fault diagnosis method for an electro-hydraulic proportional servo valve, characterized in that: The electro-hydraulic proportional servo valve fault diagnosis method comprises: Obtaining the original signals and decision weights of each sensor in the electro-hydraulic proportional servo valve to be diagnosed; the sensors include: a coil current sensor, an LVDT position sensor, a spool valve A port pressure sensor, and a spool valve B port pressure sensor; Based on the original signal of each sensor in the electro-hydraulic proportional servo valve to be diagnosed and the corresponding fault classification sub-model, the fault type obtained by diagnosis of the corresponding sensor is determined; each sensor corresponds to a fault classification sub-model, and each fault classification sub-model is obtained by training a support vector machine optimized by the Grasshopper optimization algorithm using the training set of the corresponding sensor. The fault types are: normal valve, LVDT offset, LVDT gain, valve core stuck, valve core wear, coil turn short circuit, or multiple faults; determining the fault type of the electro-hydraulic proportional servo valve to be diagnosed based on the fault type obtained through diagnosis of each sensor and the decision weight of the corresponding sensor; The training process of each fault classification sub-model includes: The grasshopper optimization algorithm is used to optimize the support vector machine to obtain the optimized support vector machine; Acquire a raw data set of the training electro-hydraulic proportional servo valve; the raw data set includes: a true value of the fault type of the training electro-hydraulic proportional servo valve and raw signals of each sensor in the training electro-hydraulic proportional servo valve; The wavelet soft threshold noise reduction method is used to reduce the noise of the original signal of each sensor in the training electro-hydraulic proportional servo valve, and the denoised signal of the corresponding sensor is obtained. Extract fault features in the time domain, frequency domain, and time-frequency domain for the denoised signal of each sensor to obtain an initial feature set for the corresponding sensor; the initial feature set includes a time domain feature parameter set, a frequency domain feature parameter set, and a time-frequency domain fault feature parameter set; The mRMR feature selection algorithm is used to screen the initial feature set of each sensor to obtain the target feature set of the corresponding sensor; Based on the target feature set of each sensor and the actual value of the fault type of the training electro-hydraulic proportional servo valve, a training set of the corresponding sensor is constructed; The optimized support vector machine is trained using the training set of each sensor to obtain the fault classification sub-model corresponding to the corresponding sensor.

2. The electro-hydraulic proportional servo valve fault diagnosis method according to claim 1, characterized in that: Based on the original signal of each sensor in the electro-hydraulic proportional servo valve to be diagnosed and the corresponding fault classification sub-model, the fault type obtained by the corresponding sensor diagnosis is determined, including: Determine any sensor in the electro-hydraulic proportional servo valve to be diagnosed as the current sensor; The wavelet soft threshold noise reduction method is used to reduce the noise of the original signal of the current sensor to obtain the denoised signal of the current sensor; For the denoised signal of the current sensor, fault features are extracted in the time domain, frequency domain, and time-frequency domain respectively to obtain the initial feature set of the current sensor, thereby determining the target feature set of the current sensor; The target feature set of the current sensor is input into the fault classification sub-model corresponding to the current sensor to obtain the fault type diagnosed by the current sensor.

3. The electro-hydraulic proportional servo valve fault diagnosis method according to claim 1, characterized in that: The optimized support vector machine is trained using the training set of each sensor to obtain the fault classification sub-model corresponding to the corresponding sensor, including: Taking the target feature set of each sensor as input and the true value of the fault type of the training electro-hydraulic proportional servo valve as output, the optimized support vector machine is trained to obtain the fault classification sub-model corresponding to the corresponding sensor.

4. The electro-hydraulic proportional servo valve fault diagnosis method according to claim 2, characterized in that: The process of using the wavelet soft threshold noise reduction method to reduce the noise of the original signal and obtain the denoised signal includes: Perform wavelet decomposition on the original signal to obtain the wavelet transform of the original signal; Performing threshold processing on the wavelet coefficients in the wavelet transform of the original signal to obtain the wavelet coefficients after threshold processing; The signal is reconstructed based on the wavelet coefficients after threshold processing to obtain the denoised signal.

5. The electro-hydraulic proportional servo valve fault diagnosis method according to claim 2, characterized in that: The time domain characteristic parameter set includes: maximum value, minimum value, peak-to-peak value, average value, effective value, variance, root mean square amplitude, kurtosis, skewness, shape factor and crest factor; The frequency domain characteristic parameter set includes: domain amplitude mean, centroid frequency, root mean square frequency, mean square frequency, frequency variance, frequency domain amplitude variance, frequency domain amplitude skewness and frequency standard deviation; The time-frequency domain fault characteristic parameter set includes: energy values ​​of 8 nodes extracted based on a wavelet packet decomposition method.

6. The electro-hydraulic proportional servo valve fault diagnosis method according to claim 1, characterized in that: The mRMR feature selection algorithm is used to screen the initial feature set of each sensor to obtain the target feature set of the corresponding sensor, including: Determine any sensor as a target sensor; Calculate the mutual information between each feature in the initial feature set of the target sensor and the true value of the fault type of the training electro-hydraulic proportional servo valve; The features in the initial feature set are screened based on the mutual information between each feature and the true value of the fault type of the training electro-hydraulic proportional servo valve to obtain a filtered feature set of the target sensor; Calculate the mutual information between any two features in the feature set after a screening; The feature set after the primary screening is screened based on the mutual information between any two features to obtain the feature set after the secondary screening; Calculate the mRMR value of each feature in the feature set after secondary screening to determine multiple candidate features; An incremental search strategy is used to determine the target feature set of the target sensor based on the initial feature set of the target sensor and multiple candidate features.

7. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the electro-hydraulic proportional servo valve fault diagnosis method according to any one of claims 1 to 6.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the electro-hydraulic proportional servo valve fault diagnosis method according to any one of claims 1 to 6 is implemented.

9. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the electro-hydraulic proportional servo valve fault diagnosis method according to any one of claims 1 to 6 is implemented.

Images