Hydraulic Pump Leakage Fault Diagnosis Method

Abstract

The present invention relates to a method for diagnosing hydraulic pump leakage faults. The method extracts the time and frequency domain features of the pressure and flow signals from the collected hydraulic pump outlet pressure and flow signals, then extracts the joint time and frequency domain features through HHT transformation, combines them to form a feature vector, and sets a corresponding fault category label. A linear discriminant analysis (LDA) algorithm with supervised learning is then used to perform dimensionality reduction processing on the feature data set. Finally, a sparrow search algorithm (SSA) is used to optimize the penalty parameter c and kernel function parameter g of a support vector machine classification model, obtaining an SVM intelligent diagnosis model with the optimal kernel function parameter g and penalty parameter c. Based on the model and a test set, the hydraulic pump leakage fault category is determined. The present invention can achieve more accurate and rapid identification of hydraulic pump leakage faults, which is conducive to ensuring the smooth and safe operation of the hydraulic pump.

Description

Technical Field

[0001] The invention belongs to the technical field of hydraulic pump fault diagnosis, and in particular to a hydraulic pump leakage fault diagnosis method based on LDA dimension reduction and SSA-SVM. Background Art

[0002] Hydraulic systems, with their fast response and smooth operation, are widely used in fields such as coal mining and aviation. However, hydraulic system failures are often hidden, random, and intertwined, making maintenance and management challenging. The hydraulic pump is a core component in the hydraulic circuit, and its proper operation directly impacts the stability of the entire hydraulic system.

[0003] Hydraulic pump leakage has always been a critical issue in the operation of hydraulic systems. When a hydraulic pump leaks, it not only affects the control accuracy of the entire mechanical system and the operation of the equipment, but can also damage hydraulic components. Therefore, timely detection of hydraulic system leaks has become a key step in hydraulic system maintenance. Failure mechanisms and failure modes in hydraulic systems are complex and diverse, making it difficult to accurately identify hydraulic pump leaks while the equipment is operating.

[0004] To ensure the smooth and safe operation of hydraulic systems, a leak fault diagnosis model with good diagnostic effectiveness and rapid diagnostic process must be established. Traditional hydraulic pump fault diagnosis methods mostly collect vibration signals as model input. However, due to the harsh operating environment of hydraulic pumps, the collected fault signals are often mixed with too many interference signals, making it difficult to obtain effective vibration signals. In addition, traditional neural network diagnosis models are prone to problems such as the curse of dimensionality and local extreme values. The SVM model can better address these problems, but the randomness of the SVM model parameter selection affects the diagnostic effectiveness. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a hydraulic pump leakage fault diagnosis method. The method reduces the dimension of the extracted fault feature data set based on the linear discriminant analysis algorithm (LDA), optimizes the SVM classification model through the sparrow search algorithm (SSA), and realizes accurate and rapid identification of hydraulic pump leakage faults.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0007] A hydraulic pump leakage fault diagnosis method comprising:

[0008] Step 1: Collect the pressure and flow signals of the hydraulic pump, and extract the time domain and frequency domain features of the pressure and flow signals to obtain time domain features and frequency domain features;

[0009] Step 2: Perform empirical mode decomposition (EMD) on the pressure and flow signals, calculate the variance contribution rate of the intrinsic mode functions IMF1, IMF2, and IMF3 obtained by decomposition, and obtain the joint characteristics in the time and frequency domains;

[0010] Step 3: Perform Hilbert-Huang transform (HHT) on the intrinsic mode function IMF1 to obtain the instantaneous frequency and instantaneous amplitude vector of the intrinsic mode function IMF1, calculate the maximum value of the instantaneous amplitude and instantaneous frequency vector, and obtain the joint feature in the time-frequency domain;

[0011] Step 4: Combine the time domain features, frequency domain features, and time-frequency domain joint features obtained in steps 1, 2, and 3 to form a first data set and set corresponding fault category labels;

[0012] Step 5: Use the linear discriminant analysis algorithm (LDA) to reduce the dimensionality of the first data set to obtain the second data set, which is divided into 50% test set and 50% training set;

[0013] Step 6: Use the training set and sparrow search algorithm (SSA) to optimize the penalty parameter c and kernel function parameter g of the support vector machine fault diagnosis model, build an SVM diagnosis model with optimal parameters, and derive the hydraulic pump leakage fault category based on the SVM diagnosis model and the test set.

[0014] Furthermore, in step one, a pressure sensor and a flow sensor are installed at the outlet of the hydraulic pump to collect the pressure and flow signals of the hydraulic pump, wherein the sampling frequency of the pressure sensor is 100 Hz and the sampling frequency of the flow sensor is 10 Hz; each sample includes 1 pressure and 1 flow signal, and the collection time is 1 minute.

[0015] Furthermore, in step one, the time domain features include 12 features in total, namely, mean, root mean square value, maximum value, peak-to-peak value, standard deviation, skewness, kurtosis, peak factor, margin factor, waveform factor, pulse factor and variance.

[0016] Furthermore, in step 1, the frequency domain features include three features: centroid frequency, mean square frequency, and frequency variance.

[0017] Furthermore, the first data set is a 40-dimensional feature vector of 20*2, and the second data set obtained after dimensionality reduction is a 6-dimensional feature vector.

[0018] Furthermore, the Hilbert-Huang transform (HHT) described in step 3 includes two parts, namely, empirical mode decomposition (EMD) and Hilbert transform; wherein, the empirical mode decomposition (EMD) decomposes the collected hydraulic pump pressure and flow signals into a series of intrinsic mode functions (IMFs);

[0019]

[0020] Where: n is the number of eigenmodes; c i (t) is the i-th eigenmode function; r n (t) is the residual function; then the variance contribution rate of IMF1, IMF2, and IMF3 obtained by decomposition is calculated:

[0021]

[0022]

[0023] Where: Di is the variance of the i-th order IMF component, Δt is the signal data acquisition time interval, which is 0.01 seconds for the pressure signal and 0.1 seconds for the flow signal.

[0024] Furthermore, in step 3, the first intrinsic mode function c1(t) obtained after EMD decomposition is subjected to Hilbert transform to obtain its instantaneous amplitude and instantaneous frequency. For any intrinsic mode function c i (t), the Hilbert transform is defined as follows:

[0025]

[0026] The analytical signal z(t) is as follows:

[0027] z(t)=c(t)+jy(t)=a(t)e jθ(t) (5)

[0028] Where a(t) is the instantaneous amplitude of c1(t), and θ(t) is the instantaneous phase of c1(t):

[0029]

[0030]

[0031] The instantaneous frequency of c1(t) is:

[0032]

[0033] Furthermore, the hydraulic pump leakage fault diagnosis accuracy of the SVM training set is used as the fitness function, and the optimal fitness value and the corresponding penalty parameter c and kernel function parameter g are retained; the principle of the SSA (Sparrow Search Algorithm, SSA) sparrow search algorithm is as follows:

[0034] Assume that the population X of sparrows is:

[0035]

[0036] Where n is the number of sparrows, which is 100, and d is the dimension of the variable to be optimized, which is 2.

[0037] f represents the fitness value, which is the fault diagnosis accuracy of the training set corresponding to the current value. The fitness function F x as follows:

[0038]

[0039] The finder is a sparrow with a good fitness value in the entire group. It is responsible for finding food for the group and providing foraging directions for followers. The number of finders is PD, which is 20. Their position is updated as follows:

[0040]

[0041] Where i = 1, 2, 3…, PD; j = 1, 2, 3…, d; t represents the current iteration number; iter max is a constant, indicating the maximum number of iterations; α∈[0,1] is a random number; R2∈[0,1] is the warning value, indicating the degree of alertness of the sparrow to predators; ST∈[0.5,1] is the warning threshold, which is 0.6; Q is a random number that follows a normal distribution; L represents a 1×d matrix where each element is 1;

[0042] The follower value is 80, and its position is updated as follows:

[0043]

[0044] Among them, X p is the optimal position currently occupied by the discoverer, X worst It represents the current global worst position. A represents a 1×d matrix, in which each element is randomly assigned a value of 1 or -1, and A + =A T (AA T ) -1When i>n / 2, it means that the i-th joiner with a lower fitness value has not obtained food and needs to fly to other places to find food to obtain a higher fitness value. The value of the vigilant is 10, and the position update method is as follows:

[0045]

[0046] Among them, X best is the current global optimal position. β is a step size control parameter, which is a random number that obeys a normal distribution with a mean of 0 and a variance of 1; K∈[-1,1] is a random number, and fi is the fitness value of the current sparrow individual; f g and f w are the current global best and worst fitness values ​​respectively.

[0047] Furthermore, the steps of constructing the SVM diagnostic model with optimal parameters are as follows:

[0048] (1) First, the extracted time domain, frequency domain, and time-frequency domain joint domain features of the hydraulic pump leakage fault are used as the input of the diagnosis model, and the fault category is used as the output value of the diagnosis model. There are three fault categories: category 0 is no leakage, category 1 is slight leakage, and category 2 is severe leakage. Each type of fault in the dataset is divided into 50% as a training set and 50% as a test set.

[0049] (2) Initialize the relevant parameters of the SSA algorithm, including: the size of the sparrow population is 100, the maximum number of iterations is 50; the parameters to be optimized are the SVM parameters c and g, both of which are in the range of [0.01, 1000]; ST is the warning threshold, which is in the range of [0.5, 1.0], and is taken as 0.6 here; the number of discoverers is 20, and the remaining 80 are followers. In each generation, SD individuals will be randomly selected from the population as vigilants, and SD is 10;

[0050] (3) Through cross-validation, the training set samples are diagnosed, and the fault diagnosis accuracy of the corresponding training set is used as the fitness of the sparrow individual. The optimal fitness value and the corresponding sparrow individual position information are retained, that is, the corresponding penalty parameter c and kernel function parameter g values;

[0051] (4) Update the position of the discoverer according to formula (11), update the position of the follower according to formula (12), and update the position of the sentinel according to formula (13);

[0052] (5) Calculate the fitness value of the new position of the sparrow individual, compare the updated fitness value with the original optimal value, and update the global optimal information;

[0053] (6) Determine whether the number of iterations reaches the maximum number of iterations of 50. If not, repeat step (3). Otherwise, stop the iteration, output the optimal parameters, obtain the SVM diagnostic model with the optimal parameters, and derive the hydraulic pump leakage fault category based on the model and the test set.

[0054] Compared to traditional hydraulic pump fault diagnosis methods based on vibration signals, the hydraulic pump outlet pressure and flow signals used in this invention are more intuitive, requiring less signal acquisition equipment and a lower sampling rate. This invention uses a linear discriminant analysis (LDA) algorithm to reduce the dimensionality of the extracted fault feature dataset and optimizes the SVM diagnostic model using a sparrow search algorithm (SSA). This solves the problem of selecting the optimal penalty parameter c and kernel function parameter g for the SVM model, improving the model's diagnostic effectiveness and speed. This invention enables more accurate and rapid identification of hydraulic pump leakage faults, helping to ensure the smooth and safe operation of the hydraulic pump and providing a new research direction for fault diagnosis of other hydraulic equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 It is a flow chart of the implementation of the present invention;

[0056] Figure 2 This is a flow chart of the sparrow search algorithm (SSA) optimized support vector machine (SVM) diagnostic model of the present invention;

[0057] Figure 3 : is a waveform diagram of the original signal of the hydraulic pump pressure in the embodiment;

[0058] Figure 4 : This is the waveform diagram of the original signal of the hydraulic pump flow in the embodiment;

[0059] Figure 5 : is a graph of instantaneous frequency and instantaneous amplitude obtained by HHT transforming the IMF1 component decomposed from the original pressure signal without leakage in the embodiment;

[0060] Figure 6 : is a graph of instantaneous frequency and instantaneous amplitude obtained by HHT transforming the IMF1 component of the original flow signal without leakage in the embodiment;

[0061] Figure 7 : is a graph of instantaneous frequency and instantaneous amplitude obtained by HHT transforming the IMF1 component decomposed from the original pressure signal of slight leakage in the embodiment;

[0062] Figure 8 : is a graph of instantaneous frequency and instantaneous amplitude obtained by HHT transforming the IMF1 component decomposed from the original flow signal of slight leakage in the embodiment;

[0063] Figure 9: is a graph of instantaneous frequency and instantaneous amplitude obtained by HHT transforming the IMF1 component decomposed from the original pressure signal of severe leakage in the embodiment;

[0064] Figure 10 : is an instantaneous frequency and instantaneous amplitude diagram obtained by HHT transforming the IMF1 component decomposed from the original flow signal of severe leakage in the embodiment;

[0065] Figure 11 Comparison diagram of actual fault types and predicted fault types in the SVM diagnostic model optimized by SSA when the test set is input in an embodiment of the present invention. DETAILED DESCRIPTION

[0066] like Figure 1 、 2 As shown, the present invention provides a hydraulic pump leakage fault diagnosis method based on LDA dimensionality reduction and SSA-SVM. The overall concept of the method is to extract time domain and frequency domain features from the collected pressure and flow signals at the hydraulic pump outlet, then extract the time and frequency domain joint features through HHT transformation, and combine them to form a feature vector, and set the corresponding fault category label; then use the linear discriminant analysis algorithm (LDA) with supervised learning to reduce the dimensionality of the feature data set; finally, use the sparrow search algorithm (SSA) to optimize the penalty parameter c and kernel function parameter g of the support vector machine classification model (SVM), and obtain the SVM intelligent diagnosis model with the optimal value of the kernel function parameter g and the optimal penalty parameter c.

[0067] The specific implementation of the present invention is further described below with reference to the accompanying drawings and specific embodiments.

[0068] The hydraulic pump leakage fault diagnosis method based on LDA dimensionality reduction and SSA-SVM provided in this embodiment is performed according to the following specific steps.

[0069] Step 1: Collect the pressure and flow signals of the hydraulic pump, and extract the time domain and frequency domain features of the pressure and flow signals to obtain the time domain features and frequency domain features (such as Figure 3 、 4 shown).

[0070] The hydraulic system was subjected to a periodic, constant load cycle, while process values ​​were measured during the load cycle. To collect the hydraulic pump's pressure and flow signals, a pressure sensor and a flow sensor were installed at the pump outlet. The pressure sensor had a sampling frequency of 100 Hz, and the flow sensor had a sampling frequency of 10 Hz. Each sample consisted of one pressure and one flow signal, and the acquisition time was 1 minute.

[0071] The extracted hydraulic signal features include time domain features, frequency domain features, and time-frequency domain features. The time domain features include 12 features: mean, root mean square value, maximum value, peak-to-peak value, standard deviation, skewness, kurtosis, crest factor, margin factor, shape factor, pulse factor, and variance. The frequency domain features include three features: center of gravity frequency, mean square frequency, and frequency variance.

[0072] Step 2: Perform empirical mode decomposition (EMD) on the pressure and flow signals, calculate the variance contribution rate of the intrinsic mode functions IMF1, IMF2, and IMF3 obtained by decomposition to obtain the time-frequency domain joint features, and obtain three time-frequency domain joint features for each signal.

[0073] Step 3: Perform a Hilbert-Huang transform (HHT) on the intrinsic mode function IMF1 to obtain its instantaneous frequency and instantaneous amplitude vector. The maximum value of the instantaneous amplitude and instantaneous frequency vector is calculated to obtain the joint time-frequency domain features. Two joint time-frequency domain features are obtained for each signal.

[0074] The Hilbert-Huang transform (HHT) includes two parts, namely, empirical mode decomposition (EMD) and Hilbert transform.

[0075] Among them, empirical mode decomposition (EMD) decomposes the collected hydraulic pump pressure and flow signals into a series of intrinsic mode functions (IMFs);

[0076]

[0077] Where: n is the number of eigenmodes; c i (t) is the i-th eigenmode function; r n (t) is the residual function.

[0078] Calculate the variance contribution rate of IMF1, IMF2, and IMF3 obtained by decomposition:

[0079]

[0080]

[0081] Where: Di is the variance of the i-th order IMF component, Δt is the signal data acquisition time interval, which is 0.01 seconds for the pressure signal and 0.1 seconds for the flow signal.

[0082] In order to further reveal the characteristics of local changes in the signal, the first intrinsic mode function IMF1 obtained after EMD decomposition is Hilbert transformed to obtain its instantaneous amplitude and instantaneous frequency. For any intrinsic mode function c i (t) Hilbert transform is defined as follows:

[0083]

[0084] The analytical signal z(t) is as follows:

[0085] z(t)=c(t)+jy(t)=a(t)e jθ(t) (5)

[0086] Where a(t) is the instantaneous amplitude of c1(t), and θ(t) is the instantaneous phase of c1(t):

[0087]

[0088]

[0089] The instantaneous frequency of c1(t) is:

[0090]

[0091] After steps 2 and 3, five time-frequency domain joint features can be obtained for each pressure and flow signal, including the corresponding variance contribution rates of IMF1, IMF2, and IMF3, and the maximum instantaneous frequency and instantaneous amplitude of IMF1.

[0092] Step 4: Combine the time domain features, frequency domain features, and time-frequency domain joint features obtained in steps 1, 2, and 3 to form a first data set and set corresponding labels.

[0093] Among them, 12 time domain features and 3 frequency domain features are obtained in step 1; 3 time-frequency domain joint features are obtained in step 2, and 2 time-frequency domain joint features are obtained in step 3; the above features are extracted from the hydraulic pump outlet pressure signal and flow signal and combined to obtain a 20*2-dimensional 40-dimensional feature vector as the first data set.

[0094] Step 5: Use the linear discriminant analysis algorithm (LDA) to reduce the dimension of the first data set to obtain a second data set consisting of a new 6-dimensional feature vector, which is divided into a 50% test set and a 50% training set.

[0095] LDA (Linear Discriminant Analysis) is a supervised dimensionality reduction method that can use prior knowledge and experience of categories in the dimensionality reduction process.

[0096] Step 6: Use the training set and sparrow search algorithm (SSA) to optimize the penalty parameter c and kernel function parameter g of the support vector machine fault diagnosis model (SVM), build the optimal parameter SVM model, and obtain the hydraulic pump leakage fault category based on the optimal parameter SVM model and the test set.

[0097] The hydraulic pump leakage fault diagnosis accuracy of the SVM training set is used as the fitness function, and the optimal fitness value and the corresponding penalty parameter c and kernel function parameter g are retained.

[0098] The principle of SSA (Sparrow Search Algorithm, SSA) sparrow search algorithm is as follows.

[0099] Assume that the population X of sparrows is:

[0100]

[0101] Where n is the number of sparrows, which is 100, and d is the dimension of the variable to be optimized, which is the number of parameters to be optimized in the SVM diagnostic model, which is 2.

[0102] f represents the fitness value, which is the fault diagnosis accuracy of the training set corresponding to the current value. The fitness function F x as follows:

[0103]

[0104] The finder is a sparrow with a good fitness value in the entire group. It is responsible for finding food for the group and providing foraging directions for followers. The number of finders is PD, which is 20. Their positions are updated as follows:

[0105]

[0106] Where i = 1, 2, 3…, PD; j = 1, 2, 3…, d; t represents the current iteration number; iter max is a constant, indicating the maximum number of iterations; α∈[0,1] is a random number; R2∈[0,1] is the warning value, indicating the degree of alertness of the sparrow to predators; ST∈[0.5,1] is the warning threshold, which is 0.6; Q is a random number that obeys the normal distribution; L represents a 1×d matrix where each element is 1.

[0107] The follower value is 80, and its position is updated as follows:

[0108]

[0109] Among them, X p is the optimal position currently occupied by the discoverer, X worstIt represents the current global worst position. A represents a 1×d matrix, in which each element is randomly assigned a value of 1 or -1, and A + =A T (AA T ) -1 When i>n / 2, this indicates that the i-th participant with a lower fitness value has not obtained food and needs to fly to other places to find food in order to obtain a higher fitness.

[0110] The value of the alert is 10, and the position is updated as follows:

[0111]

[0112] Among them, X best is the current global optimal position. β is a step size control parameter, which is a random number that obeys a normal distribution with a mean of 0 and a variance of 1; K∈[-1,1] is a random number, f i is the fitness value of the current sparrow individual; f g and f w are the current global best and worst fitness values, respectively. ε is the smallest constant to avoid zero in the denominator.

[0113] The specific steps of optimizing the SVM hydraulic pump leakage fault intelligent diagnosis model with the SSA algorithm are as follows:

[0114] (1) First, the extracted time domain, frequency domain, and time-frequency domain joint features of the hydraulic pump leakage fault are used as the input of the diagnostic model, and the fault category is used as the output value of the diagnostic model. There are three fault categories: category 0 is no leakage, category 1 is slight leakage, and category 2 is severe leakage. The dataset has a total of 2205 samples, with 1221 samples in category 0, 492 samples in category 1, and 492 samples in category 2. For each fault category in the dataset, 50% is divided into the training set and 50% is divided into the test set.

[0115] (2) Initialize the relevant parameters of the SSA algorithm, including: the size of the sparrow population is 100, the maximum number of iterations is 50; the parameters to be optimized are the SVM parameters c and g, both of which are in the range of [0.01, 1000]; ST is the warning threshold, which is in the range of [0.5, 1.0], and is taken as 0.6 here; the number of discoverers is 20, and the rest are followers. In each generation, SD individuals will be randomly selected from the population as warning individuals, and SD is 10;

[0116] (3) Through cross-validation, the training set samples are diagnosed, and the fault diagnosis accuracy of the corresponding training set is used as the fitness of the sparrow individual. The optimal fitness value and the corresponding sparrow individual position information are retained, that is, the corresponding penalty parameter c and kernel function parameter g values;

[0117] (4) Update the position of the discoverer according to formula (11), update the position of the follower according to formula (12), and update the position of the sentinel according to formula (13);

[0118] (5) Calculate the fitness value of the new position of the sparrow individual, compare the updated fitness value with the original optimal value, and update the global optimal information;

[0119] (6) Determine whether the number of iterations reaches the maximum number of iterations of 50. If not, repeat step (3). Otherwise, stop the iteration and output the optimal parameters. The optimal SVM diagnostic model is obtained and the hydraulic pump leakage fault category is obtained based on the model and the test set. The comparison diagram of the actual fault type and the predicted fault type in the SVM diagnostic model optimized by SSA for the test set is shown in the figure below. Figure 11 shown.

[0120] It should be noted that the above-mentioned specific embodiments are merely illustrative and not restrictive. Under the guidance of the present invention, ordinary technicians in this field can make various modified solutions without departing from the scope of protection of the present invention and the claims, all of which are within the scope of protection required by the present invention.

Claims

1. A method for diagnosing hydraulic pump leakage faults, characterized in that: include: Step 1: Collect the pressure and flow signals of the hydraulic pump, and extract the time domain and frequency domain features of the pressure and flow signals to obtain time domain features and frequency domain features; Step 2: Perform empirical mode decomposition on the pressure and flow signals, calculate the variance contribution rate of the intrinsic mode functions IMF1, IMF2, and IMF3 obtained by decomposition, and obtain the joint characteristics of the time and frequency domains; Step 3: Perform Hilbert-Huang transform on the intrinsic mode function IMF1 to obtain the instantaneous frequency and instantaneous amplitude vector of the intrinsic mode function IMF1, calculate the maximum value of the instantaneous amplitude and instantaneous frequency vector, and obtain the joint feature in the time-frequency domain; The Hilbert-Huang transform comprises two parts, namely, empirical mode decomposition and Hilbert transform; wherein, the empirical mode decomposition decomposes the collected hydraulic pump pressure and flow signals into a series of intrinsic mode functions (IMFs); Where: n is the number of eigenmodes; c i (t) is the i-th eigenmode function; r n (t) is the residual function; Calculate the variance contribution rate of IMF1, IMF2, and IMF3 obtained by decomposition: Where: Di is the variance of the i-th order IMF component, Δt is the signal data acquisition time interval, which is 0.01 seconds for the pressure signal and 0.1 seconds for the flow signal; The first intrinsic mode function c1(t) obtained after empirical mode decomposition is Hilbert transformed to obtain its instantaneous amplitude and instantaneous frequency. For any intrinsic mode function c i (t), the Hilbert transform is defined as follows: The analytical signal z(t) is as follows: z(t)=c(t)+jy(t)=a(t)e jθ(t) (5) Where a(t) is the instantaneous amplitude of c1(t), and θ(t) is the instantaneous phase of c1(t): The instantaneous frequency of c1(t) is: Step 4: Combine the time domain features, frequency domain features, and time-frequency domain joint features obtained in steps 1, 2, and 3 to form a first data set and set corresponding fault category labels; Step 5: Use the linear discriminant analysis algorithm to reduce the dimensionality of the first data set to obtain the second data set, which is divided into 50% test set and 50% training set; Step 6: Use the training set and sparrow search algorithm to optimize the penalty parameter c and kernel function parameter g of the support vector machine fault diagnosis model, build the SVM diagnosis model with the best parameters, and obtain the hydraulic pump leakage fault category based on the SVM diagnosis model and the test set.

2. The hydraulic pump leakage fault diagnosis method according to claim 1, characterized in that: In step 1, a pressure sensor and a flow sensor are installed at the outlet of the hydraulic pump to collect the pressure and flow signals of the hydraulic pump. The sampling frequency of the pressure sensor is 100 Hz, and the sampling frequency of the flow sensor is 10 Hz. Each sample includes 1 pressure and 1 flow signal, and the collection time is 1 minute.

3. The hydraulic pump leakage fault diagnosis method according to claim 1, characterized in that: In step 1, the time domain features include 12 features in total, namely, mean, root mean square value, maximum value, peak-to-peak value, standard deviation, skewness, kurtosis, peak factor, margin factor, shape factor, pulse factor and variance.

4. The hydraulic pump leakage fault diagnosis method according to claim 1, characterized in that: In step 1, the frequency domain features include three features: centroid frequency, mean square frequency, and frequency variance.

5. The hydraulic pump leakage fault diagnosis method according to claim 1, characterized in that: The first data set is a 40-dimensional feature vector of 20*2, and the second data set obtained after dimensionality reduction is a 6-dimensional feature vector.

6. The hydraulic pump leakage fault diagnosis method according to claim 1, characterized in that: The hydraulic pump leakage fault diagnosis accuracy of the SVM training set is used as the fitness function, and the optimal fitness value and the corresponding penalty parameter c and kernel function parameter g are retained. The principle of the SSA sparrow search algorithm is as follows. Assume that the sparrow population X is: Where n is the number of sparrows, which is 100, and d is the dimension of the variable to be optimized; f represents the fitness value, which is the fault diagnosis accuracy of the training set corresponding to the current value. The fitness function F x as follows: The finder is a sparrow with a good fitness value in the entire group. It is responsible for finding food for the group and providing foraging directions for followers. The number of finders is PD, which is 20. Their positions are updated as follows: Where i = 1, 2, 3…, PD; j = 1, 2, 3…, d; t represents the current iteration number; iter max is a constant, indicating the maximum number of iterations; α∈[0,1] is a random number; R2∈[0,1] is the warning value, indicating the degree of alertness of the sparrow to predators; ST∈[0.5,1] is the warning threshold, which is 0.6; Q is a random number that follows a normal distribution; L represents a 1×d matrix where each element is 1; The follower value is 80, and its position is updated as follows: Among them, X p is the optimal position currently occupied by the discoverer, X worst It represents the current global worst position, A represents a 1×d matrix, in which each element is randomly assigned a value of 1 or -1, and A + =A T (AA T ) -1 , when i>n / 2, this indicates that the i-th joiner with a lower fitness value has not obtained food and needs to fly to other places to find food in order to obtain higher fitness; The value of the alert is 10, and the position is updated as follows: Among them, X best is the current global optimal position; β is the step size control parameter, which is a random number that obeys the normal distribution with a mean of 0 and a variance of 1; K∈[-1,1] is a random number, fi is the fitness value of the current sparrow individual; f g and f w are the current global best and worst fitness values ​​respectively.

7. The hydraulic pump leakage fault diagnosis method according to claim 6, characterized in that: The steps to construct the SVM diagnostic model with optimal parameters are as follows: (1) First, the extracted time domain, frequency domain, and time-frequency domain joint features of the hydraulic pump leakage fault are used as the input of the diagnosis model, and the fault category is used as the output value of the diagnosis model. There are three fault categories: category 0 is no leakage, category 1 is slight leakage, and category 2 is severe leakage. Each type of fault in the dataset is divided into 50% as a training set and 50% as a test set. (2) Initialize the parameters related to the sparrow search algorithm, including: the sparrow population size is 100, the maximum number of iterations is 50; the parameters to be optimized are the SVM parameters c and g, both of which are in the range of [0.01, 1000]; ST is the warning threshold, which is in the range of [0.5, 1.0], and is taken as 0.6 here; the number of discoverers is 20, and the rest are followers. In each generation, SD individuals will be randomly selected from the population as warning individuals, and SD is 10; (3) Through cross-validation, the training set samples are diagnosed, and the fault diagnosis accuracy of the corresponding training set is used as the fitness of the sparrow individual. The optimal fitness value and the corresponding sparrow individual position information are retained, that is, the corresponding penalty parameter c and kernel function parameter g values; (4) Update the position of the discoverer according to formula (11), update the position of the follower according to formula (12), and update the position of the sentinel according to formula (13); (5) Calculate the fitness value of the new position of the sparrow individual, compare the updated fitness value with the original optimal value, and update the global optimal information; (6) Determine whether the number of iterations reaches the maximum number of iterations of 50. If not, repeat step (3). Otherwise, stop the iteration, output the optimal parameters, obtain the SVM diagnostic model with the optimal parameters, and derive the hydraulic pump leakage fault category based on the model and the test set.

Images