Underwater EHA hydraulic cylinder leakage fault diagnosis method

Abstract

The invention discloses an underwater EHA hydraulic cylinder leakage fault diagnosis method. The method comprises the following steps: preprocessing an original data set of fault data of an underwater EHA hydraulic cylinder to obtain a training data set and a test data set; improving a convolutional neural network model through a dung beetle optimization algorithm in combination with the test data set; processing the test data set through an improved convolutional neural network model to obtain a fault feature classification result of the underwater EHA hydraulic cylinder; and performing visualization processing on the fault feature classification result to obtain visual fault feature data. According to the method, the reconstruction data subjected to adaptive variable mode decomposition denoising is used as the input vector of the CNN model, the improved dung beetle optimization algorithm is adopted to optimize the hyper-parameters of the CNN model, the result judgment is more accurate, the efficiency is higher, the feature data of leakage faults of different degrees of the hydraulic cylinder are subjected to visualization processing, and the fault diagnosis accuracy is improved. And the accuracy and efficiency of fault identification are obviously improved.

Description

Technical Field

[0001] The invention relates to the technical field of data processing, and in particular to a method for diagnosing leakage faults of an underwater EHA hydraulic cylinder. Background Art

[0002] In recent years, with the advancement of artificial intelligence, big data and information processing technology, fault diagnosis technology has been widely used in various fields. Especially in underwater EHA systems, the improvement of fault diagnosis technology can effectively reduce the equipment failure rate and ensure the safety and stability of operations. The fault diagnosis system mainly includes modules such as data acquisition and processing, fault identification and location, decision support and execution control. Through the combination of these technologies, real-time monitoring and intelligent management of underwater EHA systems can be achieved.

[0003] Electro-hydraulic actuator (EHA) is a power electric actuator that integrates pump, valve, cylinder and drive control system. It is widely used in the fields of ocean exploration and deep-sea operations. EHA is a complex nonlinear system with mutual coupling of electromechanical, hydraulic and mechanical. There are many uncontrollable factors in the working environment of underwater mechanical equipment. Underwater undercurrents cause large load disturbances. The valve-controlled hydraulic system with centralized oil source has a complex structure and numerous pipelines. The hydraulic system is prone to leakage, which pollutes the underwater environment. Real-time status monitoring and fault diagnosis of electro-hydraulic actuators can not only ensure the normal operation of the actuator, but also detect problems and perform maintenance in time to avoid unnecessary losses. However, the existing hydraulic cylinder leakage fault diagnosis methods are not accurate and efficient for real-time status monitoring and fault diagnosis of electro-hydraulic actuators. Summary of the invention

[0004] The embodiment of the present invention provides an underwater EHA hydraulic cylinder leakage fault diagnosis method to solve the problem that the accuracy and efficiency of real-time status monitoring and fault diagnosis of electro-hydraulic actuators in the prior art are low.

[0005] On the one hand, an embodiment of the present invention provides a method for diagnosing a leakage fault of an underwater EHA hydraulic cylinder, comprising:

[0006] The original data set of the fault data of the underwater EHA hydraulic cylinder is preprocessed to obtain the training data set and the test data set;

[0007] Improve the convolutional neural network model by combining the dung beetle optimization algorithm with the test dataset;

[0008] Processing the test data set through the improved convolutional neural network model to obtain the fault feature classification result of the underwater EHA hydraulic cylinder;

[0009] Visualization processing is performed on the fault feature classification result to obtain visualized fault feature data.

[0010] In a possible implementation, the dung beetle optimization algorithm includes:

[0011] Initialize the dung beetle algorithm through the population of Chebyshev chaotic map;

[0012] The rolling dung beetle position update formula of the dung beetle algorithm is replaced by a curve adaptive golden sine strategy;

[0013] The reproduction and foraging position update formula of the dung beetle algorithm are improved by Levy flight of Cauchy-t mutation strategy.

[0014] In a possible implementation, improving the convolutional neural network model by using the dung beetle optimization algorithm includes:

[0015] Introducing the dung beetle optimization algorithm before training the convolutional neural network model;

[0016] Optimizing the hyperparameters of the convolutional neural network model by using the dung beetle optimization algorithm;

[0017] Encoding the optimized hyperparameters into individual parameters;

[0018] The classification error of the convolutional neural network model under each combination of the individual parameters is used as the individual fitness value for parameter optimization.

[0019] In a possible implementation, the preprocessing of the original data set of the fault data of the underwater EHA hydraulic cylinder to obtain the training data set and the test data set includes:

[0020] The original data set is preprocessed by adaptive variable mode decomposition.

[0021] In a possible implementation manner, it is characterized in that the visualizing the fault feature classification result to obtain the visualized fault feature data comprises:

[0022] The fault feature classification results are visualized using the t-SNE manifold learning method.

[0023] In one possible implementation, the population initialization of the dung beetle algorithm through the Chebyshev chaotic map includes:

[0024] Determining parameters of the Chebyshev chaotic map;

[0025] Generating initial positions of individuals in the population through the Chebyshev chaotic map;

[0026] Normalizing the individuals of the population that are not within the target range;

[0027] Performing a deduplication operation on the individuals of the population;

[0028] After the population is initialized, the fitness of the individuals in the population is evaluated.

[0029] In a possible implementation, the preprocessing of the original data set by adaptive variable mode decomposition includes:

[0030] Decomposing the original data of the original data set by adaptive variable mode decomposition to obtain IMF components;

[0031] Calculate the correlation coefficient r between the IMF component and the original signal;

[0032] The IMF components are screened by the correlation coefficient r to perform data reconstruction.

[0033] The underwater EHA hydraulic cylinder leakage fault diagnosis method of the present invention has the following advantages:

[0034] (1) The reconstructed data after adaptive variable mode decomposition and denoising is used as the input vector of the CNN model. The improved dung beetle optimization algorithm is used to optimize the hyperparameters of the convolutional neural network model, resulting in more accurate judgment and higher efficiency.

[0035] (2) The fault characteristic data of hydraulic cylinder leakage of different degrees are visualized, which significantly improves the accuracy and efficiency of fault identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0037] Figure 1 A schematic diagram of a flow chart of a method for diagnosing underwater EHA hydraulic cylinder leakage faults provided by an embodiment of the present invention;

[0038] Figure 2 A schematic diagram of the convolutional neural network structure of an underwater EHA hydraulic cylinder leakage fault diagnosis method provided by an embodiment of the present invention;

[0039] Figure 3 A schematic diagram of two-dimensional distribution scatter points of model characteristics of an underwater EHA hydraulic cylinder leakage fault diagnosis method provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0040] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0041] Figure 1 A schematic flow chart of a method for diagnosing a leakage fault of an underwater EHA hydraulic cylinder provided in an embodiment of the present invention; an embodiment of the present invention provides a method for diagnosing a leakage fault of an underwater EHA hydraulic cylinder, comprising:

[0042] The original data set of the fault data of the underwater EHA hydraulic cylinder is preprocessed to obtain the training data set and the test data set;

[0043] Improve the convolutional neural network model by combining the dung beetle optimization algorithm with the test dataset;

[0044] Processing the test data set by the improved convolutional neural network model to obtain the fault feature classification result of the underwater EHA hydraulic cylinder;

[0045] Visualization processing is performed on the fault feature classification result to obtain visualized fault feature data.

[0046] For example, by simulating the typical fault condition of EHA hydraulic cylinder leakage, the original data set is obtained, and the signal is preprocessed and reconstructed by the adaptive variable mode decomposition algorithm (ICEEMDAN). The reconstructed data after noise removal is sent to the convolutional neural network model (CNN) for training. At the same time, the dung beetle optimization algorithm (IDBO) is integrated to avoid falling into the local optimal state during the training of the convolutional neural network model, thereby improving the recognition accuracy of fault diagnosis of underwater EHA equipment, thereby achieving accurate fault classification.

[0047] Since continuous impact pulses will be generated during the normal operation of the underwater EHA, and there is noise interference, the adaptive variable mode decomposition algorithm is used to process the collected original signal. The hydraulic cylinder displacement signal under the original EHA normal working condition is decomposed by the adaptive variable mode decomposition algorithm, and the correlation coefficient between each IMF component and the original signal is calculated. The correlation coefficient reflects the similarity between the IMF component and the original vibration signal. Therefore, selecting the IMF component with higher correlation to reconstruct the signal can retain the main characteristics of the underwater EHA. The IMF above the threshold is retained, and the IMF below the threshold is discarded and does not participate in the reconstruction vector construction process.

[0048] The t-SNE manifold learning method is used to reduce the features extracted from each layer to 2 dimensions, and the outputs of the four hidden layers in the network are visualized, which improves the ability to evaluate the model's feature extraction. The model feature scatter distribution is shown in the figure below. Figure 3 shown. Figure 3 The scattered points of the same color represent samples of the same fault type, and each scattered point represents a feature point extracted and processed by the model from the original data. Its position is determined by the t-SNE algorithm based on the similarity between the feature points (the distance in the original high-dimensional space). Figure 3 It can be seen that samples of the same fault type gradually gather, samples of different fault types gradually disperse, and the normal working condition and the three types of fault modes are clearly distinguished. This shows that the t-SNE manifold learning method can automatically capture the intrinsic structure in the data and identify and classify the data in an intuitive and efficient way. Through t-SNE, the fault characteristics of hydraulic cylinders with different degrees of leakage are visualized, which significantly improves the accuracy and efficiency of fault identification.

[0049] In a possible embodiment, the dung beetle optimization algorithm includes: initializing the dung beetle algorithm through a population of Chebyshev chaotic mapping; replacing the rolling dung beetle position update formula of the dung beetle algorithm through a curve adaptive golden sine strategy; and improving the reproduction and foraging dung beetle position update formula of the dung beetle algorithm through Levy flight of the Cauchy-t mutation strategy.

[0050] Exemplarily, comprehensive improvements are made to the limitations of traditional dung beetle optimization in fault diagnosis applications. First, in order to enhance the randomness and diversity of the search, a population initialization method using Chebyshev chaotic mapping is introduced. Then, the curve adaptive golden sine strategy (CGSS) is used to replace the rolling dung beetle position update formula to improve the convergence speed and accuracy during the search process. Third, the position update formula of the breeding and foraging dung beetles is improved by using the Levy flight of the Cauchy-t mutation strategy (LCTS) to improve the exploration ability and adaptability of the search. Compared with the traditional dung beetle optimization algorithm and other optimization algorithms, the improved algorithm can converge to the optimal solution faster and has better global search ability and stability.

[0051] In a possible embodiment, improving the convolutional neural network model by using the dung beetle optimization algorithm includes: introducing the dung beetle optimization algorithm before training the convolutional neural network model; optimizing the hyperparameters of the convolutional neural network model by using the dung beetle optimization algorithm; encoding the optimized hyperparameters into individual parameters; and performing parameter optimization using the classification error of the convolutional neural network model under each combination of the individual parameters as the individual fitness value.

[0052] For example, in order to solve the problem of overfitting of the convolutional neural network model caused by too many neurons, the dung beetle optimization algorithm is introduced before training the convolutional neural network model. The convolutional neural network CNN is usually composed of an input layer, a convolutional layer, a pooling layer, a fully connected layer and an output layer. The convolutional layer mainly extracts features from the input data, and the data information from the input layer is convolved with the convolution kernel of the convolutional layer. After being processed by multiple convolutional layers and pooling layers, the feature map is passed to one or more fully connected layers. In the fully connected layer, each neuron is densely connected to all neurons in the previous layer to facilitate feature integration and further analysis. The output value of the fully connected layer of the last layer is passed to an output layer, which performs multi-category classification through the SoftMax function. The typical convolutional neural network model structure diagram is shown in the figure. Figure 2 As shown in the figure, the correlation coefficients decomposed by the adaptive variable mode decomposition algorithm are taken to form n groups of IMF components that are most relevant to the original signal, and a matrix of n*4000 is constructed. Different fault type labels are added as the input of the convolutional neural network model training model. The hyperparameters of the convolutional neural network model are optimized and selected by the dung beetle optimization algorithm method, and these parameters are encoded into individuals. The classification error of the convolutional neural network model under each parameter combination is used as the individual fitness value. The individuals continue to search in the decision space, and the fitness value continues to decrease, and finally the best parameter combination is obtained. Parameter optimization mainly optimizes the different convolution kernel sizes, quantities, and learning rates of the constructed convolutional neural network model.

[0053] In a possible embodiment, the population initialization dung beetle algorithm through Chebyshev chaotic mapping includes: determining the parameters of the Chebyshev chaotic mapping; generating the initial positions of the population individuals through the Chebyshev chaotic mapping; normalizing the population individuals that are not in the target range; deduplicating the population individuals; and evaluating the fitness of the population individuals after the population is initialized.

[0054] Exemplarily, the mapping parameters are determined, including the number of mappings, the initial value X0, the mapping range, and the number of samples.

[0055] The initial positions of individuals in the population are generated through Chebyshev mapping. The specific process is as follows:

[0056] Choose an initial value X0, usually in the range of [0,1] or [-1,1];

[0057] Apply the Chebyshev polynomials iteratively to generate new values: x n+1 =2.x n .(1-x n ) or other Chebyshev mapping forms;

[0058] Repeat the above steps until the required number of individuals is generated.

[0059] If the generated individual is not within the target range, normalization is performed to map the individual's value to the appropriate range.

[0060] Through a simple deduplication mechanism or by selecting different initial values ​​for iteration, it is ensured that the generated individuals are not repeated.

[0061] After the population is initialized, the fitness of individuals is evaluated.

[0062] In a possible embodiment, the preprocessing of the original data set of fault data of the underwater EHA hydraulic cylinder to obtain a training data set and a test data set includes: preprocessing the original data set by adaptive variable mode decomposition, and the preprocessing of the original data set by adaptive variable mode decomposition includes: decomposing the original data of the original data set by adaptive variable mode decomposition to obtain IMF components; calculating the correlation coefficient r between the IMF component and the original signal; and screening the IMF component by the correlation coefficient r for data reconstruction.

[0063] Exemplarily, the value range of the correlation coefficient r is between [-1, 1], where a value closer to 1 indicates a positive correlation, a value closer to -1 indicates a negative correlation, and a value closer to 0 indicates no correlation.

[0064] The correlation coefficient r is calculated as follows: the correlation coefficient between the IMF component and the original signal is calculated using the Pearson correlation coefficient. The calculation steps are as follows:

[0065] Prepare data: Make sure you have the original signal and the corresponding IMF components.

[0066] Calculate the mean: Calculate the mean of the original signal and the IMF component:

[0067]

[0068] in:

[0069] x i and i Represent the i-th data point of the original signal and IMF component respectively;

[0070] N is the total number of data points (i.e., signal length);

[0071] μ x and μ y are the original signal mean and the IMF component mean respectively.

[0072] Calculate the covariance: Calculate the covariance of the original signal and the IMF component:

[0073]

[0074] Calculate standard deviation: Calculate the standard deviation of the original signal and the IMF component:

[0075]

[0076] Calculate the Pearson correlation coefficient:

[0077]

[0078] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0079] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for diagnosing underwater EHA hydraulic cylinder leakage fault, characterized in that: include: The original data set of the fault data of the underwater EHA hydraulic cylinder is preprocessed to obtain the training data set and the test data set; Improve the convolutional neural network model by combining the dung beetle optimization algorithm with the test dataset; Processing the test data set through the improved convolutional neural network model to obtain the fault feature classification result of the underwater EHA hydraulic cylinder; Visualization processing is performed on the fault feature classification result to obtain visualized fault feature data.

2. The underwater EHA hydraulic cylinder leakage fault diagnosis method according to claim 1 is characterized in that: The dung beetle optimization algorithm includes: Initialize the dung beetle algorithm through the population of Chebyshev chaotic map; The rolling dung beetle position update formula of the dung beetle algorithm is replaced by a curve adaptive golden sine strategy; The reproduction and foraging position update formula of the dung beetle algorithm are improved by Levy flight of Cauchy-t mutation strategy.

3. The underwater EHA hydraulic cylinder leakage fault diagnosis method according to claim 1 is characterized in that: The method of improving the convolutional neural network model by using the dung beetle optimization algorithm includes: Introducing the dung beetle optimization algorithm before training the convolutional neural network model; Optimizing the hyperparameters of the convolutional neural network model by using the dung beetle optimization algorithm; Encoding the optimized hyperparameters into individual parameters; The classification error of the convolutional neural network model under each combination of the individual parameters is used as the individual fitness value for parameter optimization.

4. The underwater EHA hydraulic cylinder leakage fault diagnosis method according to claim 1 is characterized in that: The preprocessing of the original data set of the fault data of the underwater EHA hydraulic cylinder to obtain the training data set and the test data set includes: The original data set is preprocessed by adaptive variable mode decomposition.

5. The underwater EHA hydraulic cylinder leakage fault diagnosis method according to claim 1, characterized in that: The visualizing the fault feature classification result to obtain the visualized fault feature data comprises: The fault feature classification results are visualized using the t-SNE manifold learning method.

6. The underwater EHA hydraulic cylinder leakage fault diagnosis method according to claim 2 is characterized in that: The population initialization of the dung beetle algorithm through the Chebyshev chaos map includes: Determining parameters of the Chebyshev chaotic map; Generating initial positions of individuals in the population through the Chebyshev chaotic map; Normalizing the individuals of the population that are not within the target range; Performing a deduplication operation on the individuals of the population; After the population is initialized, the fitness of the individuals in the population is evaluated.

7. The method for diagnosing underwater EHA hydraulic cylinder leakage fault according to claim 4, characterized in that: The preprocessing of the original data set by adaptive variable mode decomposition includes: Decomposing the original data of the original data set by adaptive variable mode decomposition to obtain IMF components; Calculate the correlation coefficient r between the IMF component and the original signal; The IMF components are screened by the correlation coefficient r to perform data reconstruction.

Images