Gearbox fault diagnosis method and device under variable working conditions

Abstract

The present application relates to a method and device for diagnosing faults of a gearbox under variable working conditions. The method comprises: obtaining sample data corresponding to different fault types of the gearbox under variable working conditions; constructing a multilayer perceptron that introduces a variable working condition correction factor; wherein the multilayer perceptron uses the Harris Eagle optimization algorithm to find a target parameter combination including the number of hidden nodes, the learning rate, and the variable working condition correction factor; and using the sample data to train the multilayer perceptron to implement fault diagnosis of the gearbox under variable working conditions based on the trained multilayer perceptron. The Harris Eagle optimization algorithm is used to optimize the parameters of the multilayer perceptron to improve the ability of the multilayer perceptron to capture periodic features in the sample data of the learning sample, and to improve the accuracy of the fault diagnosis model. At the same time, a variable working condition correction factor is introduced into the multilayer perceptron, and the parameters of the multilayer perceptron can be dynamically adjusted according to different working conditions to improve the adaptability and generalization ability of the multilayer perceptron.

Description

Technical Field

[0001] The present application relates to the technical field of mechanical equipment fault diagnosis, and in particular to a method and device for diagnosing gearbox faults under variable operating conditions. Background Art

[0002] As a key transmission component in industrial equipment, the performance of the gearbox under variable operating conditions is crucial to the reliability of the entire system. However, during operation, the gearbox may fail due to fatigue, wear or overload. If these failures are not diagnosed and handled in a timely manner, they may cause equipment performance degradation or even complete failure. In the field of gearbox fault diagnosis, traditional diagnostic methods are often limited by the complexity and uncertainty brought about by changes in operating conditions, making it difficult to accurately diagnose faults. Summary of the invention

[0003] Based on this, it is necessary to provide a gearbox fault diagnosis method and device under variable working conditions that can accurately diagnose the gear fault type in response to the above technical problems.

[0004] In a first aspect, the present application provides a method for diagnosing a gearbox fault under variable operating conditions. The method comprises:

[0005] Obtain sample data corresponding to different gearbox fault types under variable working conditions;

[0006] A multilayer perceptron with variable operating condition correction factor is constructed; the multilayer perceptron uses the Harris Eagle optimization algorithm to find the target parameter combination including the number of hidden nodes, learning rate and variable operating condition correction factor;

[0007] The sample data is used to train a multilayer perceptron, so as to realize the fault diagnosis of the gearbox under variable working conditions based on the trained multilayer perceptron.

[0008] In one embodiment, obtaining sample data corresponding to different fault types of the gearbox under variable working conditions includes:

[0009] For each fault type, collect the vibration signal of the gearbox under variable working conditions;

[0010] The root mean square-mean method is used to preprocess the vibration signal to obtain the characteristic quantity distribution data corresponding to each fault type;

[0011] Linear interpolation is used to expand the feature quantity distribution data to obtain sample data.

[0012] In one embodiment, preprocessing the vibration signal using the root mean square-mean method includes:

[0013] Divide the vibration signal into several groups, and obtain the corresponding root mean square value of each group of vibration signals;

[0014] The root mean square values ​​are divided into several groups, and the characteristic quantity distribution data is obtained according to the corresponding mean of the root mean square values ​​of each group.

[0015] In one embodiment, constructing a multilayer perceptron that introduces a variable operating condition correction factor includes:

[0016] The Harris Eagle optimization algorithm is used to initialize the number of hidden nodes, learning rate and variable working condition correction factor;

[0017] The weight parameters, weight decay and regularization parameters in the multi-layer perceptron are adjusted using variable operating condition correction factors;

[0018] The Harris Eagle optimization algorithm is used to iteratively update the number of hidden nodes, learning rate and variable operating condition correction factor until the preset number of iterations is reached; the number of hidden nodes, learning rate and variable operating condition correction factor corresponding to the time when the iteration stops are obtained as the target parameter combination of the multilayer perceptron.

[0019] In one embodiment, training a multilayer perceptron using sample data includes:

[0020] The number of hidden nodes, learning rate, variable operating condition correction factor, and sample data generated during the iteration of the Harris Eagle optimization algorithm are input into the multilayer perceptron for forward propagation to obtain the prediction output for different fault types; the loss function value is calculated based on the prediction output so that the Harris Eagle optimization algorithm can update the number of hidden nodes, learning rate, and variable operating condition correction factor based on the loss function value.

[0021] In one embodiment, the loss function value is calculated using a cross entropy loss function.

[0022] In a second aspect, the present application also provides a gearbox fault diagnosis device under variable working conditions. The device comprises:

[0023] A data processing module, used to obtain sample data corresponding to different fault types of the gearbox under variable working conditions;

[0024] A model building module is used to build a multi-layer perceptron that introduces a variable operating condition correction factor; wherein the multi-layer perceptron uses the Harris Eagle optimization algorithm to find a target parameter combination including the number of hidden nodes, the learning rate, and the variable operating condition correction factor;

[0025] The model training module is used to train a multi-layer perceptron using sample data, so as to realize fault diagnosis of the gearbox under variable working conditions based on the trained multi-layer perceptron.

[0026] In a third aspect, the present application further provides a computer device, which includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps in the above-mentioned gearbox fault diagnosis method under variable working conditions when executing the computer program.

[0027] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which implements the steps in the above-mentioned gearbox fault diagnosis method under variable working conditions when the computer program is executed by a processor.

[0028] In a fifth aspect, the present application further provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, the steps in the above-mentioned gearbox fault diagnosis method under variable working conditions are implemented.

[0029] The above-mentioned method and device for diagnosing gearbox faults under variable working conditions obtain sample data corresponding to different fault types of gearbox under variable working conditions; construct a multilayer perceptron that introduces variable working condition correction factors; wherein the multilayer perceptron uses the Harris Eagle optimization algorithm to find a target parameter combination including the number of hidden nodes, the learning rate, and the variable working condition correction factor; the multilayer perceptron is trained using sample data to implement fault diagnosis of the gearbox under variable working conditions based on the trained multilayer perceptron. The Harris Eagle optimization algorithm is used to optimize the parameters of the multilayer perceptron to improve the ability of the multilayer perceptron to capture periodic features in the sample data and improve the accuracy of the fault diagnosis model. At the same time, the variable working condition correction factor is introduced into the multilayer perceptron, and the parameters of the multilayer perceptron can be dynamically adjusted according to different working conditions to improve the adaptability and generalization ability of the multilayer perceptron. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 It is a schematic flow chart of a method for diagnosing a gearbox fault under variable working conditions in one embodiment;

[0031] Figure 2 It is a flowchart of a gearbox fault diagnosis method under variable working conditions in another embodiment. DETAILED DESCRIPTION

[0032] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0033] The present application embodiment provides a gearbox fault diagnosis method under variable working conditions, such as Figure 1 As shown, the following steps are included:

[0034] Step 102, obtaining sample data corresponding to different fault types of the gearbox under variable working conditions.

[0035] Among them, the types of gearbox failures include bearing failure, gear wear, tooth pitting, broken teeth, foreign matter intrusion, etc. For gearboxes, variable operating conditions are mainly reflected in changes in speed and load. In actual engineering, the gearboxes used in many large rotating mechanical equipment do not operate under steady-state conditions of constant speed and constant load, but under variable speed and variable load conditions. This variable operating condition may cause non-periodic changes in the vibration signal of the gearbox, increasing the difficulty of fault diagnosis.

[0036] Construct different fault scenarios, collect the corresponding vibration changes of the gearbox under variable working conditions through sensors and other collection equipment, and obtain sample data corresponding to different fault types after data processing.

[0037] Step 104, constructing a multilayer perceptron that introduces a variable operating condition correction factor; wherein the multilayer perceptron uses the Harris Eagle optimization algorithm to find a target parameter combination including the number of hidden nodes, the learning rate, and the variable operating condition correction factor.

[0038] Multilayer Perceptron (MLP) is a feedforward artificial neural network model that maps multiple input nodes to one or more output nodes. This model contains at least three layers: input layer, output layer, and at least one hidden layer. These layers are composed entirely of nodes, and each node is connected to all nodes in the next layer. Each connection has a weight value, which is adjusted during the training process to minimize the prediction error of the model.

[0039] Harris Hawk Optimization (HHO) is used to solve various optimization problems, including continuous and discrete optimization tasks. Inspired by the predation behavior of the Harris Hawk, HHO imitates the three main behaviors of the Harris Hawk during the predation process: exploration phase, encirclement phase, and attack phase. Among them, the eagle in the exploration phase searches for prey in a wide area, which corresponds to the global search in the optimization process, and the purpose is to explore the solution space to find promising areas. Once the eagle in the encirclement phase finds the prey, it will gradually narrow the encirclement and approach the target more closely, which is equivalent to the local search in optimization, aiming to refine the solution and improve its quality. The eagle in the attack phase will launch an attack to capture the prey, which corresponds to convergence to the optimal solution in optimization.

[0040] The role of the variable operating condition correction factor is to dynamically adjust the weight attenuation and regularization parameters to balance the learning ability and generalization ability of the model and avoid overfitting; the weight parameters can also be adjusted to enhance the model's ability to capture these periodic features, which is crucial for accurate diagnosis of gear faults.

[0041] The Harris Eagle optimization algorithm is used to find the target parameter combination including the number of hidden nodes, learning rate and variable operating condition correction factor to improve the prediction accuracy and stability of the model.

[0042] Step 106, using the sample data to train a multi-layer perceptron, so as to implement fault diagnosis of the gearbox under variable working conditions based on the trained multi-layer perceptron.

[0043] The sample data is divided into a training set and a test set. The training set is used to adjust the parameters of the multilayer perceptron. Through repeated iterations, the prediction performance of the multilayer perceptron on the training set is improved. The performance of the model is then evaluated using the test set. If the evaluation meets expectations, the training of the multilayer perceptron is completed and can be used for gearbox fault diagnosis.

[0044] In one embodiment, step 102 includes: for each fault type, collecting the vibration signal of the gearbox under variable working conditions; preprocessing the vibration signal using the root mean square-mean method to obtain characteristic quantity distribution data corresponding to each fault type; and expanding the characteristic quantity distribution data using linear interpolation to obtain sample data.

[0045] like Figure 1 As shown in the figure, after collecting the vibration signal of the gearbox under variable working conditions, the vibration signal is divided into characteristic quantity distribution data of various fault types through the root mean square-mean preprocessing method. Then, linear interpolation is used to expand the fault characteristic distribution data to fill the missing values ​​in the data set and smooth the noise and outliers in the data.

[0046] In one embodiment, the vibration signal is preprocessed using the root mean square-mean method, including: dividing the vibration signal into several groups, and obtaining the root mean square value corresponding to each group of vibration signals; dividing the root mean square value into several groups, and obtaining characteristic quantity distribution data according to the mean corresponding to each group of root mean square values.

[0047] Taking the vibration signal with a sampling frequency of 12.8kHz as an example, 10s of vibration data are selected for each fault type under variable working conditions, totaling 128,000 data points, and these data points are divided into several groups. If each group has 200 data points, the root mean square value of the 200 data points in the same group is calculated, and the distribution of vibration characteristic quantities under a single gear fault state is calculated. There are 640 data for the root mean square value of each gear fault.

[0048] The calculation formula of the root mean square value RMS is:

[0049]

[0050] Where RMS stands for root mean square, N is the number of data points, and x i is the value of each data point.

[0051] Then, the RMS values ​​of each fault are grouped and averaged. If there are 5 RMS values ​​in each group, an average value is calculated for the 5 RMS values ​​in the same group, and the resulting data is the fault characteristic quantity distribution data. This data can provide a more stable and smoother measure of the overall energy level of the vibration signal. This process helps to reduce the impact of random fluctuations in variable operating condition signals and better reflect the periodic characteristics of the vibration signal.

[0052] In one embodiment, the linear interpolation formula is:

[0053]

[0054] Where x0≤x≤x1. The linear interpolation method is used between two known data points to estimate the value of the unknown point (x,y). In this example, there are two known points (x0,y0) and (x1,y1). The linear interpolation formula can be used to calculate the value of y at x.

[0055] In one embodiment, step 104 includes: initializing the number of hidden nodes, learning rate and variable operating condition correction factor using the Harris Hawk optimization algorithm; adjusting the weight parameters, weight decay and regularization parameters in the multilayer perceptron using the variable operating condition correction factor; iteratively updating the number of hidden nodes, learning rate and variable operating condition correction factor using the Harris Hawk optimization algorithm until a preset number of iterations is reached; and obtaining the corresponding number of hidden nodes, learning rate and variable operating condition correction factor when the iteration stops as the target parameter combination of the multilayer perceptron.

[0056] The specific steps of the Harris Hawk optimization algorithm are: (1) Define a fitness function that can evaluate the performance of a given parameter combination. In this embodiment, it is the loss function, and the smaller its value, the better the parameter combination. (2) Initialize a set of candidate solutions (i.e., parameter combinations), each candidate solution represents a possible parameter setting. (3) Iterate and update the population by simulating the hunting behavior of the Harris Hawk. In each iteration, the algorithm evaluates the fitness of each candidate solution and determines which candidate solutions are better based on the fitness. (4) Guide the population to move in a better direction based on the position of the current optimal solution, which includes exploring new areas of the parameter space and exploiting currently known good areas. (5) After a certain number of iterations, the algorithm will converge to one or several optimal parameter combinations.

[0057] like Figure 2As shown in , after constructing the multilayer perceptron model, the parameters are initialized first. By randomly initializing a group of Harris Hawks, each Hawk represents a potential hyperparameter combination, that is, a hidden layer node number and a learning rate value. The initial value of the variable working condition correction factor can be 1. Among them, the number of hidden layer nodes and the learning rate are mapped to the position in the search space. The fitness (i.e., the loss function value) of the multilayer perceptron in the initial stage is calculated, and a better fitness is searched based on the fitness to adjust the number of hidden layer nodes, the learning rate, and the variable working condition correction factor, and the escape energy factor is updated after adjusting the parameter combination. After the exploration is completed, this iteration is completed to determine whether the iteration has reached the maximum number of iterations (i.e., the preset number of iterations). If not, the number of hidden layer nodes, the learning rate, and the variable working condition correction factor will continue to be iteratively updated; if so, the number of hidden layer nodes, the learning rate, and the variable working condition correction factor at this time are output as the target parameter combination, which can be regarded as the optimal parameter combination.

[0058] Among them, the role of the variable operating condition correction factor in the multi-layer perceptron includes adjusting the weight parameters, weight attenuation and regularization parameters in the multi-layer perceptron.

[0059] The L2 regularization technology is used in the multi-layer perceptron to suppress the overfitting of the multi-layer perceptron. Regarding the adjustment of the regularization parameter by the variable working condition correction factor, please refer to the following formula:

[0060]

[0061] In the formula, L represents the loss function, L0 is the original loss function, w represents the weight parameter, k is the variable working condition correction factor, λ is the original regularization parameter, and kλ is the adjusted regularization parameter.

[0062] It can be seen from the above formula that the variable operating condition correction factor k can also be used to adjust the loss function of the multilayer perceptron.

[0063] The weight parameters in the multilayer perceptron are updated using L2 regularization and variable operating condition correction factor k to minimize the prediction error of the network. The weight parameter update formula is:

[0064]

[0065] Where w is the weight parameter, α represents the learning rate, L0 is the original loss function, and kλ is the adjusted regularization parameter.

[0066] The formula for adjusting the weight attenuation by the variable operating condition correction factor is:

[0067]

[0068] Where w is the weight parameter, k is the variable condition correction factor, β is the original attenuation factor, α represents the learning rate, and L0 is the original loss function.

[0069] By adding a variable operating condition correction factor to the weight attenuation formula, the intensity of weight attenuation can be dynamically adjusted according to different operating conditions, thereby improving the adaptability and generalization ability of the model.

[0070] In this embodiment, the weight parameters all refer to the weight coefficients connecting neurons in different layers. The difference is that the weight parameters in the regularization parameter adjustment formula are considered in the regularization term, emphasizing the impact of regularization on the loss function. The weight parameters in the weight parameter update formula are adjusted during the parameter update process, emphasizing how to update the weights through gradient descent and regularization terms. The weight parameters in the weight decay adjustment formula are also adjusted during the parameter update process, but the adjustment method is achieved through weight decay, emphasizing the direct punishment of the weight parameters to prevent overfitting.

[0071] In this embodiment, the variable working condition correction factor can be dynamically adjusted according to the changes in the actual working conditions to adapt to different working environments and conditions. Its specific functions and characteristics are as follows:

[0072] (1) Adjust the regularization parameters in the multi-layer perceptron to adapt to different working conditions. By adjusting the regularization parameters, the model's sensitivity to noise during training can be reduced and the model's generalization ability can be improved;

[0073] (2) Dynamically adjust the weight decay strength of the model. Under different working conditions, by adjusting the weight decay, the learning ability and generalization ability of the multi-layer perceptron can be balanced to avoid overfitting;

[0074] (3) Under variable working conditions, the vibration signal of the gearbox fault has periodic non-stationary characteristics. The variable working condition correction factor enhances the multilayer perceptron's ability to capture these periodic features and its ability to adapt to periodic non-stationary signals by adjusting the multilayer perceptron weight parameters.

[0075] According to the performance of the multilayer perceptron in actual applications, the value of the variable operating condition correction factor is adaptively adjusted in real time to ensure that the multilayer perceptron can maintain optimal performance under different operating conditions.

[0076] In one embodiment, each hyperparameter combination provided in the Harris Eagle algorithm is forward propagated using a multilayer perceptron to calculate the predicted output for different faults. The loss function value is calculated based on the predicted output and the true label. The loss function value can be calculated using a cross entropy loss (CrossEntropyLoss) function.

[0077] CrossEntropyLoss combines the calculations of LogSoftmax and NLLLoss. The LogSoftmax function can reduce overflow problems in numerical calculations, while NLLLoss calculates losses. If the logarithm output by LogSoftmax is very small, directly calculating the negative logarithm may cause numerical underflow. The CrossEntropyLoss function can avoid the above problems and reduce the number of layers that need to be added manually. The gradient can be calculated directly through automatic differentiation. In summary, the CrossEntropyLoss function can be used to classify different gear faults. The calculation expression of the CrossEntropyLoss function is:

[0078]

[0079] In the formula, L represents the loss function value, N is the number of samples in the training set, and K is the number of categories. k is an indicator variable. If sample i belongs to category k, then y ik =1, otherwise y ik =0.

[0080] is the probability that the model predicts that sample i belongs to category k, calculated as:

[0081]

[0082] In the formula, y ik represents the indicator variable, Z ik is the original output of the model for the i-th sample belonging to the j-th category.

[0083] In summary, the present invention uses the root mean square-mean statistical feature extraction method to divide the gear fault vibration signal into different fault frequency band data, and reduces the impact of random variable working conditions on the vibration signal by capturing the statistical characteristics of the fault data. The extracted vibration feature quantity is subjected to linear interpolation processing to enhance the density and integrity of the data set. The Harris Eagle optimization algorithm is used to optimize the multilayer perceptron that introduces the variable working condition correction factor, and the hyperparameters of the multilayer perceptron under variable working conditions are automatically optimized to enhance the periodic characteristics of the signal. A fault diagnosis classifier based on the Harris Eagle optimizer to optimize the multilayer perceptron neural network is constructed, thereby realizing the fault diagnosis of the gear.

[0084] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0085] Based on the same inventive concept, the embodiment of the present application also provides a gearbox fault diagnosis device under variable working conditions for implementing the gearbox fault diagnosis method under variable working conditions involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more embodiments of the gearbox fault diagnosis device under variable working conditions provided below can refer to the limitations of the gearbox fault diagnosis method under variable working conditions above, and will not be repeated here.

[0086] In one embodiment, a fault diagnosis device for a gearbox under variable working conditions is provided, comprising:

[0087] A data processing module, used to obtain sample data corresponding to different fault types of the gearbox under variable working conditions;

[0088] A model building module is used to build a multi-layer perceptron that introduces a variable operating condition correction factor; wherein the multi-layer perceptron uses the Harris Eagle optimization algorithm to find a target parameter combination including the number of hidden nodes, the learning rate, and the variable operating condition correction factor;

[0089] The model training module is used to train a multi-layer perceptron using sample data, so as to realize fault diagnosis of the gearbox under variable working conditions based on the trained multi-layer perceptron.

[0090] In one embodiment, the data processing module is also used to: for each fault type, collect the vibration signal of the gearbox under variable working conditions; pre-process the vibration signal using the root mean square-mean method to obtain the characteristic quantity distribution data corresponding to each fault type; and use linear interpolation to expand the characteristic quantity distribution data to obtain sample data.

[0091] In one embodiment, the data processing module is also used to: divide the vibration signal into several groups and obtain the root mean square value corresponding to each group of vibration signals; divide the root mean square value into several groups and obtain characteristic quantity distribution data according to the mean value corresponding to each group of root mean square values.

[0092] In one embodiment, the model building module is also used to: initialize the number of hidden nodes, learning rate and variable operating condition correction factor using the Harris Hawk optimization algorithm; adjust the weight parameters, weight decay and regularization parameters in the multilayer perceptron using the variable operating condition correction factor; iteratively update the number of hidden nodes, learning rate and variable operating condition correction factor using the Harris Hawk optimization algorithm until a preset number of iterations is reached; obtain the corresponding number of hidden nodes, learning rate and variable operating condition correction factor when the iteration stops as the target parameter combination of the multilayer perceptron.

[0093] In one embodiment, the model training module is also used to: input the number of hidden nodes, learning rate and variable operating condition correction factor generated during the iteration of the Harris Hawk optimization algorithm, as well as the sample data into the multi-layer perceptron for forward propagation to obtain the prediction output for different fault types; calculate the loss function value based on the prediction output, so that the Harris Hawk optimization algorithm can update the number of hidden nodes, learning rate and variable operating condition correction factor based on the loss function value.

[0094] In one embodiment, the loss function value is calculated using a cross entropy loss function.

[0095] Each module in the gearbox fault diagnosis device under variable working conditions can be implemented in whole or in part by software, hardware and a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each module.

[0096] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in all the above method embodiments when executing the computer program.

[0097] In one 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 steps in all the above method embodiments are implemented.

[0098] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in all the above method embodiments when executed by a processor.

[0099] 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 laws, regulations and standards of relevant countries and regions.

[0100] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed 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 the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can 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 can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited thereto. The processors involved in each embodiment provided in this application may be general-purpose processors, central processing units, graphics processors, digital signal processors, programmable logic devices, data processing logic devices based on quantum computing, etc., but are not limited thereto.

[0101] The technical features of the above embodiments may 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.

[0102] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for diagnosing gearbox faults under variable operating conditions, characterized in that: The method comprises: Obtain sample data corresponding to different gearbox fault types under variable working conditions; Constructing a multilayer perceptron that introduces a variable operating condition correction factor; wherein the multilayer perceptron uses a Harris Eagle optimization algorithm to find a target parameter combination including the number of hidden nodes, a learning rate, and the variable operating condition correction factor; The multilayer perceptron is trained using the sample data to implement fault diagnosis of the gearbox under the variable working condition based on the trained multilayer perceptron.

2. The method according to claim 1, characterized in that: The obtaining of sample data corresponding to different gearbox fault types under variable working conditions includes: For each of the fault types, collecting vibration signals of the gearbox under variable working conditions; The vibration signal is preprocessed by using a root mean square-mean method to obtain characteristic quantity distribution data corresponding to each fault type; The characteristic quantity distribution data is expanded by using linear interpolation to obtain the sample data.

3. The method according to claim 2, characterized in that The preprocessing of the vibration signal by using the root mean square-mean method includes: Dividing the vibration signal into several groups, and obtaining the root mean square value corresponding to the vibration signal of each group; The root mean square values ​​are divided into a plurality of groups, and the characteristic quantity distribution data is obtained according to the corresponding mean values ​​of the root mean square values ​​of each group.

4. The method according to claim 1, characterized in that The multi-layer perceptron for constructing a variable operating condition correction factor includes: Initializing the number of hidden nodes, the learning rate and the variable operating condition correction factor using the Harris Eagle optimization algorithm; Using the variable operating condition correction factor to adjust the weight parameters, weight attenuation and regularization parameters in the multi-layer perceptron; The Harris Eagle optimization algorithm is used to iteratively update the number of hidden nodes, the learning rate and the variable operating condition correction factor until a preset number of iterations is reached; the number of hidden nodes, the learning rate and the variable operating condition correction factor corresponding to when the iteration stops are obtained as the target parameter combination of the multilayer perceptron.

5. The method according to claim 4, characterized in that The using the sample data to train the multi-layer perceptron comprises: The number of hidden nodes, the learning rate and the variable operating condition correction factor generated during the iteration of the Harris Hawk optimization algorithm, as well as the sample data are input into the multilayer perceptron for forward propagation to obtain prediction outputs for different fault types; the loss function value is calculated based on the prediction output, so that the Harris Hawk optimization algorithm can update the number of hidden nodes, the learning rate and the variable operating condition correction factor based on the loss function value.

6. The method according to claim 5, characterized in that The loss function value is calculated using a cross entropy loss function.

7. A gearbox fault diagnosis device under variable working conditions, characterized in that: The device comprises: A data processing module, used to obtain sample data corresponding to different fault types of the gearbox under variable working conditions; A model building module is used to build a multi-layer perceptron that introduces a variable operating condition correction factor; wherein the multi-layer perceptron uses a Harris Eagle optimization algorithm to find a target parameter combination including the number of hidden nodes, a learning rate, and the variable operating condition correction factor; The model training module is used to train the multi-layer perceptron using the sample data, so as to implement fault diagnosis of the gearbox under the variable working condition based on the trained multi-layer perceptron.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images