A wind turbine gearbox gear fault diagnosis method and system

Abstract

The application provides a wind turbine gearbox gear fault diagnosis method and system, which comprises the following steps: collecting and preprocessing the vibration signal of the gearbox gear, constructing a gear fault feature sample set, and dividing the sample set into a training set and a test set; constructing an improved bald eagle search algorithm; constructing a deep belief network base model; obtaining the classification accuracy of the deep belief network base model on the training set as the fitness function of the improved bald eagle search algorithm, solving the optimal value of the fitness function to obtain the optimal hyperparameter combination of the deep belief network base model, completing the construction, training and optimization of the fault diagnosis model; preprocessing and feature extraction are performed on the vibration signal of the gearbox gear to be diagnosed, and the trained fault diagnosis model is input, and the gear fault classification result is output, and the fault diagnosis is completed. Through the method, the parameter optimization ability and fault recognition performance of the fault diagnosis model can be improved, and the automatic diagnosis and early warning of the wind turbine gearbox gear fault can be realized.

Description

Technical Field

[0001] This application relates to the field of wind turbine technology, and in particular to a method and system for diagnosing gear faults in wind turbine gearboxes. Background Technology

[0002] Driven by the "dual carbon" goals, wind power, as a clean and renewable energy source, has achieved rapid and large-scale development, with the global cumulative installed capacity exceeding 1 TW. Wind turbines are typically deployed in harsh environments such as remote inland and near-shore areas, and are subject to long-term exposure to variable loads, strong winds and sandstorms, and high humidity, leading to frequent equipment failures. Among these failures, the gearbox, as the core component of the wind turbine's transmission system, ranks among the most prone to failures of any mechanical part of the unit, with gear and bearing failures accounting for over 80% of all gearbox failures. Gearbox failures not only cause prolonged downtime but also account for 25%-30% of the total lifecycle cost of the unit, severely impacting the economic efficiency and operational safety of wind farms.

[0003] Currently, gear fault diagnosis in wind turbine gearboxes is mainly divided into traditional signal processing-based methods, shallow machine learning-based methods, and deep learning-based methods. Traditional signal processing methods rely on manual extraction of fault features, which requires a high level of expertise from technicians and is insufficient in extracting weak fault features under strong noise and variable operating conditions. While shallow machine learning methods such as Support Vector Machine (SVM), Extreme Learning Machine (ELM), and K-Nearest Neighbors (KNN) can achieve fault classification, their generalization ability is weak, and the diagnostic accuracy drops significantly when faced with complex fault modes.

[0004] Deep Belief Networks (DBNs), as classic deep learning models, possess powerful nonlinear feature learning and automatic representation capabilities, eliminating the need for complex manual feature engineering, and are widely used in rotating machinery fault diagnosis. However, the diagnostic performance of DBNs is highly dependent on the setting of key hyperparameters such as the number of hidden layer nodes, the number of training iterations, and the learning rate. In existing technologies, these hyperparameters are mostly set manually by experienced technicians, which not only involves a large amount of debugging work, but also easily leads to problems such as slow model convergence, susceptibility to local optima, and low classification accuracy if the parameter settings are unreasonable.

[0005] To address the hyperparameter optimization problem in deep learning network (DBN), existing research has introduced metaheuristic algorithms such as Particle Swarm Optimization (PSO) and Grey Wolf Optimizer (GWO) for automatic parameter tuning. However, these algorithms suffer from drawbacks such as slow convergence speed, insufficient global optimization capability, and premature convergence. The Bald Eagle Search (BES) algorithm, as a novel metaheuristic algorithm, boasts advantages such as simple structure and strong optimization capability. However, its use of fixed-value position control parameters in the region selection stage fails to meet the dynamic balance requirements of global search and local development during iteration, resulting in insufficient optimization accuracy and a need for improved convergence speed. When directly applied to DBN hyperparameter optimization, it struggles to achieve high-precision diagnosis of gear faults in wind turbine gearboxes. Summary of the Invention

[0006] This application provides a method and system for diagnosing gear faults in wind turbine gearboxes, in order to improve the parameter optimization capability and fault identification performance of the fault diagnosis model, and realize the automated diagnosis and early warning of gear faults in wind turbine gearboxes.

[0007] Firstly, this application provides a method for diagnosing gear faults in a wind turbine gearbox, which is executed by a computing device. The computing device can be understood as a computer or similar device, and is not limited thereto in this application. The method includes:

[0008] Vibration signals of gearbox gears were collected using a scaled-down wind turbine and monitoring system experimental platform. These vibration signals were preprocessed to construct a gear fault feature sample set, which was then divided into a training set and a test set. An improved vulture search algorithm was constructed by introducing adaptively changing position control parameters in the selection region stage of the standard vulture search algorithm, replacing the original fixed parameters. A deep belief network (DBN) basic model was then built. The construction method of the DBN basic model includes determining the number of input layer nodes based on the gear fault feature dimensions of the gear fault feature sample set, and then... The number of output layer nodes in the deep belief network (DBN) basic model is determined by category, and the number of hidden layers in the DBN basic model is determined through pre-experiments to identify the hyperparameter combination to be optimized. The classification accuracy of the DBN basic model on the training set is obtained and used as the fitness function to improve the vulture search algorithm. The optimal value of the fitness function is solved to obtain the optimal hyperparameter combination of the DBN basic model, thus completing the construction, training, and optimization of the fault diagnosis model. The gear vibration signal of the gearbox to be diagnosed is preprocessed and its features are extracted and input into the trained fault diagnosis model to output the gear fault classification result, thus completing the fault diagnosis.

[0009] Through the above methods, this application collects vibration signals from gearbox gears and preprocesses them to obtain a gear fault feature sample set, which is then divided into a training set and a test set to provide basic data for subsequent fault diagnosis. An improved vulture search algorithm is constructed to enhance global optimization accuracy and solution quality, while accelerating the convergence process, reducing computational resource consumption, and achieving enhanced robustness and stability while mitigating random interference. By constructing a deep confidence network basic model, the dependence on large-scale, high-quality labeled samples can be reduced. The fault diagnosis model is used to judge the fault classification results, enabling online status monitoring and fault early warning of wind turbine gearboxes, effectively reducing wind turbine operation and maintenance costs and improving the reliability and safety of unit operation.

[0010] In the aforementioned wind turbine gearbox gear fault diagnosis method, vibration signals of the gearbox gears are collected through the wind turbine scale-down and monitoring system experimental platform, and the vibration signals are preprocessed, including: collecting vibration signals of the gearbox gears through the wind turbine scale-down and monitoring system experimental platform, including collecting vibration signals of the gearbox gears under different speeds and different loads in three states: normal, pitting wear, and broken teeth, and preprocessing the original vibration signals.

[0011] By means of the above method, this application collects vibration signals of gearbox gears under different speeds and loads in three states: normal, pitting wear, and broken teeth, as the initial data signals for fault diagnosis. This ensures that data from multiple states of the gears are fully collected, enabling comprehensive fault diagnosis and improving the reliability of the fault diagnosis results.

[0012] In the aforementioned wind turbine gearbox gear fault diagnosis method, an improved vulture search algorithm is constructed, including:

[0013] In the selection region phase of the standard bald eagle search algorithm, the formula for updating the bald eagle's position is:

[0014] ;

[0015] in, This represents the bald eagle's current optimal position; and These represent the current and previous positions of the j-th vulture, respectively; the average distribution position of the vultures is represented by... Indicates; control parameters random numbers Introducing adaptive position control parameters Replace the fixed parameter ε, The expression is:

[0016] ;

[0017] in, Here are the control parameters for the vulture's position change; t is the current iteration number; T represents the maximum number of iterations; the improved formula for updating the vulture's position during the selected region stage is:

[0018] ;

[0019] By combining the position update rules of the search phase and the dive-hunting phase of the standard bald eagle search algorithm, an improved bald eagle search algorithm was constructed.

[0020] Through the above methods, this application constructs an improved vulture search algorithm, which improves the global optimization accuracy and solution quality, while accelerating the convergence process, reducing computational resource consumption, enhancing robustness and stability, and weakening random interference.

[0021] In the aforementioned wind turbine gearbox gear fault diagnosis method, the basic deep belief network model consists of an input layer, two hidden layers, and an output layer stacked together. The number of nodes in the input layer corresponds to the feature dimension of a single sample; the number of nodes in the output layer corresponds to the number of gear fault categories; and the hyperparameter combination to be optimized is... ,in This represents the number of nodes in the first hidden layer. This represents the number of nodes in the second hidden layer. The number of training iterations for the model. This is the learning rate.

[0022] By constructing a deep confidence network basic model, this application can reduce the dependence on large-scale high-quality labeled samples.

[0023] In the aforementioned wind turbine gearbox gear fault diagnosis method, the construction, training, and optimization of the fault diagnosis model are completed, including:

[0024] Initialize the parameters of the improved bald eagle search algorithm; the algorithm parameters include at least one of the following: population size, maximum number of iterations, upper and lower bounds of independent variables, and search space dimension; wherein, the search space dimension is consistent with the number of hyperparameters to be optimized in the deep belief network base model, and the upper and lower bounds of independent variables correspond to the value range of each hyperparameter; using the classification accuracy of the deep belief network base model on the training set as the fitness function of the improved bald eagle search algorithm, calculate the fitness value corresponding to each bald eagle in the initial population, and record the initial optimal fitness value and the corresponding bald eagle position; enter the iterative optimization process, through the selection area stage, search stage, and swoop-down hunting stage of the improved bald eagle search algorithm... The system updates the bald eagle population locations and recalculates the fitness value for each location. If the fitness value of the new location is better than the saved optimal fitness value, the optimal fitness value and the corresponding bald eagle location are updated; otherwise, the original values ​​are retained. The system checks if the maximum number of iterations has been reached. If not, it returns to the iteration optimization process to continue iterating; if the maximum number of iterations has been reached, the iteration terminates, and the bald eagle location corresponding to the optimal fitness value is output as the optimal combination of hyperparameters. The parameters of the deep belief network basic model are configured using the optimal combination of hyperparameters. Unsupervised pre-training and supervised fine-tuning of the model are completed using the training set, and the model performance is verified using the test set. Finally, the trained fault diagnosis model is obtained.

[0025] Through the above methods, this application uses a fault diagnosis model to determine the fault classification results, thereby achieving online status monitoring and fault early warning of wind turbine gearboxes, effectively reducing the operation and maintenance costs of wind turbines, and improving the reliability and safety of unit operation.

[0026] In the aforementioned wind turbine gearbox fault diagnosis method, the fitness function expression is:

[0027] ;

[0028] in, For fitness value, The classification accuracy of the deep belief network base model on the training set. The calculation formula is:

[0029] ;

[0030] Where TP is the number of positive samples correctly classified, TN is the number of negative samples correctly classified, FP is the number of negative samples misclassified as positive samples, and FN is the number of positive samples misclassified as negative samples.

[0031] In the aforementioned wind turbine gearbox gear fault diagnosis method, the vibration signal of the gearbox gear to be diagnosed is preprocessed and its features are extracted, and then input into the trained fault diagnosis model. The output gear fault classification results include:

[0032] Feature extraction includes at least one of the following: data cleaning, outlier removal, and normalization; gear fault classification results include at least one of the following: gear health status, fault type, fault confidence, and fault warning information generated based on the classification results.

[0033] Secondly, this application provides a wind turbine gearbox gear fault diagnosis system for performing the method in the first aspect of this application, including: a preprocessing module, an algorithm improvement module, a model building module, a training optimization module, and a fault diagnosis module;

[0034] The system comprises several modules: a preprocessing module, which collects vibration signals from gearbox gears using a scaled-down wind turbine and monitoring system experimental platform; a preprocessing module to construct a gear fault feature sample set, which is then divided into a training set and a test set; an algorithm improvement module, which introduces adaptively changing position control parameters to replace the original fixed parameters in the selection region stage of the standard vulture search algorithm, thus constructing an improved vulture search algorithm; and a model building module, which constructs a deep belief network (DBN) basic model. The DBN basic model construction method includes determining the number of input layer nodes based on the gear fault feature dimensions of the gear fault feature sample set. The number of output layer nodes of the deep belief network (DBN) basic model is determined based on the gear fault category, and the number of hidden layers of the DBN basic model is determined through pre-experiments to determine the hyperparameter combination to be optimized. The training optimization module is used to obtain the classification accuracy of the DBN basic model on the training set, which is used as the fitness function to improve the vulture search algorithm. The optimal value of the fitness function is solved to obtain the optimal hyperparameter combination of the DBN basic model, thus completing the construction, training and optimization of the fault diagnosis model. The fault diagnosis module is used to preprocess and extract features from the gear vibration signal of the gearbox to be diagnosed and input it into the trained fault diagnosis model, outputting the gear fault classification result to complete the fault diagnosis.

[0035] Thirdly, this application also provides a computing device, comprising: a memory for storing program instructions; and a processor for calling the program instructions stored in the memory and executing the method described in the first aspect according to the obtained program instructions.

[0036] Fourthly, this application also provides a computer-readable storage medium storing computer-readable instructions, which, when read and executed by a computer, implement the method described in the first aspect.

[0037] Fifthly, this application provides a computer program product including a computer program executable by a computer device, which, when run on the computer device, causes the computer device to perform the method described in the first aspect.

[0038] Beneficial effects: This application collects vibration signals from gearbox gears and preprocesses them to obtain a gear fault feature sample set, which is then divided into a training set and a test set, providing basic data for subsequent fault diagnosis; it constructs an improved vulture search algorithm to enhance global optimization accuracy and solution quality, while accelerating the convergence process, reducing computational resource consumption, and achieving enhanced robustness and stability while mitigating random interference; by constructing a deep confidence network basic model, the dependence on large-scale high-quality labeled samples can be reduced; the fault diagnosis model is used to judge the fault classification results, with a high degree of automation in the process, requiring no complex manual feature engineering intervention. It is suitable for gear fault simulation diagnosis in laboratory environments and can also be directly connected to the actual operating data of wind farms to realize online status monitoring and fault early warning of wind turbine gearboxes, effectively reducing the operation and maintenance costs of wind turbines and improving the reliability and safety of unit operation. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 A schematic flowchart illustrating a method for diagnosing gear faults in a wind turbine gearbox, provided in an embodiment of this application;

[0041] Figure 2 A schematic diagram of a gear fault diagnosis system for a wind turbine gearbox provided in this application embodiment;

[0042] Figure 3 This is a schematic diagram of the structure of a computing device provided in an embodiment of this application. Detailed Implementation

[0043] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0044] In the following embodiments of the present application, "and / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B may represent: the presence of A alone, the co-presence of A and B, and the presence of B alone, where A and B may be singular or plural. The character " / " generally indicates that the associated objects before and after are in an "or" relationship. "At least one (piece) of the following" or similar expressions refers to any combination of these items, including any combination of a single item (piece) or a plurality of items (pieces). For example, at least one (piece) of a, b or c may represent: a, b, c, a-b, a-c, b-c, or a-b-c, where a, b, c may be a single one or a plurality of ones. The singular expressions "a", "an", "the", "the above", "said" and "this" are intended to also cover expressions such as "one or more", unless the context clearly indicates otherwise. Also, unless otherwise stated, ordinal numbers such as "first" and "second" mentioned in the embodiments of the present application are used to distinguish a plurality of objects, and are not intended to limit the order, time sequence, priority or importance degree of the plurality of objects.

[0045] References in the description of the present application to "one embodiment" or "some embodiments" and the like mean that a particular feature, structure or characteristic described in connection with the embodiment is included in one or more embodiments of the present application. Thus, the appearances of the phrases "in one embodiment", "in some embodiments", "in other embodiments", "in still other embodiments" and the like in various places in this specification are not necessarily all referring to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in another manner. The terms "comprise", "include", "have" and their variants all mean "including but not limited to", unless otherwise specifically emphasized in another manner.

[0046] Embodiment 1

[0047] Embodiment 1 of the present application provides a gear fault diagnosis method for a wind turbine gearbox, the method is executed by a computing device, which can be understood as a device such as a computer, and the present application does not limit this herein. The process of the method is as follows Figure 1 as shown, comprising:

[0048] Step 101: Collect vibration signals of the gears in the gearbox through the scaled wind turbine and monitoring system experimental platform, preprocess the vibration signals, construct a gear fault feature sample set, and divide the gear fault feature sample set into a training set and a test set.

[0049] In one possible implementation, vibration signals of gearbox gears are collected through a wind turbine scale-down and monitoring system experimental platform. This includes collecting vibration signals of gearbox gears under different speeds and loads in three states: normal, pitting wear, and broken teeth. The original vibration signals are then preprocessed.

[0050] In one possible implementation, the training set and the test set are divided in an 8:2 ratio, with samples in the normal state, pitting wear state, and broken tooth state evenly distributed in both sets.

[0051] Step 102: In the selection area stage of the standard vulture search algorithm, an adaptively changing position control parameter is introduced to replace the original fixed parameter, thereby constructing an improved vulture search algorithm.

[0052] Specifically, in the selection region phase of the Standard Bald Eagle Search (BES) algorithm, the formula for updating the bald eagle's position is:

[0053] ;

[0054] in, This represents the bald eagle's current optimal position; and These represent the current and previous positions of the j-th vulture, respectively; the average distribution position of the vultures is represented by... Indicates; control parameters random numbers ;

[0055] Introducing adaptive position control parameters Replace the fixed parameter ε, The expression is:

[0056] ;

[0057] in, Here are the control parameters for the vulture's position change; t is the current iteration number; T represents the maximum number of iterations; the improved formula for updating the vulture's position during the selected region stage is:

[0058] ;

[0059] By combining the position update rules of the search phase and the dive-hunting phase of the standard Bald Eagle Search algorithm, an improved Bald Eagle Search (IBES) algorithm was constructed.

[0060] Step 103: Construct a deep belief network basic model; wherein, the method for constructing the deep belief network basic model includes determining the number of input layer nodes of the deep belief network basic model based on the gear fault feature dimension of the gear fault feature sample set, determining the number of output layer nodes of the deep belief network basic model based on the gear fault category, determining the number of hidden layers of the deep belief network basic model through pre-experiments, and determining the hyperparameter combination to be optimized.

[0061] Specifically, the basic model of the deep belief network consists of an input layer, two hidden layers, and an output layer stacked together. The number of nodes in the input layer is consistent with the feature dimension of a single sample, which is 26 nodes. The number of nodes in the output layer is consistent with the number of gear fault categories, which is 3 nodes, corresponding to the three states of normal, pitting wear, and broken teeth, respectively.

[0062] The hyperparameter combination to be optimized is ,in This represents the number of nodes in the first hidden layer. This represents the number of nodes in the second hidden layer. The number of training iterations for the model. This is the learning rate.

[0063] Step 104: Obtain the classification accuracy of the deep belief network base model on the training set, and use it as the fitness function of the improved vulture search algorithm. Solve for the optimal value of the fitness function to obtain the optimal hyperparameter combination of the deep belief network base model, thus completing the construction, training, and optimization of the fault diagnosis model. The fault diagnosis model is the IBES-DBN fault diagnosis model, i.e., the joint fault diagnosis model of the improved vulture search algorithm and the deep belief network.

[0064] In one possible implementation, the specific steps for constructing, training, and optimizing the fault diagnosis model include:

[0065] Initialize the parameters of the improved vulture search algorithm. The algorithm parameters include at least one of the following: population size, maximum number of iterations, upper and lower limits of independent variables, and search space dimension. The search space dimension is consistent with the number of hyperparameters to be optimized in the deep belief network base model, and the upper and lower limits of independent variables correspond to the value range of each hyperparameter. For example, initializing the parameters of the improved vulture search algorithm includes setting the population size to 30, the maximum number of iterations to 30, the upper and lower limits of independent variables to [10, 100], which represent the value range of the corresponding hyperparameters, and the search space dimension to 4.

[0066] The fitness function of the improved bald eagle search algorithm is based on the classification accuracy of the deep belief network basic model on the training set. The fitness value of each bald eagle in the initial population is calculated, and the initial optimal fitness value and the corresponding bald eagle position are recorded.

[0067] Entering the iterative optimization process, the location of the bald eagle population is updated sequentially through the selection area stage, search stage, and dive hunting stage of the improved bald eagle search algorithm. The fitness value corresponding to each location is recalculated. If the fitness value of the new location is better than the saved optimal fitness value, the optimal fitness value and the corresponding bald eagle location are updated; otherwise, the original value is kept.

[0068] Determine if the maximum number of iterations has been reached. If not, return to the iterative optimization process to continue iterating. If the maximum number of iterations has been reached, terminate the iteration and output the position of the vulture corresponding to the optimal fitness value as the optimal combination of hyperparameters.

[0069] The parameters of the deep belief network basic model are configured by using the optimal combination of hyperparameters. Unsupervised pre-training and supervised fine-tuning of the model are completed using the training set. The model performance is verified using the test set, and finally the trained fault diagnosis model is obtained.

[0070] The fitness function is expressed as follows:

[0071] ;

[0072] in, For fitness value, The classification accuracy of the deep belief network base model on the training set. The calculation formula is:

[0073] ;

[0074] Where TP is the number of positive samples correctly classified, TN is the number of negative samples correctly classified, FP is the number of negative samples misclassified as positive samples, and FN is the number of positive samples misclassified as negative samples.

[0075] Step 105: Preprocess and extract features from the gear vibration signal of the gearbox to be diagnosed, input it into the trained fault diagnosis model, output the gear fault classification result, and complete the fault diagnosis.

[0076] In one possible implementation, feature extraction includes at least one of data cleaning, outlier removal, and normalization.

[0077] The gear fault classification results include the gear's health status, fault type, fault confidence, and at least one of the following: fault warning information generated based on the classification results.

[0078] Example 2

[0079] This application, in embodiment 2, provides a specific implementation of a method for diagnosing gear faults in a wind turbine gearbox, based on embodiment 1. The method includes the following steps:

[0080] Step 201, Fault vibration signal acquisition and sample set construction;

[0081] For the experimental platform construction, this application adopts a wind turbine scale-down and monitoring system experimental platform. The platform is a single-stage planetary two-stage parallel shaft gearbox with a speed ratio of 50.4, equipped with a 4kW drive motor and an acceleration vibration sensor (sensitivity 46.5mV / m・s⁻², frequency response range 0.1~10000Hz). The vibration sensor is installed at the high-speed shaft of the gearbox to collect signals.

[0082] Specifically, the vibration signal acquisition parameters are as follows: sampling frequency 12kHz, acquisition conditions including input speeds of 780r / min, 860r / min, and 1390r / min, and loads of 0kW and 0.106kW.

[0083] Fault simulation and signal acquisition were performed by installing normal gears, pitted wear gears, and broken gears at the first-stage parallel shaft gear, setting up three sets of operating conditions:

[0084] ① Rotation speed 780 r / min, load 0.106 kW;

[0085] ② Rotation speed 860 r / min, load 0 kW;

[0086] ③ Rotation speed 1390 r / min, load 0 kW;

[0087] The sampling frequency was set to 12kHz, and 3100 sets of vibration data were collected for each state, for a total of 9300 sets of raw data.

[0088] Feature extraction and sample set construction, wherein feature extraction specifically involves calculating 26-dimensional feature indices on the preprocessed vibration signal, including 16-dimensional time-domain features, 6-dimensional frequency-domain features and 4-dimensional spectral kurtosis features;

[0089] The feature vector of a single sample is [C1,C2,...,C26], where: time-domain features include maximum, minimum, average, median, peak-to-peak value, rectified average, variance, standard deviation, kurtosis, skewness, root mean square, mean square value, root square amplitude, waveform factor, peak factor, impulse factor, and margin factor; frequency-domain features include centroid frequency, mean square frequency, root mean square frequency, frequency variance, and frequency standard deviation; spectral kurtosis features include mean spectral kurtosis, standard deviation spectral kurtosis, skewness of spectral kurtosis, and kurtosis of spectral kurtosis.

[0090] Specifically, using 1024 sampling points as a sample unit, a 26-dimensional feature index is calculated for each sample. These feature indices include time-domain features, frequency-domain features, and spectral kurtosis features; including:

[0091] Time-domain characteristics include maximum, minimum, average, median, peak-to-peak value, rectified average, variance, standard deviation, kurtosis, skewness, root mean square, root mean square value, root mean square amplitude, waveform factor, peak factor, impulse factor, and margin factor.

[0092] Frequency domain characteristics, including centroid frequency, mean square frequency, root mean square frequency, frequency variance, and frequency standard deviation;

[0093] Spectral kurtosis features include the mean of spectral kurtosis, the standard deviation of spectral kurtosis, the skewness of spectral kurtosis, and the kurtosis of spectral kurtosis.

[0094] Finally, 3000 valid samples were obtained, with 1000 samples for each state. The training set (2400 samples, 800 for each state) and the test set (600 samples, 200 for each state) were divided in an 8:2 ratio. All feature values ​​were then normalized to the [0,1] interval.

[0095] Step 202, Construction and performance verification of the IBES algorithm;

[0096] The improved IBES algorithm optimizes the region selection stage of the standard BES algorithm by introducing an adaptive position control parameter εc=exp[(Tt) / T]-1 to replace the original fixed parameter ε. The improved position update formula for the region selection stage is as follows:

[0097] ;

[0098] The search and dive-hunting phases follow the standard BES position update rules, thus completing the code implementation of the IBES algorithm.

[0099] The algorithm's performance was verified using 12 benchmark functions (single-peak functions Fun1~Fun4, multi-peak functions Fun5~Fun7, and composite functions Fun8~Fun12). Comparison algorithms included PSO, GWO, WOA, and standard BES. All algorithms had a population size of 30, a maximum of 1000 iterations, and were run independently 30 times. The results showed that, except for Fun5, the IBES algorithm outperformed the comparison algorithms in both convergence speed and optimal solution accuracy across the remaining 11 benchmark functions, validating the superiority of the IBES algorithm.

[0100] Step 203, DBN basic model construction;

[0101] A basic DBN model is constructed with the following structure: input layer (26 nodes) → first hidden layer → second hidden layer → output layer (3 nodes). The number of nodes in the input layer corresponds to 26-dimensional fault features, and the three nodes in the output layer correspond to three fault states: normal, pitting wear, and broken teeth. The hyperparameters to be optimized are the number of nodes in the first hidden layer (N1), the number of nodes in the second hidden layer (N2), the number of iterations (Niter), and the learning rate (LR), all with values ​​ranging from [10, 100].

[0102] Step 204, Training and optimization of the IBES-DBN fault diagnosis model;

[0103] IBES algorithm initialization: Set the population size to 20, the maximum number of iterations to 30, the search dimension to 4, the lower bound of the independent variable lb=[10,10,10,0.01], the upper bound of the independent variable ub=[100,100,100,0.2], and use the classification accuracy of DBN on the training set as the fitness function.

[0104] Iterative optimization: Initialize the bald eagle population, calculate the fitness value of each individual, and record the initial optimal solution; update the population position sequentially through the three stages of the IBES algorithm, and continuously update the global optimal solution during the iteration process; after reaching the maximum number of iterations of 30, the optimal hyperparameter combination is output as: N1=60, N2=20, Niter=59, LR=0.063.

[0105] Model training: The DBN model is configured with the optimal hyperparameter combination, and unsupervised pre-training is performed using the training set to train the RBM network layer by layer; then, the model parameters are fine-tuned using a supervised backpropagation algorithm to complete the training of the IBES-DBN model.

[0106] Step 205, Model performance testing and fault diagnosis;

[0107] Model Performance Comparison: The IBES-DBN model was tested using a test set and compared with SVM, ELM, DBN, the Particle Swarm Optimization-Deep Belief Network joint model PSO-DBN, and the Bald Eagle Search Algorithm-Deep Belief Network joint model BES-DBN. The fault diagnosis accuracy of each model is shown in Table 1 below:

[0108] Table 1. Fault diagnosis accuracy of each model

[0109] SVM 77.83% 77.37% 77.83% 77.27% ELM 82.17% 85.67% 78.22% 82.83% DBN 85.17% 85.03% 85.17% 84.98% PSO-DBN 90.17% 90.13% 90.17% 90.07% BES-DBN 92.83% 92.16% 92.83% 92.78% IBES-DBN 93.67% 93.67% 93.67% 93.61%

[0110] The results show that the IBES-DBN model has a classification accuracy of 93.67%, which is significantly better than the comparison model. The recall rate of broken tooth faults reaches 100%, achieving high-precision identification of gear faults.

[0111] Fault diagnosis implementation: After preprocessing and extracting 26-dimensional features from the gear vibration signal to be diagnosed, the signal is input into the trained IBES-DBN model. The model outputs the fault type and confidence level. For example, if a vibration feature sample of a broken tooth is input, the model outputs the fault type as "broken tooth" with a confidence level of 99.2%, thus completing the fault diagnosis.

[0112] Example 3

[0113] This application provides a wind turbine gearbox gear fault diagnosis system based on embodiments 1 and 2, which is used to execute the wind turbine gearbox gear fault diagnosis method proposed in embodiments 1 and 2 of this application. The system includes: a preprocessing module, an algorithm improvement module, a model building module, a training optimization module, and a fault diagnosis module.

[0114] The preprocessing module is used to collect vibration signals of gearbox gears through the wind turbine scale-down and monitoring system experimental platform, preprocess the vibration signals, construct a gear fault feature sample set, and divide the gear fault feature sample set into a training set and a test set.

[0115] The algorithm improvement module is used to introduce adaptively changing position control parameters in the selection area stage of the standard vulture search algorithm, replacing the original fixed-value parameters, and constructing an improved vulture search algorithm.

[0116] The model building module is used to build the basic model of the deep belief network. The construction method of the basic model of the deep belief network includes determining the number of input layer nodes of the basic model of the deep belief network based on the gear fault feature dimension of the gear fault feature sample set, determining the number of output layer nodes of the basic model of the deep belief network based on the gear fault category, determining the number of hidden layers of the basic model of the deep belief network through pre-experiments, and determining the hyperparameter combination to be optimized.

[0117] The training and optimization module is used to obtain the classification accuracy of the deep belief network base model on the training set, which serves as the fitness function for improving the vulture search algorithm. Solving for the optimal value of the fitness function yields the optimal hyperparameter combination of the deep belief network base model, thus completing the construction, training, and optimization of the fault diagnosis model.

[0118] The fault diagnosis module is used to preprocess and extract features from the gear vibration signals of the gearbox to be diagnosed, input them into the trained fault diagnosis model, output gear fault classification results, and complete fault diagnosis.

[0119] Example 4

[0120] Having introduced the wind turbine gearbox gear fault diagnosis system in an exemplary embodiment of this application, we will now introduce a computing device in another exemplary embodiment of this application.

[0121] Those skilled in the art will understand that various aspects of this application can be implemented as a system, method, or program product. Therefore, various aspects of this application can be specifically implemented in the following forms: a completely hardware implementation, a completely software implementation (including firmware, microcode, etc.), or a combination of hardware and software implementations, collectively referred to herein as a "circuit," "module," or "system."

[0122] In some possible implementations, the computing device according to this application may include at least one processor and at least one memory. The memory stores a computer program that, when executed by the processor, causes the processor to perform the steps of the cross-device assisted interaction method for off-site clearing according to various exemplary embodiments of this application described above.

[0123] The following reference Figure 3 To describe a computing device 130 according to this embodiment of the present application. Figure 3 The computing device 130 shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of this application. Figure 3 As shown, the computing device 130 is presented in the form of a general-purpose smart terminal (or Bluetooth headset). The components of the computing device 130 may include, but are not limited to: at least one processor 131, at least one memory 132, and a bus 133 connecting different system components (including memory 132 and processor 131).

[0124] Bus 133 represents one or more of several bus architectures, including a memory bus or memory controller, peripheral bus, processor, or local bus using any of the various bus architectures. Memory 132 may include readable media in the form of volatile memory, such as random access memory (RAM) 1321 and / or cache memory 1322, and may further include read-only memory (ROM) 1323. Memory 132 may also include a program / utility 1325 having a set (at least one) of program modules 1324, including but not limited to: an operating system, one or more application programs, other program modules, and program data, each or some combination of these examples may include an implementation of a network environment.

[0125] The computing device 130 can also communicate with one or more external devices 134 (e.g., keyboard, pointing device, etc.), and / or with any device that enables the computing device 130 to communicate with one or more other smart terminals (e.g., router, modem, etc.). This communication can be performed via input / output (I / O) interface 135. Furthermore, the computing device 130 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via network adapter 136. As shown, network adapter 136 communicates with other modules used in the computing device 130 via bus 133. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with the computing device 130, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0126] In some possible implementations, various aspects of the wind turbine gearbox gear fault diagnosis method provided in this application can also be implemented in the form of a program product, which includes a computer program. When the program product is run on a computer device, the computer program is used to cause the computer device to perform the steps in the wind turbine gearbox gear fault diagnosis method according to the various exemplary embodiments of this application described above.

[0127] The program product may employ any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0128] The program product for diagnosing gearbox gear faults in wind turbines according to embodiments of this application can be a portable compact disc read-only memory (CD-ROM) and include a computer program, and can run on a smart terminal. However, the program product of this application is not limited to this. In this document, the readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0129] A readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying a readable computer program. Such propagated data signals may take various forms, including—but not limited to—electromagnetic signals, optical signals, or any suitable combination thereof. A readable signal medium may also be any readable medium other than a readable storage medium, capable of sending, propagating, or transmitting a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0130] It should be noted that although several units or sub-units of the device have been mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, according to embodiments of this application, the features and functions of two or more units described above can be embodied in one unit. Conversely, the features and functions of one unit described above can be further divided and embodied by multiple units.

[0131] Furthermore, although the operations of the method of this application are described in a specific order in the accompanying drawings, this does not require or imply that these operations must be performed in that specific order, or that all the operations shown must be performed to achieve the desired result. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step, and / or one step may be broken down into multiple steps.

[0132] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable access frequency prediction device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable access frequency prediction device, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0133] These computer program instructions may also be stored in a computer-readable storage medium capable of directing a computer or other programmable access predictive device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0134] These computer program instructions can also be loaded onto a computer or other programmable access predictive device to cause a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable device for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0135] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0136] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for diagnosing gear faults in a wind turbine gearbox, characterized in that, include: Vibration signals of gearbox gears are collected through the wind turbine scale-down and monitoring system experimental platform. The vibration signals are preprocessed to construct a gear fault feature sample set, which is then divided into a training set and a test set. An improved vulture search algorithm is constructed by introducing adaptively changing position control parameters in the selection region stage of the standard vulture search algorithm, replacing the original fixed-value parameters. Constructing a deep belief network basic model; wherein, the method for constructing the deep belief network basic model includes determining the number of input layer nodes of the deep belief network basic model based on the gear fault feature dimension of the gear fault feature sample set, determining the number of output layer nodes of the deep belief network basic model based on the gear fault category, determining the number of hidden layers of the deep belief network basic model through pre-experiments, and determining the hyperparameter combination to be optimized; The classification accuracy of the deep belief network basic model on the training set is obtained and used as the fitness function of the improved vulture search algorithm. The optimal value of the fitness function is solved to obtain the optimal hyperparameter combination of the deep belief network basic model, thus completing the construction, training and optimization of the fault diagnosis model. The gear vibration signal of the gearbox to be diagnosed is preprocessed and its features are extracted and input into the trained fault diagnosis model. The gear fault classification result is output to complete the fault diagnosis.

2. The method according to claim 1, characterized in that, The vibration signal of the gearbox gears is collected through the wind turbine scale-down and monitoring system experimental platform, and the vibration signal is preprocessed, including: The vibration signal acquisition of gearbox gears through the wind turbine scale-down and monitoring system experimental platform includes acquiring vibration signals of gearbox gears under different speeds and loads in three states: normal, pitting wear, and broken teeth, and preprocessing the original vibration signals.

3. The method according to claim 1, characterized in that, The construction of the improved bald eagle search algorithm includes: In the selection region phase of the standard bald eagle search algorithm, the formula for updating the bald eagle's position is: ； in, This represents the bald eagle's current optimal position; and These represent the current and previous positions of the j-th vulture, respectively; the average distribution position of the vultures is represented by... Indicates; control parameters random numbers ; Introducing adaptive position control parameters Replace the fixed parameter ε, The expression is: ； in, Here are the control parameters for the vulture's position change; t is the current iteration number; T represents the maximum number of iterations; the improved formula for updating the vulture's position during the selected region stage is: ； By combining the position update rules of the search phase and the dive-hunting phase of the standard bald eagle search algorithm, an improved bald eagle search algorithm was constructed.

4. The method according to claim 1, characterized in that, The basic model for constructing a deep belief network consists of an input layer, two hidden layers, and an output layer stacked together. The number of nodes in the input layer is consistent with the feature dimension of a single sample; the number of nodes in the output layer is consistent with the number of gear fault categories. The hyperparameter combination to be optimized is: ,in This represents the number of nodes in the first hidden layer. This represents the number of nodes in the second hidden layer. The number of training iterations for the model. This is the learning rate.

5. The method according to claim 1, characterized in that, The construction, training, and optimization of the fault diagnosis model include: Initialize the parameters of the improved vulture search algorithm; the algorithm parameters include at least one of the following: population size, maximum number of iterations, upper and lower limits of independent variables, and search space dimension; wherein, the search space dimension is consistent with the number of hyperparameters to be optimized in the basic model of the deep belief network, and the upper and lower limits of independent variables correspond to the value range of each hyperparameter; The fitness function of the improved bald eagle search algorithm is based on the classification accuracy of the deep belief network basic model on the training set. The fitness value of each bald eagle in the initial population is calculated, and the initial optimal fitness value and the corresponding bald eagle position are recorded. Entering the iterative optimization process, the location of the bald eagle population is updated sequentially through the selection area stage, search stage, and dive hunting stage of the improved bald eagle search algorithm. The fitness value corresponding to each location is recalculated. If the fitness value of the new location is better than the saved optimal fitness value, the optimal fitness value and the corresponding bald eagle location are updated; otherwise, the original value is kept. Determine if the maximum number of iterations has been reached. If not, return to the iterative optimization process to continue iterating. If the maximum number of iterations has been reached, terminate the iteration and output the position of the vulture corresponding to the optimal fitness value as the optimal combination of hyperparameters. The parameters of the deep belief network basic model are configured using the optimal combination of hyperparameters. Unsupervised pre-training and supervised fine-tuning of the model are completed using the training set. The model performance is verified using the test set, and finally, the trained fault diagnosis model is obtained.

6. The method according to claim 5, characterized in that, The fitness function expression is: ； in, For fitness value, The classification accuracy of the deep belief network base model on the training set. The calculation formula is: ； Where TP is the number of positive samples correctly classified, TN is the number of negative samples correctly classified, FP is the number of negative samples misclassified as positive samples, and FN is the number of positive samples misclassified as negative samples.

7. The method according to claim 1, characterized in that, The gear vibration signal of the gearbox to be diagnosed is preprocessed and its features extracted, and then input into the trained fault diagnosis model. The output gear fault classification results include: The feature extraction includes at least one of data cleaning, outlier removal, and normalization. The gear fault classification result includes the gear's health status, fault type, fault confidence level, and at least one of the following: fault warning information generated based on the classification result.

8. A gear fault diagnosis system for a wind turbine gearbox, characterized in that, include: The preprocessing module is used to collect vibration signals of gearbox gears through the wind turbine scale-down and monitoring system experimental platform, preprocess the vibration signals, construct a gear fault feature sample set, and divide the gear fault feature sample set into a training set and a test set. The algorithm improvement module is used to introduce adaptively changing position control parameters in the selection area stage of the standard vulture search algorithm, replacing the original fixed-value parameters, and constructing an improved vulture search algorithm. The model building module is used to build a deep belief network basic model. The method for building the deep belief network basic model includes determining the number of input layer nodes of the deep belief network basic model based on the gear fault feature dimension of the gear fault feature sample set, determining the number of output layer nodes of the deep belief network basic model based on the gear fault category, determining the number of hidden layers of the deep belief network basic model through pre-experiments, and determining the hyperparameter combination to be optimized. The training optimization module is used to obtain the classification accuracy of the deep belief network basic model on the training set, which serves as the fitness function for improving the vulture search algorithm. The optimal value of the fitness function is solved to obtain the optimal hyperparameter combination of the deep belief network basic model, thus completing the construction, training, and optimization of the fault diagnosis model. The fault diagnosis module is used to preprocess and extract features from the gear vibration signals of the gearbox to be diagnosed, input them into the trained fault diagnosis model, output gear fault classification results, and complete fault diagnosis.

9. A computing device, characterized in that, Its features include: Memory, used to store program instructions; A processor is configured to invoke program instructions stored in the memory and execute the method as described in any one of claims 1-7 according to the obtained program instructions.

10. A computer-readable storage medium, characterized in that, Includes computer-readable instructions that, when read and executed by a computer, cause the method as described in any one of claims 1 to 7 to be implemented.

Images