A wind turbine bearing fault diagnosis method and system

Abstract

The application discloses a kind of wind generator bearing fault diagnosis method and system, it is related to wind power generation equipment fault diagnosis technical field.The method includes with specific sensor acquisition bearing vibration data and pre-processing;It is denoised by enhanced combination differential morphological filter ECGMF;Intrinsic time scale decomposition ITD algorithm is used to decompose signal after denoising, extract component instantaneous amplitude and spectrum analysis obtains fault characteristic;Deep support vector machine DSVM model is constructed in combination with support vector machine SVM and neural network;Characteristic vector is input into deep support vector machine DSVM model training and judging fault output result.The system contains vibration data acquisition, enhanced combination differential morphological filtering, intrinsic time scale decomposition, deep support vector machine and diagnostic result output module, corresponding to realize the function of each step above, through each link cooperation, effectively extract fault characteristic and accurately diagnose, can promote the accuracy and efficiency of wind generator bearing fault diagnosis.

Description

Technical Field

[0001] This invention relates to the field of wind power equipment fault diagnosis technology, and more specifically, to a method and system for diagnosing wind turbine bearing faults. Background Technology

[0002] As an important renewable energy power generation device, wind turbines often operate in harsh environments, facing interference from strong background noise and electromagnetic noise. These interference factors make it difficult to accurately detect fault characteristics of the generator bearings, thus posing numerous challenges to the normal operation and maintenance of wind turbines.

[0003] The existing fault diagnosis technologies have the following problems:

[0004] First, traditional methods struggle to effectively overcome the impact of strong noise interference when detecting fault feature information in rolling bearings. Although some filters in mathematical morphology can be used for signal processing, existing kurtosis criteria and signal-to-noise ratio methods are not accurate enough in selecting the scale of the structuring element, making it difficult to accurately extract weak fault feature information from strong background noise.

[0005] Secondly, the vibration signal of the self-aligning roller bearing fault in wind turbine generators has non-stationary and non-linear characteristics. Traditional fault feature extraction methods are difficult to adapt to these complex signal characteristics and cannot fully and accurately extract feature information that reflects the bearing fault state.

[0006] Finally, as a commonly used classification tool, Support Vector Machine (SVM) cannot fully explore the deep patterns in data when processing data, and its diagnostic effect has certain limitations in complex wind turbine bearing fault diagnosis scenarios.

[0007] Therefore, there is an urgent need for a more accurate, efficient, and adaptable method and system for diagnosing wind turbine bearing faults that can adapt to complex environments and diverse fault modes. Summary of the Invention

[0008] In view of the shortcomings of the existing technology, the purpose of this invention is to provide a method and system for diagnosing bearing faults in wind turbine generators.

[0009] To achieve the above objectives, the present invention provides the following technical solution:

[0010] Firstly, a method for diagnosing bearing faults in wind turbine generators includes the following steps:

[0011] Vibration data of wind turbine bearings were collected using D / SENL020 and D / SENN020 sensors and preprocessed.

[0012] The collected wind turbine bearing vibration signal was denoised by enhancing the combined differential morphological filter ECGMF.

[0013] The Inherent Time Decomposition (ITD) algorithm is used to decompose the denoised vibration signal, extract the instantaneous amplitude of the main components and perform spectrum analysis to obtain the fault characteristics of the wind turbine bearing.

[0014] A deep support vector machine (DSVM) model is constructed by combining support vector machine (SVM) with a neural network model.

[0015] The feature vectors are input into a deep support vector machine (DSVM) model for training, and the model is used to diagnose the faults of wind turbine bearings and output the diagnostic results.

[0016] Preferably, the vibration data includes main bearing axial vibration data, main bearing horizontal vibration data, main bearing vertical vibration data, rear bearing vertical vibration data, stator axial vibration data, and stator horizontal vibration data; the preprocessing includes data visualization inspection and data normalization.

[0017] Preferably, the acquired vibration signal is denoised using an enhanced combined differential morphological filter (ECGMF), including:

[0018] Based on data visualization analysis and understanding of the vibration source characteristics, the multi-scale range to be used is determined; a flat structural element with zero height is selected.

[0019] For each selected scale value, set the corresponding structural element size; clarify the combination method of various mathematical morphological operations involved in the enhanced combined differential morphological filter ECGMF transformation at the current scale; and perform mathematical morphological operations on the preprocessed vibration signal in sequence.

[0020] The mathematical morphology described consists of four basic operators: dilation, erosion, opening, and closing. Assume the original one-dimensional signal f(n) is defined as F = (0, 1, ..., N-1), and the structuring element g(m) is defined as G = (0, 1, ..., M-1), N ≥ M; where n is an independent variable representing the index or sequence number of the discrete sampling point of the one-dimensional signal f(n); and m is a variable representing the index or sequence number of the element within the domain of the structuring element g(m). The four basic operators are defined as follows:

[0021] Dilation operator:

[0022] Erosion operator: (fΘg)(n)=min[f(n+m)-g(m)]

[0023] Opening the operator:

[0024] Closing operator:

[0025] in, —Inflation operator; Θ —Erosion operator; — Opening operator; · — Closing operator;

[0026] The erosion operator can reduce the peak value of a signal and amplify the trough value of the signal, while the closing operator can suppress the negative pulse of the signal; conversely, the dilation operator can reduce the trough value of the signal and amplify the peak value of the signal, while the opening operator can suppress the positive pulse of the signal.

[0027] Based on this property, two cascade operators consisting of expansion and closure operators are defined as follows:

[0028]

[0029] A cascaded operator consisting of an erosion operator and an opening operator is defined as follows:

[0030]

[0031] F DC and F CD The ability to suppress negative pulses and extract positive pulse signals is enhanced; conversely, F EO The ability to suppress positive pulses and extract negative pulses is enhanced, thus allowing the definition of new difference operators:

[0032]

[0033] Considering that early fault information of wind turbine bearings is often severely interfered with by background noise, in order to extract weak fault feature information from strong background noise, F... CD_EO and F DC_EO Based on the characteristics of the operator, a new enhanced combinatorial difference morphology operator is proposed:

[0034] ECGMF(n)=F CD_EO (n)·F DC_EO (n)

[0035] At each scale, the vibration signal after the current-scale enhanced combined differential morphological filter (ECGMF) transformation is analyzed in terms of spectrum.

[0036] Based on the failure mechanism of the target equipment, past experience, and relevant theoretical research, determine the characteristic frequencies related to equipment failure;

[0037] For the spectrum at each scale, the characteristic frequency intensity coefficients are calculated, and then the characteristic frequency intensity coefficients C CFI The definition is as follows:

[0038]

[0039] in, —The amplitude of the i-th fault frequency; —Amplitude at each point j; M, N —Fault frequency and total number of spectrum points;

[0040] Since the fault characteristic frequency is relatively less affected by interference, the characteristic frequency intensity coefficient C is selected. CFI To determine the optimal scale of the structural element; the required characteristic frequency intensity coefficient C CFI The larger the value, the more obvious the fault characteristic signal extraction. Therefore, the largest characteristic frequency intensity coefficient C is used. CFI The scale corresponding to the value is taken as the optimal scale of the structuring element;

[0041] Based on the optimal scale of the determined structural element, set the corresponding structural element parameters; redetermine the various mathematical morphological operations involved in the enhanced combined differential morphological filter ECGMF transformation at the optimal scale; substitute the preprocessed original vibration signal into the enhanced combined differential morphological filter ECGMF for mathematical morphological operations to complete the denoising process.

[0042] Preferably, the Inherent Time Decomposition (ITD) algorithm is used to decompose the denoised vibration signal, extract the instantaneous amplitude of the main components, and perform spectral analysis to obtain the fault characteristics of the wind turbine bearing, including:

[0043] The denoised vibration signal is decomposed using Intrinsic Time Decomposition (ITD). The vibration signal can be decomposed into a series of intrinsic rotational components with different characteristic scales and a trend component, specifically:

[0044] For signal X t Define ξ as the baseline extraction operator, such that from signal X... t The remaining signal after extracting a baseline becomes an intrinsic rotation component; signal X t The first decomposition is as follows:

[0045] X t =ξX t +(1-ξ)X t =L t +H t

[0046] Among them, L t =ξX t It is the baseline signal, H t =(1-ξ)X t It is an inherent rotational component;

[0047] Assume {T KK = 1, 2, ... is the signal X t The local poles are defined with T0 = 0 as the simplified notation, and are respectively represented by X. K and L K X(T) represents K ) and L(T K ); Assume that in [0,T K Define L t and H t And X t In t∈[0,T] K+2 [It is defined; in the continuous pole interval (T] K ,T K+1 Define the piecewise linear baseline extraction factor ξ above:

[0048]

[0049] Where α∈(0,1), and is taken as 0.5; a baseline signal L is obtained through one decomposition. t and an inherent rotational component H t The inherent rotational component H t This represents the local relative high-frequency components in the signal;

[0050] The baseline signal obtained from the first decomposition is used as the new input signal for further decomposition until a monotonic signal is obtained; this transforms the original vibration signal X... t The process can be decomposed into the sum of several inherent rotational components in different frequency ranges from high to low, plus a monotonic trend component. The entire process can be represented as follows:

[0051]

[0052] Where, ψξ K X t It is the intrinsic rotation component of the (K+1)th layer, ξ P X t It is either a monotonic trend component or the baseline signal with the lowest frequency is extracted before the monotonic trend component is obtained;

[0053] Spectral analysis was performed on the instantaneous amplitudes of the first K inherent rotational components containing significant periodic impact to extract the bearing fault feature vector. Where i = 1, 2, ..., k, specifically:

[0054] The Intrinsic Time Decomposition (ITD) algorithm proposes a new method for defining instantaneous amplitude, instantaneous phase, and instantaneous frequency using the whole wave as the unit.

[0055] The instantaneous phase, in the context of full wave, refers to the signal between two adjacent upper zero-crossing points; in the context of half wave, it refers to the signal between two zero-crossing points; and in the context of monotonic interval, it refers to the signal between two adjacent extreme points. Based on this, the instantaneous phase θt It can be defined as:

[0056]

[0057] Where A1>0 and A2>0 represent the amplitudes of the positive and negative half-waves of a full wave, respectively; t1 and t5 are two times when the wave crosses zero; t2 is the maximum point (A1) of the positive half-wave; t3 is the time when the wave crosses zero below; and t4 is the minimum point (-A2) of the negative half-wave.

[0058] The instantaneous frequency, the instantaneous frequency f t It is the instantaneous phase θ t The derivative, that is:

[0059]

[0060] The instantaneous amplitude, instantaneous amplitude A t Defined in half-wave units, the extreme value of a signal between two zero-crossing points is a constant within each half-wave, i.e.:

[0061]

[0062] After completing the spectral analysis, feature information related to bearing faults is extracted from the spectrum of each inherent rotational component to form a bearing fault feature vector. Where i = 1, 2, ..., k, that is, the above feature extraction operation is performed on the first K selected inherent rotational components to obtain the corresponding bearing fault feature vector.

[0063] Preferably, a deep support vector machine (DSVM) model is constructed by combining a support vector machine (SVM) with a neural network model, including:

[0064] The overall structure of the Deep Support Vector Machine (DSVM) model includes an input layer, a hidden layer, and an output layer. The input layer receives the extracted bearing fault feature vectors. The hidden layer performs feature mapping on the input data and selects a fully connected neural network structure as the hidden layer. The output layer replaces the output layer of the neural network with a Support Vector Machine (SVM) and uses the SVM to classify the mapped data. By solving the classification hyperplane, the sample data is divided into positive and negative classes.

[0065] The objective function of Support Vector Machine (SVM) is also applicable to the classification process of the output layer of Deep Support Vector Machine (DSVM) model, assuming the original data is X. i The data after the hidden layer mapping is Φ(x) i The objective function of the Deep Support Vector Machine (DSVM) model is as follows:

[0066]

[0067] in, ξ represents the normal vector of the hyperplane; b represents the intercept of the hyperplane; C represents the penalty function; ξ i y represents a slack variable, primarily introduced to handle cases where the data may be linearly inseparable; i The true class label of a sample is typically used in binary classification problems, where y i The value can be +1 or -1, corresponding to two different categories; Φ(x) i ) represents the data after mapping through the hidden layers of the Deep Support Vector Machine (DSVM) model; l represents the number of samples;

[0068] The constraints of the objective function are rewritten as follows:

[0069]

[0070] The objective function can be rewritten as an unconstrained optimization problem, which is called the loss function of the Deep Support Vector Machine (DSVM) model, as shown below:

[0071]

[0072] The loss function of the Deep Support Vector Machine (DSVM) model is mainly divided into two parts. The first part is... Represents the interval hyperplane Distance maximization is also a regularization term in neural network models; Part Two It is the hinge loss, which allows some sample points to be misclassified; C is the penalty parameter. If C is chosen to be larger, it will minimize the number of misclassified sample points; if C is chosen to be smaller, it will maximize the distance between the margin hyperplanes, which can tolerate some misclassified training sample points.

[0073] Preferably, the feature vectors are input into a deep support vector machine (DSVM) model for training, and fault diagnosis is performed on the wind turbine bearings, outputting diagnostic results, including:

[0074] The extracted bearing fault feature vector dataset is divided into training set, validation set and test set. The feature vectors of the training set are input into the deep support vector machine (DSVM) model. After feature mapping in the hidden layer, the prediction result is obtained by the output layer support vector machine (SVM) classification.

[0075] Based on the prediction results and the true labels, the loss value is calculated according to the loss function of the deep support vector machine (DSVM) model. Then, the backpropagation algorithm is used to update the parameters of each layer of the model. The training is iterated repeatedly until the stopping condition is met.

[0076] During or after iterative training, the model is input with the feature vector of the validation set to obtain the prediction results. The evaluation metrics, including accuracy, recall, and F1 score, are calculated in combination with the corresponding real labels, and the model is optimized accordingly.

[0077] Finally, the feature vectors of the test set are input into the optimized model to obtain the prediction results, and the fault diagnosis results are output in a clear manner for bearing fault diagnosis.

[0078] In a second aspect, a wind turbine bearing fault diagnosis system includes a vibration data acquisition module, an enhanced combined differential morphological filter module, an inherent time-scale decomposition module, a deep support vector machine module, and a diagnosis result output module.

[0079] The vibration data acquisition module acquires and preprocesses vibration data of the wind turbine bearing through D / SENL020 and D / SENN020 sensors.

[0080] The enhanced combined differential morphological filter module performs noise reduction processing on the collected wind turbine bearing vibration signal through the enhanced combined differential morphological filter ECGMF.

[0081] The inherent time scale decomposition module uses the inherent time decomposition (ITD) algorithm to decompose the denoised vibration signal, extract the instantaneous amplitude of the main components and perform spectrum analysis to obtain the fault characteristics of the wind turbine bearing.

[0082] The deep support vector machine module combines support vector machine (SVM) with a neural network model to construct a deep support vector machine (DSVM) model.

[0083] The diagnostic result output module inputs the feature vector into the deep support vector machine (DSVM) model for training, performs fault judgment on the wind turbine bearing, and outputs the diagnostic result.

[0084] Compared with the prior art, the present invention has the following beneficial effects:

[0085] 1. In this invention, the Enhanced Combined Differential Morphological Filter (ECGMF) is used to detect bearing fault information, and the characteristic frequency intensity coefficient is employed to optimize the structural element scale, effectively extracting fault feature information from rolling bearing signals. Compared with other morphological filters, the Enhanced Combined Differential Morphological Filter (ECGMF) is superior in extracting fault components and suppressing noise in large wind turbine rolling bearings. Furthermore, the Enhanced Combined Differential Morphological Filter (ECGMF) shows significant advantages over other morphological filters in terms of kurtosis and fault energy ratio. For fault diagnosis of large wind turbine rolling bearings, the Enhanced Combined Differential Morphological Filter (ECGMF) has higher feature extraction efficiency than other morphological filters.

[0086] 2. In this invention, the vibration signal of the wind turbine bearing exhibits complex nonlinear, time-varying, and multi-coupling characteristics. The Intrinsic Time Decomposition (ITD) algorithm can decompose the complex signal into a sum of several intrinsic rotational components and a trend component. Spectral analysis is performed on the instantaneous amplitude of the intrinsic rotational component containing significant periodic impact components, and the amplitude at the fault characteristic frequency is extracted as the feature vector for bearing fault diagnosis. The ITD algorithm can accurately extract the dynamic characteristics of non-stationary signals, has high decomposition efficiency and frequency resolution, is suitable for analyzing non-stationary signals with time-varying spectra, has small edge effects, does not require spline interpolation, has no screening process, and can process large amounts of data in real time.

[0087] 3. In this invention, a Deep Support Vector Machine (DSVM) model is used for fault diagnosis of generator bearings. Firstly, compared to Support Vector Machines (SVMs), the hidden layers of the DSVM model can replace the manual selection of kernel functions. Furthermore, the DSVM model updates network parameters through backpropagation, eliminating the need to solve a quadratic programming problem. Secondly, compared to neural networks, the classification process of the DSVM model's output layer is more interpretable, thus enhancing the ability to uncover deeper patterns in the data. Attached Figure Description

[0088] Figure 1 This invention provides a schematic flowchart of a method for diagnosing bearing faults in wind turbine generators.

[0089] Figure 2 This invention provides a schematic flowchart of a wind turbine bearing fault diagnosis system.

[0090] Figure 3 This is a block diagram of the Deep Support Vector Machine (DSVM) model proposed in this invention. Detailed Implementation

[0091] Reference Figures 1 to 3 .

[0092] The embodiments further illustrate the wind turbine bearing fault diagnosis method and system proposed in this invention.

[0093] A method for diagnosing bearing faults in wind turbine generators includes the following steps:

[0094] Vibration data of wind turbine bearings were collected using D / SENL020 and D / SENN020 sensors and preprocessed.

[0095] The collected wind turbine bearing vibration signal was denoised by enhancing the combined differential morphological filter ECGMF.

[0096] The Inherent Time Decomposition (ITD) algorithm is used to decompose the denoised vibration signal, extract the instantaneous amplitude of the main components and perform spectrum analysis to obtain the fault characteristics of the wind turbine bearing.

[0097] A deep support vector machine (DSVM) model is constructed by combining support vector machine (SVM) with a neural network model.

[0098] The feature vectors are input into a deep support vector machine (DSVM) model for training, and the model is used to diagnose the faults of wind turbine bearings and output the diagnostic results.

[0099] The vibration data includes main bearing axial vibration data, main bearing horizontal vibration data, main bearing vertical vibration data, rear bearing vertical vibration data, stator axial vibration data, and stator horizontal vibration data; the preprocessing includes data visualization inspection and data normalization.

[0100] The data visualization inspection involves plotting the collected raw vibration signal as a time-domain waveform and a frequency-domain spectrum. This allows for a visual and intuitive observation of the signal's basic characteristics, such as whether there are obvious abnormal fluctuations, high-frequency noise interference, or the approximate frequency distribution range of the signal.

[0101] The data normalization process normalizes the vibration signal, ensuring that its value falls within a specific interval, commonly [0,1] or [-1,1]. A minimum-maximum normalization formula can be used. Where x is the original signal, x min and x max These are the minimum and maximum values ​​in the original signal; normalization helps subsequent filters better handle signals of different magnitudes and can promote convergence and stable operation in some algorithms.

[0102] The D / SENL020 sensor is characterized by high sensitivity, which can accurately capture minute changes in the bearing vibration process and plays an important role in detecting some early and relatively weak fault signals.

[0103] The D / SENN020 sensor excels in wideband response, covering a broad range of vibration frequencies to ensure that no vibration frequency components that may be related to a fault are missed.

[0104] The acquired wind turbine bearing vibration signal is denoised using an enhanced combined differential morphological filter (ECGMF), including:

[0105] The enhanced combined differential morphological filter ECGMF is a filter designed based on mathematical morphology principles. It effectively removes noise by analyzing and processing the morphological characteristics of the signal.

[0106] Based on data visualization analysis and an understanding of the characteristics of the vibration source, the multi-scale range to be used is determined.

[0107] The structuring element is mainly composed of three parts: height, length, and shape. The shapes of the structuring element are mainly triangular, elliptical, and flat. The more complex the shape of the structuring element, the longer the calculation time. Therefore, in order to improve the calculation efficiency, the enhanced combined differential morphological filter (ECGMF) selects a flat structuring element with zero height. If the scale of the structuring element is too long, the fault feature signal may be treated as noise and filtered out, resulting in the loss of feature information. Conversely, if it is too short, a large amount of noise components will remain in the signal. Therefore, it is essential to select an appropriate structuring element scale.

[0108] For each selected scale value, set the corresponding structuring element size; clarify the combination method of various mathematical morphological operations involved in the enhanced combined differential morphological filter ECGMF transformation at the current scale; and perform mathematical morphological operations on the preprocessed vibration signal in sequence.

[0109] The mathematical morphology described consists of four basic operators: dilation, erosion, opening, and closing. Assume the original one-dimensional signal f(n) is defined as F = (0, 1, ..., N-1), and the structuring element g(m) is defined as G = (0, 1, ..., M-1), N ≥ M; where n is an independent variable representing the index or sequence number of the discrete sampling point of the one-dimensional signal f(n); and m is a variable representing the index or sequence number of the element within the domain of the structuring element g(m). The four basic operators are defined as follows:

[0110] Dilation operator:

[0111] Erosion operator: (fΘg)(n)=min[f(n+m)-g(m)]

[0112] Opening the operator:

[0113] Closing operator:

[0114] in, —Inflation operator; Θ —Erosion operator; — Opening operator; · — Closing operator.

[0115] The erosion operator can reduce the peak value of a signal and amplify the trough value of the signal, while the closing operator can suppress the negative pulse of the signal; conversely, the expansion operator can reduce the trough value of the signal and amplify the peak value of the signal, while the opening operator can suppress the positive pulse of the signal.

[0116] Based on this property, two cascade operators consisting of expansion and closure operators are defined as follows:

[0117]

[0118] A cascaded operator consisting of an erosion operator and an opening operator is defined as follows:

[0119]

[0120] F DC and F CD The ability to suppress negative pulses and extract positive pulse signals is enhanced; conversely, F EO The ability to suppress positive pulses and extract negative pulses is enhanced, thus allowing the definition of new difference operators:

[0121]

[0122] Considering that early fault information of wind turbine bearings is often severely interfered with by background noise, in order to extract weak fault feature information from strong background noise, F... CD_EO and F DC_EO Based on the characteristics of the operator, a new enhanced combinatorial difference morphology operator is proposed:

[0123] ECGMF(n)=F CD_EO (n)·F DC_EO (n)

[0124] At each scale, the vibration signal after undergoing the ECGMF transformation operation of the current scale enhanced combined differential morphological filter is subjected to spectral analysis. Typically, methods such as Fast Fourier Transform (FFT) are used to convert the time-domain data into frequency-domain data to obtain the spectrum of the vibration data at the current scale. The spectrum shows the energy distribution of the vibration signal at different frequencies, that is, the amplitude of each frequency component.

[0125] Based on the failure mechanism of the target equipment (such as wind turbine bearings), past experience, and relevant theoretical research, determine the characteristic frequencies related to equipment failure;

[0126] For the spectrum at each scale, the characteristic frequency intensity coefficients are calculated, and then the characteristic frequency intensity coefficients C CFI The definition is as follows:

[0127]

[0128] in, —The amplitude of the i-th fault frequency; —Amplitude at each j point; M, N —Fault frequency and total number of spectrum points.

[0129] Since the fault characteristic frequency is relatively less affected by interference, the characteristic frequency intensity coefficient C is selected. CFI To determine the optimal scale of the structural element; the required characteristic frequency intensity coefficient C CFI The larger the value, the more obvious the fault characteristic signal extraction. Therefore, the largest characteristic frequency intensity coefficient C is used. CFI The scale corresponding to the value is taken as the optimal scale of the structuring element;

[0130] Based on the optimal scale of the determined structural element, set the corresponding structural element parameters; redetermine the various mathematical morphological operations involved in the enhanced combined differential morphological filter ECGMF transformation at the optimal scale; substitute the preprocessed original vibration signal into the enhanced combined differential morphological filter ECGMF for mathematical morphological operations to complete the denoising process.

[0131] The Intrinsic Time Decomposition (ITD) algorithm is used to decompose the denoised vibration signal, extract the instantaneous amplitude of the main components, and perform spectral analysis to obtain the fault characteristics of the wind turbine bearing, including:

[0132] The components decomposed by the Inherent Time Decomposition (ITD) algorithm have certain physical meanings and can effectively reflect the characteristics of the original signal, making it suitable for analyzing signals containing amplitude modulation (AM) and frequency modulation (FM) components.

[0133] The denoised vibration signal is subjected to intrinsic time decomposition (ITD). The vibration signal can be decomposed into a series of intrinsic rotational components with different characteristic scales and a trend component, specifically:

[0134] For signal X t Define ξ as the baseline extraction operator, such that from signal X... t The remaining signal after extracting a baseline becomes an intrinsic rotation component; signal X t The first decomposition is as follows:

[0135] X t =ξX t +(1-ξ)X t =L t +H t

[0136] Among them, L t =ξX t It is the baseline signal, H t =(1-ξ)X t It is an inherent rotational component;

[0137] Assume {T K K = 1, 2, ... is the signal X t The local poles are defined with T0 = 0 as the simplified notation, and are respectively represented by X. K and LK X(T) represents K ) and L(T K ); Assume that in [0,T K Define L t and H t And X t In t∈[0,T] K+2 [It is defined; in the continuous pole interval (T] K ,T K+1 Define the piecewise linear baseline extraction factor ξ above:

[0138]

[0139] Where α∈(0,1), and is taken as 0.5; a baseline signal L is obtained through one decomposition. t and an inherent rotational component H t The inherent rotational component H t This represents the local relative high-frequency components in a signal.

[0140] The baseline signal obtained from the first decomposition is used as the new input signal for further decomposition until a monotonic signal is obtained; this transforms the original vibration signal X... t The process can be decomposed into the sum of several inherent rotational components in different frequency ranges from high to low, plus a monotonic trend component. The entire process can be represented as follows:

[0141]

[0142] Where, ψξ K X t It is the intrinsic rotation component of the (K+1)th layer, ξ P X t It is the monotonic trend component or the baseline signal with the lowest frequency that is extracted before the monotonic trend component is obtained.

[0143] Spectral analysis was performed on the instantaneous amplitudes of the first K inherent rotational components containing significant periodic impact to extract the bearing fault feature vector. Where i = 1, 2, ..., k, specifically:

[0144] The Intrinsic Time Decomposition (ITD) algorithm proposes a new method for defining instantaneous amplitude, instantaneous phase, and instantaneous frequency using the whole wave as the unit.

[0145] The instantaneous phase, in the context of full wave, refers to the signal between two adjacent upper zero-crossing points; in the context of half wave, it refers to the signal between two zero-crossing points; and in the context of monotonic interval, it refers to the signal between two adjacent extreme points. Based on this, the instantaneous phase θ t It can be defined as:

[0146]

[0147] Where A1>0 and A2>0 represent the amplitudes of the positive and negative half-waves of a full wave, respectively; t1 and t5 are two times when the wave crosses zero; t2 is the maximum point (A1) of the positive half-wave; t3 is the time when the wave crosses zero below; and t4 is the minimum point (-A2) of the negative half-wave.

[0148] The instantaneous amplitude, instantaneous amplitude A t Defined in half-wave units, the extreme value of a signal between two zero-crossing points is a constant within each half-wave, i.e.:

[0149]

[0150] After completing the spectral analysis, feature information related to bearing faults is extracted from the spectrum of each inherent rotational component to form a bearing fault feature vector. Where i = 1, 2, ..., k, that is, the above feature extraction operation is performed on the first K selected inherent rotational components to obtain the corresponding bearing fault feature vector.

[0151] The deep support vector machine (DSVM) model is constructed by combining support vector machines (SVM) with neural network models, including:

[0152] The overall structure of the Deep Support Vector Machine (DSVM) model includes an input layer, a hidden layer, and an output layer. The input layer receives the extracted bearing fault feature vectors. The hidden layer performs feature mapping on the input data and selects a fully connected neural network structure as the hidden layer. The output layer replaces the output layer of the neural network with an SVM model and uses the SVM model to classify the mapped data. By solving the classification hyperplane, the sample data is divided into positive and negative classes. The classification process is simple and easy to understand, enhancing the interpretability of the DSVM model.

[0153] The objective function of Support Vector Machine (SVM) is also applicable to the classification process of the output layer of Deep Support Vector Machine (DSVM) model, assuming the original data is X. i The data after the hidden layer mapping is Φ(x) i The objective function of the Deep Support Vector Machine (DSVM) model is as follows:

[0154]

[0155] in, ξ represents the normal vector of the hyperplane; b represents the intercept of the hyperplane; C represents the penalty function; ξ i y represents a slack variable, primarily introduced to handle cases where the data may be linearly inseparable; i The true class label of a sample is typically used in binary classification problems, where y iThe value can be +1 or -1, corresponding to two different categories; Φ(x) i ) represents the data after mapping through the hidden layers of the Deep Support Vector Machine (DSVM) model; l represents the number of samples;

[0156] The constraints of the objective function are rewritten as follows:

[0157]

[0158] The objective function can be rewritten as an unconstrained optimization problem, which is called the loss function of the Deep Support Vector Machine (DSVM) model, as shown below:

[0159]

[0160] The loss function of the Deep Support Vector Machine (DSVM) model is mainly divided into two parts. The first part is... Represents the interval hyperplane Distance maximization is also a regularization term in neural network models; Part Two It is the hinge loss, which allows some sample points to be misclassified; C is the penalty parameter. If C is chosen to be larger, it will minimize the number of misclassified sample points; if C is chosen to be smaller, it will maximize the distance between the margin hyperplanes, which can tolerate some misclassified training sample points.

[0161] The feature vectors are input into a deep support vector machine (DSVM) model for training, and fault diagnosis is performed on the wind turbine bearings. The diagnostic results are output, including:

[0162] The extracted bearing fault feature vector dataset is divided into training, validation, and test sets according to a certain ratio; a common ratio is 80% as training set, 10% as validation set, and 10% as test set; ensure that the data format and labeling are accurate so that the model can correctly identify and process it.

[0163] The feature vectors of the training set are input into the input layer of the deep support vector machine (DSVM) model, and then passed through the hidden layer for feature mapping. The data processed by the hidden layer is then input into the output layer of the support vector machine (SVM) for classification to obtain the predicted classification result.

[0164] Based on the predicted classification results and the actual sample labels, the loss value is calculated according to the loss function of the Deep Support Vector Machine (DSVM) model. Using the calculated loss value, the model parameters, including the connection weights and biases of the hidden layers and the parameters of the SVM, are updated sequentially from the output layer to the input layer through the backpropagation algorithm.

[0165] Repeat the steps of forward propagation, loss calculation, backpropagation, and parameter update to perform multiple iterations of training until the model reaches the predetermined stopping condition. Through continuous iterative training, the deep support vector machine (DSVM) model can gradually learn the features and patterns of the data, improving the accuracy of classification and the model's generalization ability.

[0166] During or after each iteration of training, the validation set feature vectors are input into the trained deep support vector machine (DSVM) model, and the predicted classification results are obtained according to the forward propagation steps described above. Then, based on the predicted results and the true fault labels corresponding to the validation set feature vectors, evaluation metrics, including accuracy, recall, and F1 score, are calculated, and optimization is performed based on the calculated evaluation metrics.

[0167] The accuracy refers to the proportion of correctly predicted samples out of the total number of samples; the calculation formula is: Wherein, TP represents true positives, which is the number of samples that are actually positive and predicted as positive; TN represents true negatives, which is the number of samples that are actually negative and predicted as negative; FP represents false positives, which is the number of samples that are actually negative but predicted as positive; and FN represents false negatives, which is the number of samples that are actually positive but predicted as negative.

[0168] Recall is the proportion of correctly predicted positive samples to the actual number of positive samples, and it is calculated using the following formula:

[0169] The F1 score is the harmonic mean of precision and recall, which takes into account the balance between precision and recall. The calculation formula is as follows:

[0170] The test set feature vectors are input into the trained deep support vector machine (DSVM) model after evaluation and optimization, and the predicted classification results are obtained by following the forward propagation steps described above; the predicted fault diagnosis results are output in a clear and easy-to-understand manner.

[0171] A wind turbine bearing fault diagnosis system includes a vibration data acquisition module, an enhanced combined differential morphological filter module, an inherent time-scale decomposition module, a deep support vector machine module, and a diagnosis result output module.

[0172] The vibration data acquisition module acquires and preprocesses vibration data of the wind turbine bearing through D / SENL020 and D / SENN020 sensors.

[0173] The enhanced combined differential morphological filter module performs noise reduction processing on the collected wind turbine bearing vibration signal through the enhanced combined differential morphological filter ECGMF.

[0174] The inherent time scale decomposition module uses the inherent time decomposition (ITD) algorithm to decompose the denoised vibration signal, extract the instantaneous amplitude of the main components and perform spectrum analysis to obtain the fault characteristics of the wind turbine bearing.

[0175] The deep support vector machine module combines support vector machine (SVM) with a neural network model to construct a deep support vector machine (DSVM) model.

[0176] The diagnostic result output module inputs the feature vector into the deep support vector machine (DSVM) model for training, performs fault judgment on the wind turbine bearing, and outputs the diagnostic result.

[0177] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for diagnosing bearing faults in wind turbine generators, characterized in that, Includes the following steps: Vibration data of wind turbine generator bearings are collected using D / SENL020 and D / SENN020 sensors and preprocessed. The vibration data includes axial vibration data of the main bearing, horizontal vibration data of the main bearing, vertical vibration data of the main bearing, vertical vibration data of the rear bearing, axial vibration data of the stator, and horizontal vibration data of the stator. The preprocessing includes data visualization inspection and data normalization. The collected wind turbine bearing vibration signal was denoised by enhancing the combined differential morphological filter ECGMF. Based on data visualization analysis and understanding of the vibration source characteristics, the multi-scale range to be used is determined; a flat structural element with zero height is selected. For each selected scale value, set the corresponding structural element size; clarify the combination method of various mathematical morphological operations involved in the enhanced combined differential morphological filter ECGMF transformation at the current scale; and perform mathematical morphological operations on the preprocessed vibration signal in sequence. The mathematical morphology consists of four basic operators: dilation, erosion, opening, and closing; assuming the original one-dimensional signal... The definition of structural elements The definition is: where, It is used to represent one-dimensional signals The independent variable is the discrete sampling point location index or sequence number; It is used to represent structural elements Define a variable representing the position index or ordinal number of an element within a domain; then the four basic operators are defined as follows: Dilation operator: ; Corrosion operator: ; Opening the operator: ; Closing operator: ; in, —Inflation operator; —Erosion operator; — Opening operator; — Closing operator; The erosion operator can reduce the peak value of a signal and amplify the trough value of the signal, while the closing operator can suppress the negative pulse of the signal; conversely, the dilation operator can reduce the trough value of the signal and amplify the peak value of the signal, while the opening operator can suppress the positive pulse of the signal. Based on this property, two cascade operators consisting of expansion and closure operators are defined as follows: ； ; A cascaded operator consisting of an erosion operator and an opening operator is defined as follows: ； and The ability to suppress negative pulses and extract positive pulse signals is enhanced; conversely, The ability to suppress positive pulses and extract negative pulses is enhanced, thus allowing the definition of new difference operators: ； ； Considering that early fault information of wind turbine bearings is often severely interfered with by background noise, in order to extract weak fault feature information from strong background noise, it is necessary to combine... Based on the characteristics of the operator, a new enhanced combinatorial difference morphology operator is proposed: ； At each scale, the vibration signal after the current-scale enhanced combined differential morphological filter (ECGMF) transformation is analyzed in terms of spectrum. Based on the failure mechanism of the target equipment, past experience, and relevant theoretical research, determine the characteristic frequencies related to equipment failure; For the spectrum at each scale, the characteristic frequency intensity coefficients are calculated. The definition is as follows: ； in, ——No. The amplitude of each fault frequency; --Every The amplitude at each point; —Fault frequency and total spectrum quantity; Since the fault characteristic frequency is relatively less affected by interference, the characteristic frequency intensity coefficient is selected. To determine the optimal scale of the structural element; the required characteristic frequency intensity coefficients. The larger the value, the more obvious the fault characteristic signal extraction; therefore, the largest characteristic frequency intensity coefficient is used. The scale corresponding to the value is taken as the optimal scale of the structuring element; Based on the optimal scale of the determined structural element, set the corresponding structural element parameters; redetermine the various mathematical morphological operations involved in the enhanced combined differential morphological filter ECGMF transformation at the optimal scale; substitute the preprocessed original vibration signal into the enhanced combined differential morphological filter ECGMF for mathematical morphological operations to complete the denoising process. The Inherent Time Decomposition (ITD) algorithm is used to decompose the denoised vibration signal, extract the instantaneous amplitude of the main components and perform spectrum analysis to obtain the fault characteristics of the wind turbine bearing. A deep support vector machine (DSVM) model is constructed by combining support vector machine (SVM) with a neural network model. The feature vectors are input into a deep support vector machine (DSVM) model for training, and the model is used to diagnose the faults of wind turbine bearings and output the diagnostic results.

2. The method for diagnosing wind turbine bearing faults according to claim 1, characterized in that, The feature vectors are input into a deep support vector machine (DSVM) model for training, and fault diagnosis is performed on the wind turbine bearings. The diagnostic results are output, including: The extracted bearing fault feature vector dataset is divided into training set, validation set and test set. The feature vectors of the training set are input into the deep support vector machine (DSVM) model. After feature mapping in the hidden layer, the prediction result is obtained by the output layer support vector machine (SVM) classification. Based on the prediction results and the true labels, the loss value is calculated according to the loss function of the deep support vector machine (DSVM) model. Then, the backpropagation algorithm is used to update the parameters of each layer of the model in turn, and the training is iterated repeatedly until the stopping condition is met. During or after iterative training, the model is input with the feature vector of the validation set to obtain the prediction results. The evaluation metrics, including accuracy, recall, and F1 score, are calculated in combination with the corresponding real labels, and the model is optimized accordingly. Finally, the feature vectors of the test set are input into the optimized model to obtain the prediction results, and the fault diagnosis results are output in a clear manner for bearing fault diagnosis.

3. A wind turbine bearing fault diagnosis system, applied to the wind turbine bearing fault diagnosis method described in claim 1 or 2, characterized in that, It includes a vibration data acquisition module, an enhanced combined differential morphological filter module, an inherent time-scale decomposition module, a deep support vector machine module, and a diagnostic result output module; The vibration data acquisition module collects and preprocesses wind turbine bearing vibration data using D / SENL020 and D / SENN020 sensors. The vibration data includes axial vibration data of the main bearing, horizontal vibration data of the main bearing, vertical vibration data of the main bearing, vertical vibration data of the rear bearing, axial vibration data of the stator, and horizontal vibration data of the stator. The preprocessing includes data visualization inspection and data normalization. The enhanced combined differential morphological filter module performs noise reduction processing on the collected wind turbine bearing vibration signal through the enhanced combined differential morphological filter ECGMF; based on data visualization analysis and understanding of the vibration source characteristics, it determines the multi-scale range to be used; and selects a flat structural element with zero height. For each selected scale value, set the corresponding structural element size; clarify the combination method of various mathematical morphological operations involved in the enhanced combined differential morphological filter ECGMF transformation at the current scale; and perform mathematical morphological operations on the preprocessed vibration signal in sequence. The mathematical morphology consists of four basic operators: dilation, erosion, opening, and closing; assuming the original one-dimensional signal... The definition of structural elements The definition of ;in, It is used to represent one-dimensional signals The independent variable is the discrete sampling point location index or sequence number; It is used to represent structural elements Define a variable representing the position index or ordinal number of an element within a domain; then the four basic operators are defined as follows: Dilation operator: ; Corrosion operator: ; Opening the operator: ; Closing operator: ; in, —Inflation operator; —Erosion operator; — Opening operator; — Closing operator; The erosion operator can reduce the peak value of a signal and amplify the trough value of the signal, while the closing operator can suppress the negative pulse of the signal; conversely, the dilation operator can reduce the trough value of the signal and amplify the peak value of the signal, while the opening operator can suppress the positive pulse of the signal. Based on this property, two cascade operators consisting of expansion and closure operators are defined as follows: ； ； A cascaded operator consisting of an erosion operator and an opening operator is defined as follows: ； and The ability to suppress negative pulses and extract positive pulse signals is enhanced; conversely, The ability to suppress positive pulses and extract negative pulses is enhanced, thus allowing the definition of new difference operators: ； ； Considering that early fault information of wind turbine bearings is often severely interfered with by background noise, in order to extract weak fault feature information from strong background noise, it is necessary to combine... and Based on the characteristics of the operator, a new enhanced combinatorial difference morphology operator is proposed: ； At each scale, the vibration signal after the current-scale enhanced combined differential morphological filter (ECGMF) transformation is analyzed in terms of spectrum. Based on the failure mechanism of the target equipment, past experience, and relevant theoretical research, determine the characteristic frequencies related to equipment failure; For the spectrum at each scale, the characteristic frequency intensity coefficients are calculated. The definition is as follows: ； in, ——No. The amplitude of each fault frequency; — Each The amplitude of the point; —Fault frequency and total spectrum quantity; Since the fault characteristic frequency is relatively less affected by interference, the characteristic frequency intensity coefficient is selected. To determine the optimal scale of the structural element; the required characteristic frequency intensity coefficients. The larger the value, the more obvious the fault characteristic signal extraction; therefore, the largest characteristic frequency intensity coefficient is used. The scale corresponding to the value is taken as the optimal scale of the structuring element; Based on the optimal scale of the determined structural element, set the corresponding structural element parameters; redetermine the various mathematical morphological operations involved in the enhanced combined differential morphological filter ECGMF transformation at the optimal scale; substitute the preprocessed original vibration signal into the enhanced combined differential morphological filter ECGMF for mathematical morphological operations to complete the denoising process. The inherent time scale decomposition module uses the inherent time decomposition (ITD) algorithm to decompose the denoised vibration signal, extract the instantaneous amplitude of the main components and perform spectrum analysis to obtain the fault characteristics of the wind turbine bearing. The deep support vector machine module combines support vector machine (SVM) with a neural network model to construct a deep support vector machine (DSVM) model. The diagnostic result output module inputs the feature vector into the deep support vector machine (DSVM) model for training, performs fault judgment on the wind turbine bearing, and outputs the diagnostic result.

Images