Wind power gear crack online monitoring method, system and equipment based on acoustic emission and medium

Abstract

The invention discloses a wind power gear crack online monitoring method, system and equipment based on acoustic emission and a medium. The method comprises the steps that an acoustic emission sensor is installed on a gearbox; acquiring an acoustic emission source signal through an acoustic emission sensor; denoising the acoustic emission source signal to obtain a time-frequency domain feature; according to the method, the system, the equipment and the medium, the cracks of the wind power gear can be accurately monitored on line.

Description

Technical Field

[0001] The invention belongs to the field of wind power generation, and relates to an online monitoring method, system, equipment and medium for cracks in wind power gears based on acoustic emission. Background Art

[0002] In a complex and harsh working environment, the transmission gears of wind turbines need to withstand great shear stress and impact force due to their high operating speed and heavy load. They will be damaged under cyclic load conditions. This damage is called gear fatigue damage.

[0003] Gear fatigue damage generally produces cracks. Currently, the methods for detecting gear cracks basically require on-site tests and shutdown inspections. These methods not only bring inconvenience to production, but also lack real-time performance. When cracks are detected, they are already very serious. If cracks can be found in the early stage of crack generation and reasonable repair and reinforcement measures are taken, the generation of serious cracks can be effectively avoided, ensuring the continuous, efficient and stable operation of wind turbines, while reducing unnecessary maintenance costs. Summary of the invention

[0004] The purpose of the present invention is to overcome the shortcomings of the above-mentioned prior art and provide a method, system, device and medium for online monitoring of wind turbine gear cracks based on acoustic emission. The method, system, device and medium can accurately monitor wind turbine gear cracks online.

[0005] In order to achieve the above object, the present invention adopts the following technical scheme:

[0006] In one aspect, the present invention provides an online monitoring method for cracks in wind turbine gears based on acoustic emission, comprising:

[0007] Install an acoustic emission sensor on the gearbox;

[0008] Acquiring acoustic emission source signals through an acoustic emission sensor;

[0009] Denoising the acoustic emission source signal to obtain time-frequency domain features;

[0010] Evaluate whether cracks occur in wind turbine gears based on time-frequency domain characteristics.

[0011] The on-line crack monitoring method for wind turbine gears based on acoustic emission of the present invention is further improved in that:

[0012] Furthermore, the process of denoising the acoustic emission source signal to obtain time-frequency domain features is as follows:

[0013] Decomposing the acoustic emission source signal using independent component separation to determine noise;

[0014] Removing certain noise from the acoustic emission source signal to obtain a denoised acoustic emission source signal;

[0015] The denoised acoustic emission source signal is decomposed by wavelet packet to obtain the time-frequency domain characteristics.

[0016] Furthermore, the process of evaluating the cracks of wind turbine gears according to the time-frequency domain characteristics is as follows:

[0017] The time-frequency domain features are input into a kernel extreme learning machine to identify whether cracks occur in the wind turbine gear.

[0018] Furthermore, the output weights and output results of the kernel extreme learning machine are:

[0019]

[0020] Among them, I is a diagonal matrix; C is the penalty coefficient; P is the number of training samples.

[0021] In a second aspect of the present invention, the present invention provides an online monitoring system for cracks in wind turbine gears based on acoustic emission, comprising:

[0022] A mounting module for mounting an acoustic emission sensor on a gearbox;

[0023] An acquisition module, used for acquiring an acoustic emission source signal through an acoustic emission sensor;

[0024] A denoising module, used to denoise the acoustic emission source signal to obtain time-frequency domain features;

[0025] The evaluation module is used to evaluate whether cracks occur in wind turbine gears based on time-frequency domain characteristics.

[0026] The wind turbine gear crack online monitoring system based on acoustic emission described in the present invention is further improved in that:

[0027] Furthermore, the denoising module includes:

[0028] A decomposition module, used for decomposing the acoustic emission source signal using independent component separation to determine noise;

[0029] A removal module, used to remove the determined noise from the acoustic emission source signal to obtain a denoised acoustic emission source signal;

[0030] The wavelet packet decomposition module is used to perform wavelet packet decomposition on the denoised acoustic emission source signal to obtain time-frequency domain features.

[0031] Furthermore, the process of evaluating the cracks of wind turbine gears according to the time-frequency domain characteristics is as follows:

[0032] The time-frequency domain features are input into a kernel extreme learning machine to identify whether cracks occur in the wind turbine gear.

[0033] Furthermore, the output weights and output results of the kernel extreme learning machine are:

[0034]

[0035] Among them, I is a diagonal matrix; C is the penalty coefficient; P is the number of training samples.

[0036] In a third aspect of the present invention, the present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method for autonomous navigation of live operations in a distribution network when executing the computer program.

[0037] In a fourth aspect of the present invention, the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the autonomous navigation method for live operations in a distribution network are implemented.

[0038] The present invention has the following beneficial effects:

[0039] In the specific operation of the wind turbine gear crack online monitoring method, system, equipment and medium based on acoustic emission described in the present invention, when cracks occur in the wind turbine gear, an acoustic emission source signal will be generated. The acoustic emission source signal is transmitted to the surface of the gear box through media such as gears and gear boxes and causes slight vibrations. Based on the above principle, the present invention installs an acoustic emission sensor on the gear box to obtain the acoustic emission source signal through the acoustic emission sensor; denoises the acoustic emission source signal to obtain time-frequency domain characteristics; evaluates whether cracks occur in the wind turbine gear according to the time-frequency domain characteristics, so as to timely identify when cracks occur in the transmission gear of the wind turbine generator, thereby reducing the risk of abnormal operation of the unit. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the accompanying drawings:

[0041] Figure 1 is a flow chart of the method of the present invention;

[0042] Figure 2 It is a roadmap for realizing the present invention;

[0043] Figure 3 It is the original collected acoustic emission signal diagram;

[0044] Figure 4 This is the acoustic emission signal diagram after noise reduction. DETAILED DESCRIPTION

[0045] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that the embodiments and features in the embodiments of the present application can be combined with each other without conflict.

[0046] The following detailed description is an exemplary description, which is intended to provide further detailed description of the present invention. Unless otherwise specified, all technical terms used in the present invention have the same meaning as those generally understood by those skilled in the art to which the present application belongs. The terms used in the present invention are only for describing specific embodiments, and are not intended to limit exemplary embodiments according to the present invention.

[0047] Embodiment 1

[0048] refer to Figures 1 to 4 The wind turbine gear crack online monitoring method based on acoustic emission of the present invention comprises the following steps:

[0049] When cracks appear in the fan gear, an acoustic emission source signal is generated. The acoustic emission source signal is transmitted to the surface of the gear box through media such as gears and gear boxes and causes slight vibrations. The acoustic emission sensor installed on the gear box captures the vibration signal and converts it into an electrical signal, which is amplified to obtain the acoustic emission signal.

[0050] Specifically, considering signal attenuation, the acoustic emission sensor is installed on the gear.

[0051] When the wind turbine is running, the collected gear acoustic emission signals contain noise from gear meshing, generator operation, wind turbine vibration, etc., and noise reduction processing is required to obtain usable signal data.

[0052] Specifically, the acoustic emission signal is first decomposed using independent component analysis (ICA) to remove the determined noise, and then the remaining signal is decomposed using wavelet packets to achieve the purpose of noise reduction.

[0053] Specifically, ICA analysis is a signal processing technology that has emerged in recent years. It has the advantages of not being affected by the aliasing of frequency bands between source signals, and not being interfered by the different strengths of source signals. The ICA method is: assuming that the observed data is a linear combination of the source signals, when the source signal has at most one Gaussian component and is statistically independent of each other, the ICA method can successfully separate the source signals. At the same time, the ICA transform can enhance the non-Gaussianity of the signal, which is conducive to the sparse coding of the signal. On this basis, the basis function extracted by ICA is used as a filter, and the threshold method can be used to effectively denoise the signal.

[0054] Specifically, the ICA independent source separation principle is:

[0055] When the number of noise components and independent components is not very large, especially when their total number is not greater than the number of mixed signals, the basic ICA model can be used to estimate the independent components. Let the independent component S = [S1,…,Sk,N1,…Nl] T , where Si,i=1,…,k is a useful independent component, and Ni,i=1,…,l is a noise variable, that is, the number of mixed signals is k+l. X=[X1,X2,…,Xn] T is a set of observation signals, where n≥k+l, and each component in X is linearly combined by each independent source in S through the mixing matrix A, that is:

[0056] X=AS

[0057] The basic principle of ICA is: when the coefficient matrix A and the source matrix S are unknown, the source signals are separated from the observed signal X by unmixing the coefficient matrix W, so that the output Y is the optimal approximation of S, that is:

[0058] Y=WX

[0059] Among them, A is a mixing matrix combining the true independent components and the noise covariance structure. Therefore, by finding m directions with extremely large non-Gaussianity, the true independent components can be estimated.

[0060] Based on the fixed-point fast ICA algorithm, the denoising steps are as follows:

[0061] 11) Determine the number m of independent components to be estimated;

[0062] 12) Select initial weights: Randomly select a weight matrix as the starting point;

[0063] 13) Initialize all w i , i = 1,…,m, where each w i All have unit norm.

[0064] 14) Iterative Optimization: Use the gradient ascent optimization algorithm to maximize the statistical independence of independent components.

[0065] 15) Convergence test: Check whether the algorithm has converged to stable weights, usually by checking whether the weight change is below a certain threshold.

[0066] 16) Calculate independent components: Use the final weight matrix to transform the observed data to obtain independent components.

[0067] The steps of the iterative optimization part are:

[0068] 21) Objective function:

[0069] In ICA, non-Gaussian measures are usually used as objective functions. Common ones are negative entropy or fourth-order cumulant (kurtosis). Negentropy expression: Negentropy(Y) = E[h(Y)]-h(E[Y]), where Y is the signal, E is the expected value, and h(Y) is the entropy of signal Y;

[0070] 22) Initial weight;

[0071] The optimization process starts with an initial weight matrix that is randomly selected or determined by some heuristic method;

[0072] 23) Gradient calculation;

[0073] At each iteration, the gradient of the objective function with respect to the weight matrix is ​​calculated.

[0074] 24) Optimization algorithm;

[0075] Use the gradient ascent method to increase the value of the objective function by adjusting the weights in the direction of the gradient.

[0076] 25) Learning rate;

[0077] The learning rate determines the size of the step in the direction of the gradient.

[0078] 26) Update rules;

[0079] In each iteration, the calculated gradient and learning rate are used to update the weight matrix. The update rule is usually: weights = weights + learning rate * gradient.

[0080] 27) Convergence test;

[0081] After each iteration, check whether the change in the weight matrix is ​​less than a certain threshold, or whether the increase in the objective function is no longer significant. When the convergence condition is met, stop the iteration.

[0082] 28) Output independent components;

[0083] Once the algorithm converges, the final weight matrix is ​​used to obtain the independent components through linear transformation.

[0084] After the initial denoising by ICA, the acoustic emission signal is further decomposed by wavelet packet to achieve deep denoising. The specific process is as follows:

[0085] The noise-reduced acoustic emission signals before cracks are generated and the noise-reduced acoustic emission signals when cracks are generated, which are collected multiple times, are constructed into a diagnostic database.

[0086] The denoised acoustic emission signal is extracted through feature extraction to obtain time-frequency domain features.

[0087] The time-frequency domain features are input into the kernel extreme learning machine to identify whether cracks occur. The specific process is as follows:

[0088] 1a) 70% of the diagnosis database is used for training the kernel extreme learning machine; 30% is used for testing the kernel extreme learning machine.

[0089] 2a) Kernel extreme learning machine diagnosis classification step;

[0090] Kernel Extreme Learning Machines (KELM) is an improvement of Extreme Learning Machines (ELM) combined with kernel functions. KELM replaces the random mapping of ELM with a pre-selected kernel function to improve the generalization and stability degradation caused by random assignment of hidden layer neurons and improve the calculation accuracy of the model.

[0091] The kernel extreme learning machine consists of an input layer, a hidden layer and an output layer, where x 1 、x 2 to x m is the input information; 1 ,y 2 To n is the output information, and the output matrix is ​​Y; b is the number of neurons in the hidden layer 1 、b 2 To b l The threshold matrix composed of ij is the weight between the i-th input layer neuron and the j-th hidden layer neuron, and the weight matrix composed of the above weights is w; similarly, β ij is the weight between the jth hidden layer neuron and the kth output layer neuron, and the weight matrix is ​​constructed as β.

[0092] The threshold matrix, the weight matrix between the input layer and the hidden layer, and the weight matrix between the hidden layer and the output layer are expressed as:

[0093]

[0094] The ELM is trained by training samples. Suppose there are P input samples (X, Y), X and Y are the input matrix and output matrix, and the dimensions of X and Y are m and n respectively, expressed as:

[0095]

[0096] Use X to find the output matrix Q:

[0097]

[0098] Among them, w i =[wi1 ,w i2 ,…,w in ]; x j =[x 1j ,x 2j ,…,x nj ] T , g(x) is the activation function of a single hidden layer, and the above formula is simplified to:

[0099] Hβ=Q

[0100] Among them, H is the output matrix of a single hidden layer. ELM minimizes the error Y between the output matrix Q and the actual output matrix of the sample by finding the optimal weight combination:

[0101]

[0102] Finally, once the input layer weight matrix w and threshold matrix b are obtained, the hidden layer output matrix H will be uniquely determined. At this time, the optimal solution of ELM is:

[0103]

[0104] Among them, H + It is the generalized inverse of Moore-Penrose of the output matrix H. Therefore, ELM uses the training data to find the optimal weight combination. Given the sample input X that needs to be predicted or classified, X and the optimal weight combination are input into ELM together to get the output result Q.

[0105] The hidden layer output matrix H in ELM is replaced by the kernel function matrix Ω, which becomes a kernel extreme learning machine. The KELM obtained in this way not only retains the advantages of ELM, but also improves its generalization performance and stability. The KELM kernel matrix Ω ELM It is expressed as:

[0106]

[0107] Among them, h(x) is the hidden layer output function, K(x i ,x j ) is the kernel function, and the Gaussian kernel function with fewer parameters and strong versatility is selected. The output weights and output results of KELM are:

[0108]

[0109] Among them, I is a diagonal matrix; C is the penalty coefficient; P is the number of training samples.

[0110] Embodiment 2

[0111] The wind turbine gear crack online monitoring system based on acoustic emission of the present invention comprises:

[0112] A mounting module for mounting an acoustic emission sensor on a gearbox;

[0113] An acquisition module, used for acquiring an acoustic emission source signal through an acoustic emission sensor;

[0114] A denoising module, used to denoise the acoustic emission source signal to obtain time-frequency domain features;

[0115] The evaluation module is used to evaluate whether cracks occur in wind turbine gears based on time-frequency domain characteristics.

[0116] The wind turbine gear crack online monitoring system based on acoustic emission described in the present invention is further improved in that:

[0117] Furthermore, the denoising module includes:

[0118] A decomposition module, used for decomposing the acoustic emission source signal using independent component separation to determine noise;

[0119] A removal module, used to remove the determined noise from the acoustic emission source signal to obtain a denoised acoustic emission source signal;

[0120] The wavelet packet decomposition module is used to perform wavelet packet decomposition on the denoised acoustic emission source signal to obtain time-frequency domain features.

[0121] Furthermore, the process of evaluating the cracks of wind turbine gears according to the time-frequency domain characteristics is as follows:

[0122] The time-frequency domain features are input into a kernel extreme learning machine to identify whether cracks occur in the wind turbine gear.

[0123] Furthermore, the output weights and output results of the kernel extreme learning machine are:

[0124]

[0125] Among them, I is a diagonal matrix; C is the penalty coefficient; P is the number of training samples.

[0126] The division of modules in the embodiments of the present application is schematic and is only a logical function division. There may be other division methods in actual implementation. In addition, each functional module in each embodiment of the present application may be integrated into a processor, or may exist physically separately, or two or more modules may be integrated into one module. The above-mentioned integrated modules may be implemented in the form of hardware or in the form of software functional modules.

[0127] Embodiment 3

[0128] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the online monitoring method for cracks in wind turbine gears based on acoustic emission are implemented, including: installing an acoustic emission sensor on a gearbox; obtaining an acoustic emission source signal through the acoustic emission sensor; denoising the acoustic emission source signal to obtain time-frequency domain features; and evaluating whether cracks occur in wind turbine gears according to the time-frequency domain features. The memory may include a memory, such as a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk memory, etc. The processor, the network interface, and the memory are interconnected through an internal bus, which may be an industrial standard architecture bus, a peripheral component interconnection standard bus, an extended industrial standard architecture bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. The memory is used to store programs. Specifically, the program may include a program code, and the program code includes computer operation instructions. The memory may include a memory and a non-volatile memory, and provide instructions and data to the processor.

[0129] Embodiment 4

[0130] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the online monitoring method for wind turbine gear cracks based on acoustic emission are implemented, including: installing an acoustic emission sensor on a gearbox; acquiring an acoustic emission source signal through the acoustic emission sensor; denoising the acoustic emission source signal to obtain time-frequency domain characteristics; and evaluating whether a wind turbine gear has cracks based on the time-frequency domain characteristics. Specifically, the computer-readable storage medium includes, but is not limited to, for example, volatile memory and / or non-volatile memory. The volatile memory may include a random access memory (RAM) and / or a cache memory (cache), etc. The non-volatile memory may include a read-only memory (ROM), a hard disk, a flash memory, an optical disk, a magnetic disk, etc.

[0131] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0132] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0133] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0134] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0135] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. An online monitoring method for wind turbine gear cracks based on acoustic emission, characterized in that: include: Install an acoustic emission sensor on the gearbox; Acquiring acoustic emission source signals through an acoustic emission sensor; Denoising the acoustic emission source signal to obtain time-frequency domain features; Evaluate whether cracks occur in wind turbine gears based on time-frequency domain characteristics.

2. The method for online monitoring cracks of wind turbine gears based on acoustic emission according to claim 1 is characterized in that: The process of denoising the acoustic emission source signal to obtain time-frequency domain features is as follows: Decomposing the acoustic emission source signal using independent component separation to determine noise; Removing certain noise from the acoustic emission source signal to obtain a denoised acoustic emission source signal; The denoised acoustic emission source signal is decomposed by wavelet packet to obtain the time-frequency domain characteristics.

3. The online monitoring method for wind turbine gear cracks based on acoustic emission according to claim 1 is characterized in that: The process of evaluating the cracks of wind turbine gears according to the time-frequency domain characteristics is as follows: The time-frequency domain features are input into a kernel extreme learning machine to identify whether cracks occur in the wind turbine gear.

4. The method for online crack monitoring of wind turbine gears based on acoustic emission according to claim 3 is characterized in that: The output weights and output results of the kernel extreme learning machine are: Among them, I is a diagonal matrix; C is the penalty coefficient; P is the number of training samples.

5. An online monitoring system for wind turbine gear cracks based on acoustic emission, characterized in that: include: A mounting module for mounting an acoustic emission sensor on a gearbox; An acquisition module, used for acquiring an acoustic emission source signal through an acoustic emission sensor; A denoising module, used to denoise the acoustic emission source signal to obtain time-frequency domain features; The evaluation module is used to evaluate whether cracks occur in wind turbine gears based on time-frequency domain characteristics.

6. The wind turbine gear crack online monitoring system based on acoustic emission according to claim 5 is characterized in that: The denoising module comprises: A decomposition module, used for decomposing the acoustic emission source signal using independent component separation to determine noise; A removal module, used to remove the determined noise from the acoustic emission source signal to obtain a denoised acoustic emission source signal; The wavelet packet decomposition module is used to perform wavelet packet decomposition on the denoised acoustic emission source signal to obtain time-frequency domain features.

7. The wind turbine gear crack online monitoring system based on acoustic emission according to claim 5 is characterized in that: The process of evaluating the cracks of wind turbine gears according to the time-frequency domain characteristics is as follows: The time-frequency domain features are input into a kernel extreme learning machine to identify whether cracks occur in the wind turbine gear.

8. The wind turbine gear crack online monitoring system based on acoustic emission according to claim 7 is characterized in that: The output weights and output results of the kernel extreme learning machine are: Among them, I is a diagonal matrix; C is the penalty coefficient; P is the number of training samples.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method for autonomous navigation of live-line operations in a distribution network as described in any one of claims 1 to 4 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method for autonomous navigation of live-line operations in a distribution network as described in any one of claims 1 to 4 are implemented.

Images