A detection method, device and equipment for small defects at the connection of wind turbine blades

Abstract

The present invention discloses a detection method, device and equipment for tiny defects at the connection of wind turbine blades. Among them, the method includes: calculating the optimal working frequency of an eddy current array sensor; obtaining different fatigue data sequences of wind turbine blades through the optimal working frequency, converting the fatigue data sequences of wind turbine blades into equidistant sequences, and extracting the midpoint value data therein; constructing a fatigue life monitoring model through different fatigue data sequences of wind turbine blades; training the fatigue life monitoring model by using a deep learning network and the midpoint value data; and predicting tiny defects at the connection of wind turbine blades according to the trained fatigue life monitoring model. The present invention ensures the accuracy of data by using the optimal working frequency, and the midpoint value is used to update the deep learning network in the reverse direction during the training process to ensure the convergence speed and accuracy of the model. By using the above technical means, the accuracy of the model is improved, so that it can monitor tiny structural defects through an eddy current array sensor.

Description

Technical Field

[0001] The present invention relates to the technical field of information-based detection of bolts on wind turbine blades, and particularly to a detection method, device and equipment for minute defects at the connection of wind turbine blades. Background Art

[0002] Corrosion has always been regarded as a fatal threat during the aging process of blades. Due to aging, the ability to monitor the structural corrosion of blades is insufficient, which seriously affects the safety management of wind turbine blades by major wind power companies. During the aging process of blades, since the stress on the blade root is large, higher requirements are imposed on the bolts thereon. Once defects appear at the bolt edges, serious losses will be caused under the action of long-term large stress. Currently, the detection of blades mostly uses an eddy current sensor array structure. This structure can not only achieve the microscopic expansion of the detection space of traditional eddy current detection technology, but also improve the detection sensitivity, macroscopically expand the detection space, and improve the detection speed. However, the existing eddy current sensor array scheme has low sensitivity to minute defects at the blade root; this is because the signals of minute defects are relatively weak, and it is difficult to judge the detection results based on the sensing signals measured by the eddy current array sensor alone. Therefore, the applicant proposes a detection method, device and equipment for minute defects at the connection of wind turbine blades. Summary of the Invention

[0003] In view of this, the purpose of the present invention is to propose a detection method, device and equipment for minute defects at the connection of wind turbine blades, which can realize the monitoring of minute structural defects of wind turbine blades through an eddy current sensor array.

[0004] According to one aspect of the present invention, a detection method for minute defects at the connection of wind turbine blades is provided, including: calculating the optimal operating frequency β(v, L1) of the eddy current array sensor; obtaining different fatigue data sequences of wind turbine blades through the optimal operating frequency β(σ, L1), converting the fatigue data sequences of the wind turbine blades into equidistant sequences, and extracting the midpoint value data therein; constructing a fatigue life monitoring model based on the different fatigue data sequences of the wind turbine blades; training the constructed fatigue life monitoring model by using a deep learning network and the midpoint value data; and predicting the minute defects at the connection of the wind turbine blades according to the trained fatigue life monitoring model.

[0005] According to another aspect of the present invention, there is provided a detection device for tiny defects at the connection of a wind turbine blade, comprising: an analysis module, a database module, a construction module, a deep learning module, and a prediction module; Analysis module: calculating the optimal operating frequency β(σ, L1) of the eddy current array sensor; Database module: obtaining different fatigue data sequences of wind turbine blades through the optimal operating frequency β(σ, L1), converting the fatigue data sequences of the wind turbine blades into equidistant sequences, and extracting the midpoint value data therein; Construction module: constructing a fatigue life monitoring model based on the different fatigue data sequences of wind turbine blades; Deep learning module: training the constructed fatigue life monitoring model using a deep learning network and the midpoint value data; Prediction module: predicting the tiny defects at the connection of the wind turbine blade according to the trained fatigue life monitoring model.

[0006] According to yet another aspect of the present invention, there is provided a detection device for tiny defects at the connection of a wind turbine blade, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the detection method for tiny defects at the connection of a wind turbine blade as described in any one of the above.

[0007] It can be found that in the above solution, the optimal operating frequency β(σ, L1) of the eddy current array sensor is calculated; through such calculation, it can be ensured that the signal of tiny defects can be effectively detected during the data collection process, ensuring the effectiveness of the data, and solving the problem that it is difficult to judge the detection result of the sensing signal measured by the eddy current sensor array alone due to the relatively weak signal of the tiny defects. According to the operating mechanism of the eddy current array sensor, the accuracy of the automatic monitoring model depends on the transformation coefficient, and the transformation coefficient depends on the background value in the interval [g i , g i+1 . During the model training process, the background value takes S (1) (g i ) and S (1) (g g+1)'s midpoint value, and subtracting the midpoint value from the training result and backpropagating it to the network to update the network parameters can make the model have a small monitoring error when the data sequence changes smoothly, ensure that the model is not affected by the sequence change, greatly reduce the monitoring error, and improve the automatic monitoring accuracy of the model. The signal may have different degrees of offset (polynomial trend). Using the least squares method to cancel the trend term can reduce noise and eliminate the offset, and reduce the error. The present invention uses the optimal operating frequency β(σ,L1) to ensure the accuracy of the data. The midpoint value is used for backpropagation to update the deep learning network during the training process to ensure the convergence speed and accuracy of the model. The least squares method is used to solve the problem of the noise of the original data. The prediction accuracy of the fatigue life monitoring model is improved from three technical means of data, preprocessing, and backpropagation method, so that it can realize the monitoring of the small structural defects of the wind turbine blade through the eddy current sensor array. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0009] Figure 1 FIG. is a schematic flow chart of an embodiment of the method for detecting small defects at the connection of the wind turbine blade of the present invention;

[0010] Figure 2 FIG. is a schematic structural diagram of an embodiment of the device for detecting small defects at the connection of the wind turbine blade of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0011] The present invention will be further described in detail below with reference to the drawings and embodiments. It should be particularly noted that the following embodiments are only used to illustrate the present invention, but do not limit the scope of the present invention. Similarly, the following embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present invention.

[0012] The present invention provides a method for detecting small defects at the connection of a wind turbine blade, which can realize the monitoring of small structural defects of the wind turbine blade through an eddy current sensor array.

[0013] Please refer to Figure 1 , Figure 1 FIG. is a schematic flow chart of an embodiment of the method for detecting small defects at the connection of the wind turbine blade of the present invention. It should be noted that if there are substantially the same results, the method of the present invention is not limited to Figure 1 the flow sequence shown. For exampleFigure 1 As shown, the method includes the following steps:

[0014] S101. Calculate the optimal operating frequency β(σ, L1) of the eddy current array sensor;

[0015] In this embodiment, the spatial diffusion equation of the magnetic vector B is derived, and the analytical general solution of the magnetic vector B is determined based on this spatial diffusion equation; the magnetic field intensity Q and the electromagnetic field quantity boundary condition at the P section are calculated in the layered medium space based on the analytical general solution of the fixed magnetic vector B; the optimal operating frequency β(σ, L1) is obtained by using the magnetic field intensity Q and the electromagnetic field quantity boundary condition. Specifically, since the construction process of the semi-analytical model of the eddy current array sensor is very complex, it is necessary to first analyze the relationship between the magnetic vector B and the line current density K at the P section. Without considering the change of displacement current, the spatial diffusion equation of the magnetic vector B can be derived as follows:

[0016] L 2 = ωμσ (1)

[0017] Using the method of separation of variables in the layered medium space, the analytical general solution of the magnetic vector B is determined, and the calculation formula is as follows:

[0018]

[0019]

[0020] k y = 2πn / λ, n = 1, 2... (4)

[0021] In the above formula: B Δ,n and B 0,n respectively represent the transformation coefficients of the magnetic vector B at the upper and lower interfaces in the layered medium space; λ represents the wavelength of the excitation distribution in the layered medium space; L represents the thickness of the layered medium space; μ represents the magnetic permeability of the layered medium space; σ represents the conductivity of the layered medium space; ω is the current change period, k x and k y respectively represent the transverse and longitudinal vectors of the magnetic vector, and ε represents the dielectric constant of the layered medium space.

[0022] In the layered medium space, from formula (2), we can get:

[0023] Q = B / μ (5)

[0024] In the above formula: Q represents the magnetic field intensity.

[0025] When the thickness of the layered medium space is infinite, the magnetic field intensity and the magnetic vector can be decoupled, and thus the electromagnetic field quantity boundary condition at the P section can be obtained:

[0026] B z (x = P+ ) = B z (x = P - ) (6)

[0027] Q y (x = P + ) - Q y (x = P - ) = ρ z (x = P) (7)

[0028] In Formulas (6) and (7): B z (x = P - ) represents the magnetic vector at the positive cross-section; Q y (x = P + ) represents the magnetic field strength at the positive cross-section; Q y (x = P - ) represents the magnetic field strength at the reverse cross-section; ρ z (x = P) represents the linear current density at the cross-section B z (x = P + ) represents the magnetic vector at the reverse cross-section. Based on the above formulas and the electromagnetic field quantities at the boundary conditions of the electromagnetic field quantities, the specific values of the magnetic vector at the positive cross-section and the magnetic field strength at the positive cross-section at the x = S cross-section can be determined.

[0029] At the x = S cross-section, by combining the linear current density at the cross-section and the magnetic vector at the reverse cross-section, the transformation coefficient relationship can be determined:

[0030] ρ z,n (x = P) = B z,n (x = P) · β(σ, L1) (8)

[0031] In Formula (8), β(σ, L1) represents the fixed transformation coefficient term, where L1 represents the lift-off distance, that is, the distance between the eddy current array sensor and the y vane. The magnitude of the lift-off distance determines the optimal operating frequency under the action of the electromagnetic field induced by the eddy current array sensor. The larger the value of this distance, the smaller the automation monitoring error.

[0032] After determining the fixed transformation coefficient β(σ, L1) of the eddy current array sensor, a large amount of fatigue data can be obtained. Although these data seem to be determined on the surface, they are actually just a kind of whitened value. Under the same data state, there is no definite relationship information between different data. Therefore, the monitoring accuracy of the model constructed by the original method is not high. Since these original data sequences are required to be smooth, an automated monitoring model for fatigue life needs to be constructed.

[0033] The advantage of such a setting is that since the signals of micro defects are relatively weak, it is difficult to judge the detection results of the sensing signals measured by the eddy current sensor array alone. Therefore, the optimal operating frequency β(σ, L1) of the eddy current array sensor is calculated. Through such calculation, it can be ensured that the signals of micro defects can be effectively detected during the data collection process, ensuring the effectiveness of the data.

[0034] S102. Obtain different fatigue data sequences of wind turbine blades through the optimal operating frequency β(σ, L1), convert the fatigue data sequences of the wind turbine blades into equidistant sequences, and extract the midpoint value data therein.

[0035] Specifically, use the eddy current array sensor with the set optimal operating frequency β(σ, L1) to detect different fatigue data of wind turbine blades; each fatigue data of the wind turbine blade forms a sequence S (0) =(g1, g1, g1,... g i ); and judge whether the sequence is an equidistant sequence. If so, extract the midpoint value data of the sequence; if not, construct a differential equation to convert the sequence into an equidistant sequence and then extract the midpoint value data of the sequence.

[0036] In this embodiment, let S (0) =(g1, g1, g1,... g i ) be a smooth original data sequence. If Δg i =g i+1 -g i , it means that the smooth original data sequence is an equidistant sequence; if Δg i ≠g i+1 -g i , it means that the original data sequence is a non-equidistant sequence. Therefore, S(0) needs to be transformed as follows:

[0037] Where:

[0038] Perform a cumulative sum on S (0) to generate S (1) :

[0039]

[0040] Construct a differential equation on [g i , g i+1 according to formula (9):

[0041]

[0042] According to formula (10), it can be obtained:

[0043]

[0044] Let be s (1)(g) In the interval [g i , g i+1 , the background value can transform (11) into:

[0045] S (0) (g i+1 ) Δg i+1 + az (1) (g i+1 ) = bΔg i+1 (12)

[0046] Through the above transformation, the wind turbine blade fatigue data sequence can be converted into an equidistant sequence. The advantage of this setting is that according to the operating mechanism of the eddy current array sensor, the accuracy of the automatic monitoring model depends on the transformation coefficient, and the transformation coefficient depends on the background value in the interval [g i , g i+1 . During the model training process, the background value takes the midpoint value of S (1) (g i ) and S (1) (g g+1 ), and the difference between the midpoint value and the training result is backpropagated to the network to update the network parameters, which can make the monitoring error smaller when the data sequence changes smoothly, ensure that the model is not affected by the sequence change, greatly reduce the monitoring error, and improve the automatic monitoring accuracy of the model.

[0047] S103. Construct a fatigue life monitoring model based on the different wind turbine blade fatigue data sequences;

[0048] In this embodiment, the fatigue life monitoring model refers to a deep learning model, and there are various ways to construct it, which will not be elaborated here.

[0049] S104. Train the constructed fatigue life monitoring model using a deep learning network and midpoint value data;

[0050] Specifically, use the least squares method to eliminate the trend of the different wind turbine blade fatigue data sequences, use a deep learning network to label the abnormal sequences in the different wind turbine blade fatigue data sequences after eliminating the trend, extract the sequence features of the different wind turbine blade fatigue data sequences from the different wind turbine blade fatigue data sequences after eliminating the trend, and extract the mean, variance, kurtosis, and maximum value from the feature vectors as the training input of the deep learning network, and use the deep learning network with the mean, variance, kurtosis, and maximum value as the training input to train the constructed fatigue life monitoring model, and use the difference between the midpoint value data and the analysis result of the fatigue life monitoring model to backpropagate to the deep learning network to update the deep learning network parameters.

[0051] In this embodiment, deep learning adopts an anomaly detection technology for one-dimensional signals, which includes signal processing in the early stage, signal feature extraction, and a deep learning network.

[0052] Among them, the signal (i.e., different fatigue data sequences of wind turbine blades, the same below) is input into the computer in the form of a one-dimensional signal through a signal collector (i.e., an eddy current array sensor) in the early stage. Since there will be different degrees of noise in the process of signal generation, transmission, and conversion, it is necessary to perform noise reduction processing on the signal. First, since the signal may have different degrees of offset (polynomial trend), the least squares method is used to cancel the trend term. Let the sampling data of the measured vibration signal be x k (k = 1, 2, 3,..., n). Since the sampling data is at equal time intervals, for simplicity, let the sampling interval Δt = 1, and set a polynomial:

[0053]

[0054] where is the fitting data; a i is the undetermined coefficient; m is the order.

[0055] Calculations are performed on all sampling data, and we can obtain:

[0056]

[0057] where:

[0058]

[0059]

[0060] A T = [a0, a1,..., a m

[0061] Combined with the above calculation formula (14), find the partial derivative of E with respect to a i , then we have:

[0062]

[0063] Let the partial derivative be 0, then the trend elimination formula can be obtained:

[0064]

[0065] After that, FFT (Fast Fourier Transform) is used to transform the collected signal x(n) from a time-domain signal to a frequency-domain signal X(k), and its conversion formula is as follows:

[0066]

[0067] ​After that, the mean, variance, kurtosis, and maximum value of the processed signal are extracted as the feature vector of the signal, and the abnormal signal is labeled. The problematic signal is marked as 1, and the signal without problems is marked as 0. The training set is put into the neural network for training. During the training process, the difference between the midpoint value data and the result of the fatigue life monitoring model is used for backpropagation to the deep learning network to update the parameters of the deep learning network. After a finite number of iterative updates, a fatigue life monitoring model is generated. The advantage of this setting is that the signal may have different degrees of offset (polynomial trend). Using the least squares method to cancel the trend term can reduce noise and eliminate the offset, reducing the error. Subtracting the midpoint value from the training result and performing backpropagation to the network to update the network parameters can make the model have a small monitoring error when the data sequence changes smoothly, ensure that the model is not affected by the sequence change, greatly reduce the monitoring error, and improve the automatic monitoring accuracy of the model.

[0068] S105. Predict the small defects at the connection of the wind turbine blade according to the trained fatigue life monitoring model.

[0069] In this embodiment, specifically, according to the trained fatigue life monitoring model, the mean, variance, kurtosis, and maximum value associated with the data are extracted from the data measured by the eddy current sensor array. The extracted mean, variance, kurtosis, and maximum value associated with the data are input into the trained fatigue life monitoring model to predict the small defects corresponding to the blade root of the blade root. The predicted small defects are obtained through the trained fatigue life monitoring model. The advantage of this setting is that the optimal operating frequency β(σ, L1) ensures the accuracy of the data; the midpoint value is used for backpropagating and updating the deep learning network during the training process; the least squares method is used to solve the noise problem. The prediction accuracy of the fatigue life monitoring model is improved from three technical means of data, preprocessing, and backpropagation method, enabling it to monitor the small structural defects of the wind turbine blade through the eddy current sensor array.

[0070] It can be found that in this embodiment, the optimal operating frequency β(σ, L1) of the eddy current array sensor is calculated. By such calculation, it can be ensured that the signal of the small defect can be effectively detected during the data collection process, ensuring the effectiveness of the data, and solving the problem that it is difficult to judge the detection result of the sensing signal measured by the eddy current sensor array alone due to the weak signal of the small defect. According to the operating mechanism of the eddy current array sensor, the accuracy of the automatic monitoring model depends on the transformation coefficient, and the transformation coefficient depends on the background value in the interval [g i , g i+1 . During the model training process, the background value takes S (1) (g i ) and S (1) (gg+1 ) and subtracting the midpoint value from the training result and backpropagating it to the network to update the network parameters can make the model have a small monitoring error when the data sequence changes smoothly, ensure that the model is not affected by the sequence change, greatly reduce the monitoring error, and improve the automatic monitoring accuracy of the model. The signal may have different degrees of offset (polynomial trend). Using the least squares method to cancel the trend term can reduce noise and eliminate the offset, and reduce the error. The present invention uses the optimal working frequency β(σ, L1) to ensure the accuracy of the data. The midpoint value is used for backpropagating and updating the deep learning network during the training process to ensure the convergence speed and accuracy rate of the model. The least squares method is used to solve the problem of the noise of the original data. The prediction accuracy rate of the fatigue life monitoring model is improved from three technical means of data, preprocessing, and backpropagation method, so that it can realize the monitoring of the small structural defects of the wind turbine blade through the eddy current sensor array.

[0071] The present invention also provides a detection device for small defects at the connection of a wind power blade, which can realize the monitoring of small structural defects of the wind turbine blade through an eddy current sensor array.

[0072] Please refer to Figure 2 , Figure 2 which is a schematic structural diagram of an embodiment of the detection device for small defects at the connection of the wind power blade of the present invention. In this embodiment, the detection device for small defects at the connection of the wind power blade includes an analysis module, a database module, a construction module, a deep learning module, and a prediction module;

[0073] Analysis module: Calculate the optimal working frequency β(σ, L1) of the eddy current array sensor;

[0074] Database module: Obtain different fatigue data sequences of wind power blades through the optimal working frequency β(σ, L1), convert the fatigue data sequences of the wind power blades into equidistant sequences, and extract the midpoint value data thereof;

[0075] Construction module: Construct a fatigue life monitoring model based on the different fatigue data sequences of the wind power blades;

[0076] Deep learning module: Use a deep learning network and the midpoint value data to train the constructed fatigue life monitoring model;

[0077] Prediction module: Predict the small defects at the connection of the wind power blade according to the trained fatigue life monitoring model.

[0078] Optionally, the analysis module may be specifically configured to: derive the spatial diffusion equation of the magnetic vector B, determine the analytical general solution of the magnetic vector B based on the spatial diffusion equation; calculate the magnetic field strength Q and the electromagnetic field quantity boundary condition at the P section in the layered medium space based on the analytical general solution of the fixed magnetic vector B; and obtain the optimal operating frequency β(σ, L1) by using the magnetic field strength Q and the electromagnetic field quantity boundary condition.

[0079] Optionally, the database module may be specifically configured to: use the eddy current array sensor with the set optimal operating frequency β(σ, L1) to detect different fatigue data of wind turbine blades; each fatigue data of the wind turbine blade forms a sequence S (0) =(g1, g1, g1,... g i ); and determine whether the sequence is an equidistant sequence. If so, extract the midpoint value data of the sequence; if not, construct a differential equation to convert the sequence into an equidistant sequence and then extract the midpoint value data of the sequence.

[0080] Optionally, the deep learning module may be specifically configured to: perform detrending processing on the different wind turbine blade fatigue data sequences by using the least squares method, label the abnormal sequences in the detrended different wind turbine blade fatigue data sequences by using a deep learning network, extract the sequence features of the different wind turbine blade fatigue data sequences from the detrended different wind turbine blade fatigue data sequences, and extract the mean, variance, kurtosis, and maximum value from the feature vectors as the training input of the deep learning network. And train the constructed fatigue life monitoring model by using the deep learning network with the mean, variance, kurtosis, and maximum value as the training input, and use the difference between the midpoint value data and the result analyzed by the fatigue life monitoring model to backpropagate to the deep learning network to update the deep learning network parameters, and generate a fatigue life monitoring model after a finite number of iterative updates.

[0081] Each unit module of the detection device for small defects at the connection of the wind turbine blade can respectively execute the corresponding steps in the above method embodiments, so the unit modules will not be elaborated here. For details, please refer to the description of the corresponding steps above.

[0082] The present invention further provides a detection device for small defects at the connection of a wind turbine blade, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein, the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the above detection method for small defects at the connection of the wind turbine blade.

[0083] Among them, the memory and the processor are connected in a bus manner. The bus can include any number of interconnected buses and bridges, and the bus connects various circuits of one or more processors and memories together. The bus can also connect various other circuits such as peripheral devices, voltage regulators, and power management circuits, etc., which are well known in the art, so they will not be further described herein. The bus interface provides an interface between the bus and the transceiver. The transceiver can be a component or multiple components, such as multiple receivers and transmitters, and provides a unit for communicating with various other devices on the transmission medium. The data processed by the processor is transmitted on the wireless medium through the antenna. Further, the antenna also receives data and transmits the data to the processor.

[0084] The processor is responsible for managing the bus and general processing, and can also provide various functions, including timing, peripheral interface, voltage regulation, power management, and other control functions. The memory can be used to store the data used by the processor when executing operations.

[0085] The present invention further provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the above method embodiments are implemented.

[0086] It can be found that in the above solution, since the signals of the micro defects are relatively weak, it is difficult to judge the detection results of the sensing signals measured by the eddy current sensor array alone; therefore, the optimal operating frequency β(σ,L1) of the eddy current array sensor is calculated; through such calculation, it can be ensured that the signals of the micro defects can be effectively detected during the process of collecting data, and the validity of the data is guaranteed. According to the operating mechanism of the eddy current array sensor, the accuracy of the automatic monitoring model depends on the transformation coefficient, and the transformation coefficient depends on the background value in the interval [g i ,g i+1 . During the model training process, the background value takes S (1) (g i ) and S (1) (g g+1)'s midpoint value, and subtract the midpoint value from the training result and backpropagate it to the network to update the network parameters. When the data sequence changes smoothly, the monitoring error of this model can be small, ensuring that the model is not affected by the sequence change, greatly reducing the monitoring error, and improving the automatic monitoring accuracy of the model. The signal may have different degrees of offset (polynomial trend). Using the least squares method to cancel the trend term can reduce noise and eliminate the offset, reducing the error. The present invention uses the optimal operating frequency β(σ,L1) to ensure the accuracy of the data. The midpoint value is used for backpropagation to update the deep learning network during the training process to ensure the convergence speed and accuracy of the model. The least squares method is used to solve the problem of the noise of the original data. From three technical means of data, preprocessing, and backpropagation method, the prediction accuracy of the fatigue life monitoring model is improved, enabling it to monitor the small structural defects of the fan blade through the eddy current sensor array.

[0087] In several embodiments provided by the present invention, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of modules or units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces. The indirect couplings or communication connections of devices or units can be in electrical, mechanical, or other forms.

[0088] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place, or they can be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0089] In addition, the functional units in various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.

[0090] When the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the methods according to various embodiments of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs.

[0091] The above are only partial embodiments of the present invention, and thus do not limit the protection scope of the present invention. Any equivalent device or equivalent process transformation made by using the content of the specification and drawings of the present invention, or directly or indirectly applied in other related technical fields, shall similarly be included in the patent protection scope of the present invention.

Claims

1. A detection method for small defects at the connection of wind turbine blades, characterized in that, Including: Calculate the Optimal Operating Frequency of Eddy Current Array Sensors ; Through the optimal operating frequency Obtain different fatigue data sequences of wind turbine blades, convert the fatigue data sequences of the wind turbine blades into equidistant sequences, and extract the midpoint value data therein; Construct a fatigue life monitoring model based on the different wind turbine blade fatigue data sequences; Use a deep learning network and midpoint data to train the constructed fatigue life monitoring model; According to the trained fatigue life monitoring model, predict the small defects at the connection of the wind turbine blade; The step of using a deep learning network and midpoint data to train the constructed fatigue life monitoring model includes: Using the least squares method to eliminate the trend of the different wind turbine blade fatigue data sequences, using a deep learning network to label the abnormal sequences in the different wind turbine blade fatigue data sequences after eliminating the trend, extracting the sequence features of the different wind turbine blade fatigue data sequences from the different wind turbine blade fatigue data sequences after eliminating the trend, and extracting the mean, variance, kurtosis, and maximum value from the feature vector as the training input of the deep learning network, and using the deep learning network with the mean, variance, kurtosis, and maximum value as the training input to train the constructed fatigue life monitoring model, and using the difference between the midpoint data and the analysis result of the fatigue life monitoring model to backpropagate to the deep learning network to update the deep learning network parameters, and generating a fatigue life monitoring model after a finite number of iterative updates.

2. The detection method for minute defects at the connection of wind turbine blades as claimed in claim 1, wherein, Calculating the optimal operating frequency of the eddy current array sensor , including: Derive the spatial diffusion equation of the magnetic vector B, and determine the analytical general solution of the magnetic vector B based on this spatial diffusion equation; calculate the magnetic field strength Q and the electromagnetic field quantity boundary conditions at the P section in the layered medium space based on the analytical general solution of the determined magnetic vector B; obtain the optimal operating frequency by using the magnetic field strength Q and the electromagnetic field quantity boundary conditions .

3. The detection method for minute defects at the connection of a wind power blade as claimed in claim 1, characterized in that, The above-mentioned through the optimal operating frequency Obtain different fatigue data sequences of wind turbine blades, convert the fatigue data sequences of the wind turbine blades into equidistant sequences, and extract the midpoint value data therein; including: Using the set optimal operating frequency of the eddy current array sensor to detect different fatigue data of wind turbine blades; each fatigue data of the wind turbine blade forms a sequence ; and determine whether the sequence is an equidistant sequence. If so, extract the midpoint value data of the sequence; if not, construct a differential equation to convert the sequence into an equidistant sequence and then extract the midpoint value data of the sequence.

4. A detection device for small defects at the connection of a wind turbine blade, characterized in that Including: An analysis module, a database module, a construction module, a deep learning module, and a prediction module; Analysis module: Calculate the optimal operating frequency of the eddy current array sensor ; Database module: through the optimal operating frequency Obtain different fatigue data sequences of wind turbine blades, convert the fatigue data sequences of wind turbine blades into equidistant sequences, and extract the midpoint value data therein; The construction module: Construct a fatigue life monitoring model based on the different wind turbine blade fatigue data sequences; The deep learning module: Use a deep learning network and midpoint data to train the constructed fatigue life monitoring model; The deep learning module is specifically used for: Using the least squares method to eliminate the trend of the different wind turbine blade fatigue data sequences, using a deep learning network to label the abnormal sequences in the different wind turbine blade fatigue data sequences after eliminating the trend, extracting the sequence features of the different wind turbine blade fatigue data sequences from the different wind turbine blade fatigue data sequences after eliminating the trend, and extracting the mean, variance, kurtosis, and maximum value from the feature vector as the training input of the deep learning network, and using the deep learning network with the mean, variance, kurtosis, and maximum value as the training input to train the constructed fatigue life monitoring model, and using the difference between the midpoint data and the analysis result of the fatigue life monitoring model to backpropagate to the deep learning network to update the deep learning network parameters, and generating a fatigue life monitoring model after a finite number of iterative updates; The prediction module: According to the trained fatigue life monitoring model, predict the small defects at the connection of the wind turbine blade.

5. The detection device for small defects at the connection of a wind turbine blade according to claim 4, wherein The analysis module is specifically used for: Derive the spatial diffusion equation of the magnetic vector B, and determine the analytical general solution of the magnetic vector B based on this spatial diffusion equation; calculate the magnetic field intensity Q and the electromagnetic field quantity boundary conditions at the P cross-section in the layered medium space based on the analytical general solution of the determined magnetic vector B; obtain the optimal operating frequency by using the magnetic field intensity Q and the electromagnetic field quantity boundary conditions .

6. The detection device for small defects at the connection of a wind turbine blade according to claim 4, wherein The database module is specifically used for: Using the set optimal operating frequency The eddy current array sensor detects different fatigue data of wind turbine blades; Each fatigue data of a wind turbine blade forms a sequence ; and determine whether the sequence is an equidistant sequence. If so, extract the midpoint value data of the sequence; if not, construct a differential equation to convert the sequence into an equidistant sequence and then extract the midpoint value data of the sequence.

7. A detection device for small defects at the connection of a wind turbine blade, characterized in that, Including: At least one processor; And, A memory communicatively connected to the at least one processor; wherein, the memory stores instructions executable by the at least one processor, and when the instructions are executed by the at least one processor, the at least one processor is enabled to execute the method for detecting minute defects at the connection of a wind turbine blade as recited in any one of claims 1 to 3.

8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, the method for detecting minute defects at the connection of a wind turbine blade as recited in any one of claims 1 to 3 is implemented.

Images