Crack propagation state detection method, device, equipment, medium and product

Abstract

The invention relates to the technical field of material structure performance detection, and discloses a crack propagation state detection method, device and equipment, a medium and a product, and aims to improve the purity of a signal and eliminate the interference of other damage signals by carrying out noise reduction and blind separation treatment on the acoustic emission signal. Furthermore, through feature classification processing, a voiceprint signal feature parameter set which can most accurately reflect the crack propagation state can be found from numerous features, and the effectiveness and pertinence of the features are improved. Furthermore, the change relationship between the voiceprint signal characteristic parameters and the crack propagation rate is established by using a fatigue crack propagation voiceprint signal test method and a Paris's formula, and the relationship between the voiceprint signal characteristic parameters and the crack propagation state can be quantified from the perspective of combining theory and experiment. Finally, in combination with the voiceprint signal feature parameter set and the change relation, the crack propagation state can be accurately recognized in the high-noise environment, and the accuracy and reliability of the detection result are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of material structure performance detection, and in particular to a crack extension state detection method, device, equipment, medium and product. Background Art

[0002] Crack propagation status monitoring of materials is an important industrial issue that is related to the safety and reliability of materials. At present, through the analysis of various fracture scenarios, it is found that the materials that fracture in wind turbines are mainly metal materials and composite materials.

[0003] At present, there are the following problems when using acoustic emission technology to monitor crack growth: the acoustic emission signal generated by crack growth may be mixed with the signals generated by other types of damage inside the material (such as fiber breakage, matrix cracking, etc.), making signal analysis complicated; the acoustic emission signal is usually very weak and may be masked by environmental noise, mechanical noise or electronic instrument interference noise, which makes signal collection and identification difficult; the prediction of crack growth rate is crucial to preventing structural failure, but how to accurately predict the crack growth rate from the acoustic emission signal is a technical challenge. Therefore, the existing methods of using acoustic emission technology to detect crack growth have low accuracy and reliability. Summary of the invention

[0004] In view of this, the present invention provides a method, device, equipment, medium and product for detecting crack propagation status, so as to solve the problem of low accuracy and reliability of the existing method for crack propagation detection using acoustic emission technology.

[0005] In a first aspect, the present invention provides a method for detecting crack propagation status, which is used for a fan component; the method comprises:

[0006] The initial acoustic emission signal set of the fan is obtained; the initial acoustic emission signal set is subjected to noise reduction and blind separation processing to obtain the crack extension signal set of the fan component; the crack extension signal set is subjected to feature classification processing to obtain the acoustic print signal characteristic parameter set reflecting the crack extension state; the relationship between the characteristic parameters of the acoustic print signal and the crack extension rate is established by using the fatigue crack extension acoustic print signal test method and the Paris formula; the crack extension state of the fan component is detected based on the acoustic print signal characteristic parameter set and the change relationship to obtain the crack extension state detection result.

[0007] The crack extension state detection method provided by the present invention can effectively remove environmental noise, mechanical noise, and electronic instrument interference noise by performing noise reduction processing on the initial acoustic emission signal set, thereby improving the purity of the signal. At the same time, through blind separation processing, it is possible to focus on the crack extension signal, eliminate the interference of other damage signals, and improve the accuracy of crack extension state judgment. Further, by performing feature classification processing on the crack extension signal set, the soundprint signal feature parameter set that can most accurately reflect the crack extension state can be found from a large number of features, redundant or irrelevant features are removed, and the effectiveness and pertinence of the features are improved. Further, the relationship between the characteristic parameters of the soundprint signal and the crack extension rate is established by using the fatigue crack extension soundprint signal test method and the Paris formula, which can quantify the relationship between the characteristic parameters of the soundprint signal and the crack extension state from the perspective of combining theory and experiment, thereby improving the scientificity and accuracy of the crack extension state judgment. Finally, combined with the soundprint signal feature parameter set and the change relationship, the crack extension state can be accurately identified in a high noise environment, the accuracy and reliability of the detection results are improved, the false alarm rate is effectively reduced, and a strong guarantee is provided for the safe operation of the fan.

[0008] In an optional implementation, the initial acoustic emission signal set is subjected to noise reduction and blind separation processing to obtain a crack extension signal set of the fan component, including:

[0009] The preset signal denoising combination algorithm is used to perform denoising on the initial acoustic emission signal set to obtain the target acoustic emission signal set; the nonlinear principal component analysis algorithm is used to perform blind separation on the target acoustic emission signal set to obtain the crack propagation signal set of the fan component.

[0010] The crack extension state detection method provided by the present invention can effectively remove environmental noise, mechanical noise, and electronic instrument interference noise by using a preset signal noise reduction combination algorithm, thereby improving the purity of the signal. Furthermore, by using a nonlinear principal component analysis algorithm for blind separation, the acoustic emission signal generated by crack extension can be separated from the mixed signal generated by other types of damage inside the material (such as fiber breakage, matrix cracking, etc.), thereby focusing on the crack extension signal, eliminating the interference of other damage signals, and improving the accuracy of the crack extension state judgment.

[0011] In an optional implementation, a nonlinear principal component analysis algorithm is used to blindly separate the target acoustic emission signal set to obtain a crack extension signal set of the fan component, including:

[0012] The target acoustic emission signal set is subjected to time-varying linear mixing to obtain a mixed acoustic emission signal set; the objective function is constructed using the minimum mean square error approximation criterion; based on the mixed acoustic emission signal set and the objective function, the crack propagation signal set is obtained through gradient descent algorithm processing.

[0013] The crack propagation state detection method provided by the present invention can recombine the original acoustic emission signal by performing time-varying linear mixing on the target acoustic emission signal set. Furthermore, the objective function is constructed using the minimum mean square error approximation criterion, so that the objective function can be as close to the real crack propagation signal as possible, and at the same time, the signal processing target can be converted into a mathematically optimizable problem. Finally, the gradient descent algorithm is used to process the mixed acoustic emission signal set and the objective function, which can quickly and effectively find the minimum value of the objective function, further reducing the influence of other interference signals on the crack propagation signal, making the subsequent crack state analysis based on the separated signal more reliable, and greatly improving the accuracy of the entire crack propagation state monitoring.

[0014] In an optional implementation, the crack extension signal set is subjected to feature classification processing to obtain a voiceprint signal feature parameter set reflecting the crack extension state, including:

[0015] The crack extension signal set is processed by a one-dimensional convolutional neural network algorithm to obtain a fault feature set;

[0016] The fault feature set is processed by the bird flock algorithm to obtain the acoustic signal feature parameter set reflecting the crack propagation state.

[0017] The crack extension state detection method provided by the present invention processes the crack extension signal set through a one-dimensional convolutional neural network algorithm, which can effectively extract fault features, further reduce the interference of human factors, and improve the accuracy and efficiency of feature extraction. Furthermore, by using the intelligent optimization characteristics of the bird flock algorithm, the features are further screened and optimized, and the soundprint signal feature parameter set that can most accurately reflect the crack extension state is screened from a large number of features, thereby removing redundant or irrelevant features and improving the effectiveness and pertinence of the features.

[0018] In an optional implementation, the fatigue crack extension soundprint signal test method and the Paris formula are used to establish a relationship between the characteristic parameters of the soundprint signal and the crack extension rate, including:

[0019] The fatigue crack extension soundprint signal test method is used to establish the first change relationship curve between the crack extension rate and the stress intensity factor range, and the second change relationship curve between the soundprint signal characteristic parameters and the cycle number; based on the first change relationship curve and the second change relationship curve, the Paris formula is used to determine the third correlation relationship between the soundprint signal characteristic parameters and the stress intensity factor range, and the fourth correlation relationship between the crack extension rate and the stress intensity factor range; according to the third correlation relationship and the fourth correlation relationship, the change relationship between the soundprint signal characteristic parameters and the crack extension rate is determined.

[0020] The crack extension state detection method provided by the present invention can directly obtain the actual crack extension data and the corresponding change of the characteristic parameters of the soundprint signal with the number of cycles by conducting fatigue crack extension soundprint signal tests. Furthermore, the Paris formula can be used to further study the intrinsic connection between the characteristic parameters of the soundprint signal and the range of the stress intensity factor, and to more accurately define the quantitative relationship between the crack extension rate and the range of the stress intensity factor. Finally, the changing relationship between the characteristic parameters of the soundprint signal and the crack extension rate can be finally determined based on the determined third and fourth correlations. Therefore, by implementing the present invention, from the perspective of combining theory and experiment, the connection between the characteristic parameters of the soundprint signal and the crack extension state is quantified, thereby improving the scientificity and accuracy of the crack extension state judgment.

[0021] In an optional implementation, based on the characteristic parameter set and the change relationship of the voiceprint signal, the crack extension state of the fan component is detected to obtain the crack extension state detection result, including:

[0022] Determine changes in characteristic parameters of the voiceprint signal and multiple first crack extension states based on the voiceprint signal characteristic parameter set; determine multiple second crack extension states based on the changes and change relationships; determine crack extension state detection results based on the multiple first crack extension states and the multiple second crack extension states.

[0023] The crack extension state detection method provided by the present invention can obtain the change trend of the characteristic parameters of the voiceprint signal by analyzing the characteristic parameter set of the voiceprint signal, and can also perform preliminary identification and detection of the crack extension state and obtain the corresponding multiple first crack extension states. Furthermore, combined with the change of the characteristic parameters of the voiceprint signal and the established relationship between the characteristic parameters of the voiceprint signal and the crack extension rate, the change of the voiceprint signal can be linked to the actual crack extension state, which improves the scientificity and accuracy of the judgment of the crack extension state. Compared with the traditional single judgment method, it can more comprehensively and accurately identify the characteristics of the crack at different stages. Finally, by comprehensively considering the multiple second crack extension states and the multiple first crack extension states determined in the previous steps, a comprehensive and accurate crack extension state detection result can be obtained, which improves the accuracy and reliability of the detection result.

[0024] In a second aspect, the present invention provides a crack extension state detection device for a fan component; the device comprises:

[0025] An acquisition module is used to acquire the initial acoustic emission signal set of the fan; a first processing module is used to perform noise reduction and blind separation processing on the initial acoustic emission signal set to obtain a crack extension signal set of the fan component; a second processing module is used to perform feature classification processing on the crack extension signal set to obtain a soundprint signal characteristic parameter set reflecting the crack extension state; an establishment module is used to establish a changing relationship between the soundprint signal characteristic parameters and the crack extension rate by using the fatigue crack extension soundprint signal test method and the Paris formula; a detection module is used to detect the crack extension state of the fan component based on the soundprint signal characteristic parameter set and the changing relationship to obtain a crack extension state detection result.

[0026] In a third aspect, the present invention provides a computer device, comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the crack extension state detection method of the first aspect or any corresponding embodiment thereof by executing the computer instructions.

[0027] In a fourth aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the crack extension state detection method of the first aspect or any corresponding embodiment thereof.

[0028] In a fifth aspect, the present invention provides a computer program product, comprising computer instructions for causing a computer to execute the crack extension state detection method of the first aspect or any corresponding embodiment thereof. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0030] Figure 1 is a schematic flow chart of a crack extension state detection method according to an embodiment of the present invention;

[0031] Figure 2 is a schematic flow chart of another crack extension state detection method according to an embodiment of the present invention;

[0032] Figure 3 is a schematic flow chart of another crack extension state detection method according to an embodiment of the present invention;

[0033] Figure 4 is a structural block diagram of a crack extension state detection device according to an embodiment of the present invention;

[0034] Figure 5 It is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0035] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0036] Crack propagation state monitoring of materials is an important industrial issue that is related to the safety and reliability of materials. Existing common methods include:

[0037] Ultrasonic testing (UT): By emitting ultrasonic waves and receiving the reflected signals, cracks inside the material can be detected. Cracks will change the propagation path and speed of the sound waves, thus being detected.

[0038] Radiographic testing (RT): uses X-rays or gamma rays to penetrate the material. Cracks block the rays, causing them to form shadows on film or digital detectors, revealing the presence of cracks.

[0039] Magnetic Particle Testing (MT): Utilizing the principle that cracks in a magnetic field will attract magnetic particles, the MT method detects cracks by observing the distribution of magnetic particles.

[0040] Penetrant Testing (PT): A penetrant containing a fluorescent or dye is applied to the surface of the material. The cracks will absorb the penetrant and then treated with a developer to make the penetrant in the cracks visible.

[0041] Eddy current testing (ET): By generating eddy currents on the surface of the material, cracks will change the distribution of eddy currents, and cracks can be found by detecting changes in eddy currents.

[0042] Acoustic Emission (AE): When a crack propagates, sound waves are generated. By capturing these sound waves with sensors installed on the surface of the material, the crack propagation can be monitored in real time.

[0043] Optical microscope: Under laboratory conditions, an optical microscope can be used to observe cracks on the surface of materials.

[0044] Electron microscope: has higher resolution than optical microscope and can observe smaller cracks.

[0045] Digital Image Correlation (DIC): By comparing images before and after crack propagation, the displacement and propagation of the crack can be calculated.

[0046] Stress wave analysis: By analyzing the propagation characteristics of stress waves inside the material, the existence and expansion of cracks can be inferred.

[0047] Fiber optic sensor: The optical fiber is embedded in the material, and the expansion of the crack will cause the breakage of the optical fiber or the change of the optical signal, thereby monitoring the crack.

[0048] Smart materials and structures: Use smart materials such as piezoelectrics or shape memory alloys, which can generate electrical signals or change shape as cracks propagate and can be monitored.

[0049] The method for detecting the crack propagation state in a material depends on factors such as the type of material, the size of the crack, the frequency of monitoring, whether real-time monitoring is required, and cost. In practical applications, due to different usage scenarios, it may be necessary to combine and optimize multiple crack detection methods to improve the accuracy and reliability of crack monitoring.

[0050] As for wind turbines, their key components include blades, towers, generators, gearboxes, bearings, etc. These components are subjected to huge mechanical stress and environmental impacts during operation. Offshore wind power is subjected to severe tests due to the salt spray environment. Some key structural parts are very likely to break, which in turn causes significant economic losses. According to statistics, the fractures caused by cracks in wind turbines generally occur in blades, towers, generators, gearboxes, various shafts and bearings, bolts and connectors.

[0051] In the blade, the tip of the blade: due to the large bending stress and centrifugal force at the tip of the blade, it is easy to break here; the root of the blade: the connection between the blade and the hub is subject to large bending moment and torque, which is a common location of fracture; the surface of the blade: due to fatigue, corrosion or impact damage, cracks may appear on the surface of the blade and extend to fracture.

[0052] In the tower, connecting flange: the flange where the tower is connected to the foundation, blades or generator may break due to the large load; bottom of the tower: the connection between the bottom of the tower and the foundation may break due to the weight and wind load of the entire wind turbine; middle section of the tower: the middle section of the tower may break due to corrosion, fatigue or manufacturing defects.

[0053] In the generator: Shaft: The shaft of the generator may break due to fatigue or overload during long-term operation; Bearing: The generator bearing may be damaged due to poor lubrication, overheating or fatigue, resulting in shaft breakage.

[0054] In the gearbox, gears: The gears in the gearbox may break due to fatigue, overload or manufacturing defects; bearings: The bearings in the gearbox may be damaged due to poor lubrication, overheating or fatigue, causing the gears to break.

[0055] Among various types of bearings, balls / rollers: the balls or rollers inside the bearings may break due to fatigue or impact; bearing rings: the bearing rings may break due to overload or manufacturing defects.

[0056] Among bolts and connectors, bolts: bolts connecting various parts of the fan may break due to fatigue, overload or corrosion; connectors: connectors may break due to fatigue, overload or corrosion.

[0057] Therefore, by analyzing various fracture scenarios, it is not difficult to conclude that the materials that fracture in the wind turbine are mainly metal materials and composite materials.

[0058] Furthermore, the acoustic emission signals of metal materials during the fracture process usually show three stages: Crack nucleation: Under the action of stress, micro defects inside the material begin to expand and form tiny cracks, at which time acoustic emission signals are generated. Crack steady-state crack expansion: With the continuous action of the load, the crack gradually expands, accompanied by the continuous generation of acoustic emission signals. Crack rapid fracture: When the crack expands to a critical size, the remaining material cannot withstand the applied stress, resulting in fracture, at which time a large number of acoustic emission signals are generated.

[0059] The fracture process of composite materials may involve a variety of different failure mechanisms, including but not limited to: Fiber fracture: When the reinforcing fibers in the composite material break, energy is released, generating an AE signal. Matrix fracture: Fracture of the matrix material in the composite material will also generate an AE signal. Interface debonding: Interface debonding between the fiber and the matrix will also generate an AE signal. Crack propagation: During the crack propagation process in the composite material, the formation of new surfaces will also generate an AE signal.

[0060] Therefore, only by establishing the corresponding relationship between the acoustic emission signal parameters and the crack propagation state in metal materials and composite materials can the real-time monitoring and analysis of the crack propagation state of key structural parts in the wind turbine be realized using acoustic emission technology, and the initiation and propagation of cracks can be discovered in time, thereby preventing structural failure.

[0061] According to an embodiment of the present invention, an embodiment of a crack propagation state detection method is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0062] In this embodiment, a crack extension state detection method is provided for a fan component; Figure 1 is a flow chart of a crack extension state detection method according to an embodiment of the present invention. Figure 1 As shown, the process includes the following steps:

[0063] Step S101, obtaining an initial acoustic emission signal set of a wind turbine.

[0064] Among them, the initial acoustic emission signal set refers to the set of original sound wave signals collected by acoustic emission sensors installed on the surface of key components of the wind turbine (such as blades, towers, generators, gearboxes, bearings, etc.) during the operation of the wind turbine. It may include various sound wave information generated by the wind turbine components under various working conditions such as mechanical stress and environmental influences (such as salt spray environment of offshore wind turbines).

[0065] Step S102: performing noise reduction and blind separation processing on the initial acoustic emission signal set to obtain a crack extension signal set of the fan component.

[0066] Among them, the noise reduction process can adopt methods such as wavelet transform method, adaptive filtering method (such as least mean square algorithm and recursive least square algorithm, etc.).

[0067] Furthermore, the blind separation process may adopt a nonlinear principal component analysis (NLPCA) method, an independent component analysis method (ICA), a clustering-based blind separation method, and the like.

[0068] Furthermore, the crack propagation signal set represents a set of acoustic emission signals directly related to the crack propagation process of the fan components, which is extracted from the initial acoustic emission signal set after noise reduction and blind separation processing, and can reflect the acoustic print information generated at various stages of the crack from initiation, stable propagation to possible unstable propagation.

[0069] Specifically, by performing noise reduction processing on the initial acoustic emission signal set, environmental noise, mechanical noise, and electronic instrument interference noise can be effectively removed, thereby improving the purity of the signal. At the same time, through blind separation processing, the interference of other damage signals can be eliminated, and then the crack extension signal can be focused on and the corresponding crack extension signal set can be obtained.

[0070] Step S103, performing feature classification processing on the crack extension signal set to obtain a voiceprint signal feature parameter set reflecting the crack extension state.

[0071] Among them, feature classification processing can adopt deep learning methods, machine learning methods, statistical-based methods, etc.; the voiceprint signal feature parameter set can include time domain feature parameters (such as ring count, rise time, duration, etc.) and frequency domain feature parameters (such as center frequency, bandwidth, etc.), etc.

[0072] Specifically, by extracting and classifying the features of the crack extension signal set, it is possible to find the voiceprint signal feature parameter set that most accurately reflects the crack extension state from numerous features, thereby removing redundant or irrelevant features and improving the effectiveness and pertinence of the features.

[0073] In an optional embodiment, feature classification processing is performed on the crack extension signal set through a deep neural network.

[0074] First, the crack extension signal set is converted into a form suitable for neural network input, such as digitizing and normalizing the signal.

[0075] Secondly, a neural network structure with multiple hidden layers is constructed, and the number of network layers and the number of neurons in each layer are determined according to the dimension of the input data and the complexity of the problem.

[0076] Furthermore, the network is trained using labeled training data (data with known crack propagation states), and the network weights are adjusted through the back-propagation algorithm to minimize the error between the predicted results and the true labels.

[0077] Finally, after multiple iterations of training, the network can learn the mapping from the input signal to the crack propagation state, and then classify the new crack propagation signal and obtain the corresponding voiceprint signal feature parameter set.

[0078] Step S104, using the fatigue crack extension voiceprint signal test method and the Paris formula, establish the relationship between the characteristic parameters of the voiceprint signal and the crack extension rate.

[0079] Among them, the fatigue crack extension acoustic print signal test method refers to a test method used to study the change law of acoustic emission signals (acoustic print signals) during the crack extension process of materials under fatigue load. During the test, by applying periodic fatigue loads to the test samples (usually the materials used in the key components of the wind turbine, such as metals or composite materials), the alternating stresses that the components are subjected to under actual working conditions are simulated. At the same time, acoustic emission sensors can be used to monitor the acoustic emission signals generated during the process of crack initiation, extension and fracture inside the material.

[0080] Furthermore, the Paris Law represents a formula that describes the relationship between fatigue crack growth rate and the range of stress intensity factors.

[0081] Specifically, by combining the fatigue crack extension soundprint signal test method and the Paris formula, it is possible to quantify the relationship between the characteristic parameters of the soundprint signal and the crack extension state from the perspective of combining theory and experiment, thereby improving the scientificity and accuracy of the crack extension state judgment.

[0082] Step S105: Based on the characteristic parameter set and the change relationship of the voiceprint signal, the crack propagation state of the fan component is detected to obtain a crack propagation state detection result.

[0083] Specifically, by analyzing the current state and change trend of the characteristic parameters of the voiceprint signal and then combining the change relationship, the crack propagation state can be accurately identified in a high noise environment and finally the crack propagation state detection result can be obtained, thereby improving the accuracy and reliability of the detection result.

[0084] The crack extension state detection method provided in this embodiment can effectively remove environmental noise, mechanical noise, and electronic instrument interference noise by performing noise reduction processing on the initial acoustic emission signal set, thereby improving the purity of the signal. At the same time, through blind separation processing, it is possible to focus on the crack extension signal, eliminate the interference of other damage signals, and improve the accuracy of crack extension state judgment. Further, by performing feature classification processing on the crack extension signal set, the soundprint signal feature parameter set that can most accurately reflect the crack extension state can be found from many features, redundant or irrelevant features are removed, and the effectiveness and pertinence of the features are improved. Further, the relationship between the characteristic parameters of the soundprint signal and the crack extension rate is established by using the fatigue crack extension soundprint signal test method and the Paris formula, which can quantify the relationship between the characteristic parameters of the soundprint signal and the crack extension state from the perspective of combining theory and experiment, thereby improving the scientificity and accuracy of the crack extension state judgment. Finally, combined with the soundprint signal feature parameter set and the change relationship, the crack extension state can be accurately identified in a high noise environment, the accuracy and reliability of the detection results are improved, the false alarm rate is effectively reduced, and a strong guarantee is provided for the safe operation of the fan.

[0085] In this embodiment, a crack extension state detection method is provided for a fan component; Figure 2 is a flow chart of a crack extension state detection method according to an embodiment of the present invention. Figure 2 As shown, the process includes the following steps:

[0086] Step S201, obtain the initial acoustic emission signal set of the fan. Figure 1 Step S101 of the illustrated embodiment will not be described in detail here.

[0087] Step S202: performing noise reduction and blind separation processing on the initial acoustic emission signal set to obtain a crack extension signal set of the fan component.

[0088] Specifically, the above step S202 includes:

[0089] Step S2021: Use a preset signal noise reduction combination algorithm to perform noise reduction processing on the initial acoustic emission signal set to obtain a target acoustic emission signal set.

[0090] Among them, the preset signal noise reduction combination algorithm represents a method that integrates multiple noise reduction techniques, which is used to remove noise from the initial acoustic emission signal set to improve signal quality and obtain a purer target acoustic emission signal set. In this embodiment, the preset signal noise reduction combination algorithm is a Bayesian optimization-wavelet threshold-variational mode decomposition (BO-WT-VMD) algorithm.

[0091] First, wavelet transform is used to determine the number of wavelet layers and remove noise. Wavelet transform is a time-frequency analysis method that can decompose the signal into wavelet coefficients of different frequencies.

[0092] Specifically, the characteristics of the original signal are analyzed and the appropriate number of wavelet layers is determined, and then the original signal is denoised according to the wavelet threshold method.

[0093] Among them, wavelet threshold denoising means setting the coefficients with smaller absolute values ​​in the wavelet coefficients (usually considered to be caused by noise) to zero or shrinking them, while retaining the coefficients with larger absolute values ​​(considered to be the main components of the signal), thereby achieving the purpose of removing noise.

[0094] Secondly, Bayesian optimization is used to determine the parameters K and α of VMD, where VMD (Variational Mode Decomposition) represents an adaptive signal decomposition method; K represents the number of decomposed modes; and α represents a parameter related to bandwidth constraints.

[0095] Specifically, the hyperparameter combination and the prior data set are randomly initialized. Further, the sampling points can be increased through the Gaussian model (GP) to obtain the posterior data set.

[0096] Furthermore, if the posterior data set meets the termination condition (for example, a certain accuracy requirement or number of iterations is reached), the optimized parameter combination is determined to obtain the parameters of the wavelet transform; if the termination condition is not met, the Gaussian model and the previous steps are repeated until the posterior data set meets the termination condition, and finally the optimized parameter combination is determined to obtain the parameters K and α of the VMD.

[0097] Finally, the obtained parameters K and α are substituted into the VMD algorithm to obtain the decomposed mode and select the optimal mode.

[0098] Specifically, the obtained parameters K and α are substituted into the VMD algorithm and the signal is decomposed to obtain K modes. Furthermore, the optimal mode can be selected by the Pearson coefficient, and then the remaining modes can be eliminated to finally obtain the denoised reconstructed data, i.e., the target acoustic emission signal set.

[0099] Among them, the Pearson coefficient represents an indicator to measure the linear correlation between two variables.

[0100] Step S2022: Use a nonlinear principal component analysis algorithm to blindly separate the target acoustic emission signal set to obtain a crack propagation signal set of the fan component.

[0101] Among them, the nonlinear principal component analysis (NLPCA) algorithm represents a principal component analysis method for processing nonlinear data, which can map high-dimensional nonlinear data to a low-dimensional space while retaining the main information and structure of the data as much as possible.

[0102] In some optional implementations, the above step S2022 includes:

[0103] Step a1, performing time-varying linear mixing on the target acoustic emission signal set to obtain a mixed acoustic emission signal set.

[0104] Step a2: construct the objective function using the minimum mean square error approximation criterion.

[0105] Step a3, based on the mixed acoustic emission signal set and the objective function, a crack extension signal set is obtained by processing with a gradient descent algorithm.

[0106] Specifically, assuming that the target acoustic emission signal set is s(t), after the time-varying system The mixed acoustic emission signal set x(t) is obtained by linear mixing, that is,

[0107] Among them, time-varying linear mixing represents the coefficients in the mixing process (given by It can simulate the time-varying interference and mixing that the actual signal may be subjected to during the transmission or acquisition process.

[0108] Furthermore, the minimum mean square error approximation criterion can be used as the independence criterion and the corresponding objective function J(W) can be constructed, as shown in the following equation (1):

[0109]

[0110] Where: represents the time-varying system to be solved Generalized inverse matrix; represents a nonlinear odd function; E represents the mathematical expectation.

[0111] Furthermore, the purpose of the above objective function is to find the optimal Make the separated signal as close as possible to the original independent signal source.

[0112] Furthermore, the gradient descent algorithm is used to solve the problem, and its iterative formula is shown in the following equation (2):

[0113]

[0114] Where: μ represents the learning rate, which is used to control the step size of each iteration; LT represents a lower triangular matrix.

[0115] Furthermore, in the iterative process, according to the mixed acoustic emission signal set x(t) and the current Calculate the gradient of the objective function and then update it according to the above relationship (2) Iterate continuously until convergence (for example, the value of the objective function no longer changes significantly or reaches the preset number of iterations), and then finally obtain Used to separate mixed signals.

[0116] Finally, the crack extension signal set y(t) of the fan component is obtained, that is,

[0117] Furthermore, when y(t) is a sub-Gaussian function, When it is a super-Gaussian signal y(t),

[0118] Step S203, performing feature classification processing on the crack extension signal set to obtain a voiceprint signal feature parameter set reflecting the crack extension state.

[0119] Specifically, the above step S203 includes:

[0120] Step S2031, processing the crack extension signal set through a one-dimensional convolutional neural network algorithm to obtain a fault feature set.

[0121] The fault feature set may include a plurality of key information that can reflect the crack propagation state.

[0122] Furthermore, the one-dimensional convolutional neural network (1DCNN) algorithm represents a deep learning model for processing one-dimensional sequence data. It can automatically learn feature representation from input data through structures such as convolutional layers, pooling layers, and fully connected layers, and can be applied to tasks such as feature extraction and classification of time series data.

[0123] Specifically, the crack extension signal set can be sorted and segmented to form a fixed-segment data sample set, which can include laboratory runner material fatigue test soundprint signals, model runner test soundprint signals under different water environments, etc., to ensure that the data is representative and diverse.

[0124] Furthermore, the obtained data sample set is input into the convolution layer of 1DCNN, and then the convolution kernel in the convolution layer can slide on the input data and extract local features through convolution operation.

[0125] Among them, the weights of the convolution kernel can be automatically learned through training, and can adaptively capture various patterns and features in the crack extension signal, such as different frequency components and changing trends in time series.

[0126] Furthermore, after being processed by the convolutional layer, the extracted features can be passed to subsequent layers (which may include a pooling layer for feature dimensionality reduction and a fully connected layer for feature integration), and finally output by the fully connected layer and sent to the output layer, which then outputs the fault feature set extracted after being processed by the one-dimensional convolutional neural network algorithm.

[0127] Step S2032: Process the fault feature set using a bird flock algorithm to obtain a voiceprint signal feature parameter set reflecting the crack propagation state.

[0128] Among them, the bird flock algorithm represents an intelligent optimization algorithm based on the behavior of bird groups. It can simulate the cooperation and competition mechanism of bird flocks in the process of foraging, flying, etc., and find the optimal solution by continuously updating the position and speed of each individual in the flock.

[0129] Specifically, the bird flock algorithm can be used to optimize the values ​​of the penalty parameter c and the parameter g of the radial basis kernel function in the support vector machine (SVM) model.

[0130] In the bird flock algorithm, each bird is regarded as a possible solution (i.e., a set of values ​​of c and g). The flock of birds flies and forages in the search space, and the optimal parameter combination is found by continuously updating the position of the bird (i.e., updating the parameter value).

[0131] Furthermore, the individual (bird) keyi1 in the bird swarm algorithm adjusts its speed and position according to its own historical optimal position, the historical optimal position of the group, and some random factors, thereby exploring and utilizing the entire search space to find the parameter value that optimizes the performance of the SVM model.

[0132] Furthermore, the fault feature set is input into the SVM classifier after bird flock optimization.

[0133] Furthermore, the optimized SVM classifier can classify the fault feature set according to the optimized parameters c and g and finally obtain the acoustic signal feature parameter set reflecting the crack propagation state.

[0134] Step S204, using the fatigue crack extension voiceprint signal test method and Paris formula, establish the relationship between the characteristic parameters of the voiceprint signal and the crack extension rate. Figure 1 Step S104 of the illustrated embodiment will not be described in detail here.

[0135] Step S205: Based on the characteristic parameter set and the change relationship of the voiceprint signal, the crack extension state of the fan component is detected to obtain the crack extension state detection result. Figure 1 Step S105 of the illustrated embodiment will not be described in detail here.

[0136] The crack extension state detection method provided in this embodiment can effectively remove environmental noise, mechanical noise, and electronic instrument interference noise by using a preset signal noise reduction combination algorithm, thereby improving the purity of the signal. Further, by performing time-varying linear mixing on the target acoustic emission signal set, the original acoustic emission signal can be recombined. Further, the objective function is constructed using the minimum mean square error approximation criterion, so that the objective function can be as close to the real crack extension signal as possible, and the signal processing target can be converted into a mathematically optimizable problem. Finally, the gradient descent algorithm is used to process the mixed acoustic emission signal set and the objective function, which can quickly and effectively find the minimum value of the objective function, further reducing the influence of other interference signals on the crack extension signal. Further, the crack extension signal set is processed by a one-dimensional convolutional neural network algorithm, which can effectively extract fault features, further reducing the interference of human factors and improving the accuracy and efficiency of feature extraction. Further, by using the intelligent optimization characteristics of the bird flock algorithm, the features are further screened and optimized, and the voiceprint signal feature parameter set that can most accurately reflect the crack extension state is screened from many features, removing redundant or irrelevant features, and improving the effectiveness and pertinence of the features.

[0137] In this embodiment, a crack extension state detection method is provided for a fan component; Figure 3 is a flow chart of a crack extension state detection method according to an embodiment of the present invention. Figure 3 As shown, the process includes the following steps:

[0138] Step S301, obtain the initial acoustic emission signal set of the fan. Figure 1 Step S101 of the illustrated embodiment will not be described in detail here.

[0139] Step S302: Perform noise reduction and blind separation on the initial acoustic emission signal set to obtain a crack extension signal set of the fan component. Figure 2 Step S202 of the illustrated embodiment will not be described in detail here.

[0140] Step S303, perform feature classification processing on the crack extension signal set to obtain a voiceprint signal feature parameter set reflecting the crack extension state. Figure 2 Step S203 of the illustrated embodiment will not be described in detail here.

[0141] Step S304, using the fatigue crack extension voiceprint signal test method and the Paris formula, establish the relationship between the characteristic parameters of the voiceprint signal and the crack extension rate.

[0142] Specifically, the above step S304 includes:

[0143] Step S3041, using the fatigue crack growth soundprint signal test method to establish a first change relationship curve between the crack growth rate and the stress intensity factor range, and a second change relationship curve between the soundprint signal characteristic parameters and the cycle number.

[0144] Specifically, conventional metal materials for fans can be selected and metal material fatigue crack extension acoustic print signal tests can be carried out.

[0145] Among them, during the test, multiple key parameters can be recorded and measured, such as fatigue crack length a, stress intensity factor range ΔK, number of cycles N, fatigue crack growth rate da / dN and acoustic emission characteristic parameters (such as ringing count, amplitude, energy count, etc.).

[0146] Furthermore, the experimental data can be used to draw a relationship curve between the fatigue crack growth rate da / dN and the stress intensity factor range ΔK, that is, a first relationship curve, which can reflect the law of the crack growth rate changing with the stress intensity factor range.

[0147] At the same time, based on the change data of the acoustic emission characteristic parameters (such as ring count, amplitude, energy count, etc.) recorded in the test with the cycle number N, a second change relationship curve between the characteristic parameters of the soundprint signal and the cycle number N can be drawn, which can show the evolution of the characteristic parameters of the soundprint signal with the number of cycles during the fatigue process. For example, some characteristic parameters may change little in the early stage of crack initiation, and their values ​​will gradually increase or show a specific change pattern as the crack expands and the number of cycles increases.

[0148] Step S3042: Based on the first change relationship curve and the second change relationship curve, the third correlation relationship between the characteristic parameters of the voiceprint signal and the range of the stress intensity factor and the fourth correlation relationship between the crack growth rate and the range of the stress intensity factor are determined using the Paris formula.

[0149] Specifically, according to the first variation relationship curve, the fourth correlation relationship between the crack growth rate and the stress intensity factor range can be directly determined by citing the Paris formula, as shown in the following relationship (3):

[0150]

[0151] Where: C and n represent material constants, which can be obtained by fitting the test data.

[0152] Further, according to the second variation relationship curve, the cumulative value H can be taken for each characteristic parameter of the voiceprint signal (such as ring count, amplitude, energy count, etc.). At the same time, the introduction of the Paris formula can determine the third correlation relationship between the characteristic parameters of the voiceprint signal and the range of the stress intensity factor, as shown in the following relationship (4):

[0153]

[0154] Where: C1 and n1 represent material constants, which can be obtained by fitting the test data.

[0155] Step S3043: determine the changing relationship between the voiceprint signal characteristic parameter and the crack growth rate according to the third association relationship and the fourth association relationship.

[0156] Specifically, the above equations (3) and (4) can be logarithmized and further deduced to obtain the change rate of the cumulative value of the crack growth rate and the acoustic print characteristic parameter with the number of cycles: The quantitative relationship between them is shown in the following equation (5):

[0157]

[0158] Furthermore, the activity parameters of the voiceprint signal characteristic parameters along with the cycle times can be used to Reflects the crack growth rate Then, the relationship between the characteristic parameters of the voiceprint signal and the crack growth rate can be established, namely the above relationship (5).

[0159] Step S305: Based on the characteristic parameter set and the change relationship of the voiceprint signal, the crack propagation state of the fan component is detected to obtain a crack propagation state detection result.

[0160] Specifically, the above step S305 includes:

[0161] Step S3051, determining the change of the voiceprint signal characteristic parameters and a plurality of first crack extension states according to the voiceprint signal characteristic parameter set.

[0162] The multiple first crack extension states may include three main states: crack initiation, crack extension, and crack instability.

[0163] Specifically, by performing time domain analysis and frequency domain analysis on the characteristic parameter set of the voiceprint signal, the change of the characteristic parameters of the voiceprint signal can be obtained.

[0164] First, the characteristic parameters of the voiceprint signal can be arranged in chronological order to form time series data.

[0165] Secondly, the difference between each characteristic parameter at adjacent time points can be calculated, so as to observe its instantaneous change. For example, for ring count, the change range of the ring count can be determined by calculating the difference between two adjacent monitorings.

[0166] Furthermore, a curve of each characteristic parameter changing over time can be drawn to intuitively display its changing trend, including rising, falling, stable or fluctuating conditions. Furthermore, the overall changing trend of the characteristic parameter during the entire monitoring period can be clearly seen through the curve.

[0167] Furthermore, by calculating the statistical indicators of each feature parameter, such as mean, variance, standard deviation, etc., its central tendency and dispersion degree in the entire data set can be determined. Furthermore, the rate of change of the feature parameter can be calculated.

[0168] Furthermore, methods such as correlation coefficient (such as Pearson correlation coefficient) can be used to analyze the correlation between different feature parameters and determine the association relationship between the feature parameters.

[0169] Finally, by combining the time domain analysis results and the frequency domain analysis results, the changes in the characteristic parameters of the voiceprint signal can be ultimately determined.

[0170] Further, multiple first crack extension states can be judged and determined in the following manner:

[0171] (1) If the characteristic parameters of most voiceprint signals change slowly, the rate of change is small and close to zero, and the values ​​of the characteristic parameters fluctuate within a small range, it can be preliminarily determined that the crack is in the initiation stage;

[0172] (2) If some characteristic parameters of the voiceprint signal show a relatively stable linear or approximately linear change trend, the rate of change remains near a relatively stable value, and the coordinated changes between different characteristic parameters are relatively regular, it can be preliminarily judged that the crack is in the expansion stage.

[0173] (3) If the changes in certain characteristic parameters of the voiceprint signal become drastic, such as a sudden and significant increase in the rate of change, a sharp increase in the value of the characteristic parameter, or a significant increase in the fluctuation amplitude, it can be preliminarily determined that the crack is in an unstable state.

[0174] Step S3052: determining a plurality of second crack extension states based on the change conditions and change relationships.

[0175] The multiple second crack extension states may include three main states: crack initiation, crack extension, and crack instability.

[0176] Specifically, according to the obtained changes in the characteristic parameters of the voiceprint signal and in combination with the changing relationship between the characteristic parameters of the voiceprint signal and the crack growth rate, the changes in the crack growth rate corresponding to the changes in the characteristic parameters of each voiceprint signal can be further determined.

[0177] Furthermore, the state of the crack, that is, a plurality of second crack extension states, can be determined according to the change in the crack extension rate.

[0178] Step S3053, determining a crack extension state detection result according to the plurality of first crack extension states and the plurality of second crack extension states.

[0179] Specifically, by comprehensively considering the multiple first crack extension states and the multiple second crack extension states, the final crack extension state detection result can be further determined.

[0180] For example, when the first crack extension state is completely consistent with the second crack extension state and the characteristics of this state are very significant, it can be directly determined that the final crack extension state detection result is this state. Through comprehensive analysis, a comprehensive and accurate crack extension state detection result can be obtained, thereby improving the accuracy and reliability of the detection results.

[0181] The crack extension state detection method provided in this embodiment can directly obtain the actual crack extension data and the corresponding change of the characteristic parameters of the soundprint signal with the number of cycles by conducting fatigue crack extension soundprint signal tests. Furthermore, the Paris formula can be used to further study the intrinsic relationship between the characteristic parameters of the soundprint signal and the range of the stress intensity factor, and to more accurately define the quantitative relationship between the crack extension rate and the range of the stress intensity factor. Finally, according to the determined third and fourth correlation relationships, the changing relationship between the characteristic parameters of the soundprint signal and the crack extension rate can be finally determined, and the relationship between the characteristic parameters of the soundprint signal and the crack extension state is quantified from the perspective of combining theory and experiment. Furthermore, by analyzing the characteristic parameter set of the soundprint signal, the changing trend of the characteristic parameters of the soundprint signal can be obtained, and at the same time, the crack extension state can be preliminarily identified and detected and the corresponding multiple first crack extension states can be obtained. Furthermore, by combining the changes in the characteristic parameters of the voiceprint signal and the established relationship between the characteristic parameters of the voiceprint signal and the crack growth rate, the changes in the voiceprint signal can be linked to the actual crack growth state, which improves the scientificity and accuracy of the crack growth state judgment. Compared with the traditional single judgment method, it can more comprehensively and accurately identify the characteristics of cracks at different stages. Finally, by comprehensively considering multiple second crack growth states and multiple first crack growth states determined in the previous steps, a comprehensive and accurate crack growth state detection result can be obtained, which improves the accuracy and reliability of the detection result.

[0182] In one example, a non-destructive testing optimization combination method for monitoring the crack propagation state of key components of a wind turbine is provided. Mainly from the perspective of the time domain and frequency domain of the acoustic emission signal, a new method for noise reduction, separation, feature extraction and pattern recognition of the acoustic emission signal is proposed. From the perspective of the characteristic parameters of the acoustic emission signal, the correlation between the change in the characteristic parameter and the change in the crack stress intensity factor is established. Combining the above two methods, the change in the acoustic emission signal is combined and analyzed to obtain the corresponding relationship between the acoustic emission signal and the crack propagation state, thereby improving the accuracy and reliability of the acoustic emission signal in monitoring the crack propagation state. Specifically including:

[0183] Step 1: Signal noise reduction.

[0184] Specifically, the working environment of the fan is complex, various types of acoustic emission signals are strong, and the background noise is strong. When using acoustic emission signals to evaluate the crack state, it will be interfered by other types of acoustic emission signals and noise. Using the Bayesian optimization-wavelet threshold-variational mode decomposition (BO-WT-VMD) combined algorithm for signal denoising can effectively improve the signal-to-noise ratio and solve the interference of noise.

[0185] Firstly, the wavelet transform is used to determine the number of wavelet layers and denoise the original signal.

[0186] Secondly, Bayesian optimization is used to determine the parameters K and α of VMD, that is: first, randomly initialize the hyperparameter combination and the prior data set, increase the sampling points through the Gaussian model GP, and obtain the posterior data set. If the posterior data set meets the termination condition, the optimized parameter combination is determined to obtain the parameters of the wavelet transform; if the posterior data set does not meet the termination condition, the Gaussian model and the previous steps are repeated until the posterior data set meets the termination condition, and the optimized parameter combination is determined to obtain the parameters K and α of VMD.

[0187] Finally, the obtained parameters K and α are substituted into the VMD algorithm to obtain the decomposed K modes. Then, the optimal mode is selected through the Pearson coefficient, and the remaining modes are eliminated to finally obtain the denoised reconstructed data.

[0188] Step 2: Blind signal separation.

[0189] Specifically, when cracks occur in key structural parts of a fan, they are often accompanied by acoustic emission signals generated by impact, friction, etc. Accurately separating the crack extension signal from the collected mixed signal for detailed analysis is of great significance for accurately judging the crack extension state in the fan. The nonlinear principal component analysis algorithm (NLPCA) is used to separate the noise-reduced signal to effectively obtain the required signal.

[0190] First, the denoised signal s(t) is transformed into Linear mixing gives x(t), and the separation process is simplified to finding the time-varying system The generalized inverse matrix of The process is shown in the following equation (6):

[0191]

[0192] Secondly, according to the minimum mean square error approximation criterion as the independence criterion, the objective function is constructed, as shown in the above relationship (1).

[0193] Finally, the gradient descent algorithm is used to solve the problem, as shown in the above equation (2).

[0194] Step 3: Signal feature recognition: Starting from the time domain and frequency domain of acoustic emission signals, an effective method for pattern recognition of acoustic emission signals during crack propagation is established, and the bird flock optimization-one-dimensional convolutional neural network algorithm (BS-1DCNN) is used for feature recognition.

[0195] Specifically, in the BS-1DCNN fault diagnosis model, the BS-1DCNN algorithm takes advantage of the complementary characteristics of 1DCNN and SVM, combines the advantages of 1DCNN and SVM and improves their respective limitations. The relevant parameters in SVM are optimized by the bird flock algorithm, which effectively solves the local optimal problem and improves the classification accuracy. Finally, the bird flock optimized support vector machine classifier is used to finely identify the extracted fault features.

[0196] The main diagnostic process is as follows:

[0197] Firstly, the impeller crack soundprint signals appearing in this experiment were sorted and segmented, including the laboratory impeller material fatigue experiment soundprint signals and the model impeller experiment soundprint signals under different water environments, to form a fixed-segment data sample set.

[0198] Secondly, the data sample set is input into the 1CNN convolution layer, and the convolution kernel is used to adaptively extract fault features.

[0199] Furthermore, after the convolutional layer, the extracted features are output by the fully connected layer and sent to the output layer.

[0200] Furthermore, the bird flock algorithm is used to optimize the penalty parameter c in the SVM model and the parameter g of the radial basis kernel function.

[0201] Finally, the conventional softmax classifier is replaced by the bird flock optimized SVM classifier in the output layer of 1DCNN to further improve the classification accuracy and complete the fault diagnosis process.

[0202] Step 4: Construct a numerical model and establish the corresponding relationship between the changes in characteristic parameters of the acoustic emission signal and the crack propagation state.

[0203] Specifically, the characteristic parameters of the acoustic emission signal are used to reflect the crack growth rate da / dN along with the activity parameter dH / dN of the cycle number, and the corresponding relationship between the spectrum characteristics of the crack acoustic emission signal and the macroscopic expansion state of the crack is constructed.

[0204] Firstly, conventional metal materials for fans were selected to carry out the acoustic emission signal test of fatigue crack propagation of metal materials. The relationship curves of fatigue crack length a and stress intensity factor range ΔK with the number of cycles N, the relationship curves of fatigue crack propagation rate da / dN and ΔK, and the relationship diagram of acoustic emission characteristic parameters with the process of cycle number N were obtained.

[0205] Secondly, the Paris formula is used to establish the relationship between the characteristic parameters of the voiceprint signal and , as shown in the above relationship (4).

[0206] Furthermore, from the Paris formula, it can be seen that the relationship between the crack growth rate and ΔK is as shown in the above relationship (3).

[0207] Furthermore, by taking the logarithm of the above equations (3) and (4) and further deducing, we can obtain the rate of change of the cumulative value of the crack growth rate and the acoustic print characteristic parameters with the number of cycles: The quantitative relationship between them is shown in the above equation (5).

[0208] Finally, the activity parameters of the voiceprint signal characteristic parameters along with the cycle times are used Reflects the crack growth rate Then, an expression for the relationship between the characteristic parameters of the voiceprint signal and the crack growth rate can be established.

[0209] Step 5: Comprehensive analysis of crack propagation status.

[0210] Specifically, through steps one to three, the intelligent algorithm can be used to identify the crack propagation state, and finally the three main state labels of crack initiation, crack propagation, and crack instability will be obtained. Through step four, the relationship between the change of the characteristic parameters of the acoustic emission signal and the crack propagation rate can be analyzed to obtain the crack propagation rate in a specific period of time. Corresponding to the crack propagation rate curve, the state of the crack can be obtained, and the three main states of crack initiation, stable crack propagation, and unstable crack propagation can be output. Corresponding to the labels obtained by the intelligent algorithm, the crack propagation state can be jointly judged to improve the accuracy of crack propagation state identification.

[0211] The non-destructive testing optimization combination method for monitoring the crack propagation status of key components of wind turbines provided in this example has the following effects:

[0212] 1. This example proposes a new signal noise reduction, signal blind separation, and signal feature recognition method. Using the new method, the acoustic emission signal is processed in a whole set of signal noise reduction, separation, and recognition. It can analyze the acoustic emission signal from the perspective of time domain and frequency domain and obtain the state of the crack.

[0213] 2. This example proposes a method to establish the relationship between the characteristic parameters of the acoustic emission signal and the crack propagation state, which can realize the judgment of the crack propagation state from the perspective of the acoustic emission signal parameter method analysis;

[0214] 3. This example proposes to combine the time domain and frequency domain methods with the parameter method, and use the intelligent algorithm and the crack extension state relationship to jointly judge the state of the crack, rather than using a single method to evaluate the crack extension state, which can improve the accuracy of crack extension state judgment.

[0215] Therefore, the innovative crack propagation state identification method proposed in this example can ensure the accuracy of crack propagation state identification in a high noise environment, reduce the false alarm rate, and improve the accuracy of fan crack monitoring.

[0216] In this embodiment, a crack extension state detection device is also provided, which is used to implement the above-mentioned embodiments and preferred implementation modes, and the descriptions that have been made will not be repeated. As used below, the term "module" can implement a combination of software and / or hardware of a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0217] This embodiment provides a crack extension state detection device for a fan component; Figure 4 As shown, the device comprises:

[0218] The acquisition module 401 is used to acquire an initial acoustic emission signal set of the wind turbine.

[0219] The first processing module 402 is used to perform noise reduction and blind separation processing on the initial acoustic emission signal set to obtain a crack extension signal set of the wind turbine component.

[0220] The second processing module 403 is used to perform feature classification processing on the crack extension signal set to obtain a voiceprint signal feature parameter set reflecting the crack extension state.

[0221] The module 404 is established to establish the relationship between the characteristic parameters of the voiceprint signal and the crack growth rate by using the fatigue crack growth voiceprint signal test method and the Paris formula.

[0222] The detection module 405 is used to detect the crack propagation state of the fan component based on the voiceprint signal characteristic parameter set and the change relationship to obtain the crack propagation state detection result.

[0223] In some optional implementations, the first processing module 402 includes:

[0224] The first processing submodule is used to perform noise reduction processing on the initial acoustic emission signal set by using a preset signal noise reduction combination algorithm to obtain a target acoustic emission signal set.

[0225] The separation submodule is used to blindly separate the target acoustic emission signal set using a nonlinear principal component analysis algorithm to obtain a crack propagation signal set of the fan component.

[0226] In some optional embodiments, the separation submodule comprises:

[0227] The mixing unit is used to perform time-varying linear mixing on the target acoustic emission signal set to obtain a mixed acoustic emission signal set.

[0228] The construction unit is used to construct the objective function using the minimum mean square error approximation criterion.

[0229] The processing unit is used for obtaining a crack extension signal set based on the mixed acoustic emission signal set and the objective function through gradient descent algorithm processing.

[0230] In some optional implementations, the second processing module 403 includes:

[0231] The second processing submodule is used to process the crack extension signal set through a one-dimensional convolutional neural network algorithm to obtain a fault feature set.

[0232] The third processing submodule is used to process the fault feature set through a bird flock algorithm to obtain a soundprint signal feature parameter set reflecting the crack extension state.

[0233] In some optional implementations, the establishing module 404 includes:

[0234] A submodule is established to establish a first variation relationship curve between crack growth rate and stress intensity factor range and a second variation relationship curve between characteristic parameters of the soundprint signal and the number of cycles by using a fatigue crack growth soundprint signal test method.

[0235] The first determination submodule is used to determine the third correlation relationship between the characteristic parameters of the voiceprint signal and the range of the stress intensity factor and the fourth correlation relationship between the crack growth rate and the range of the stress intensity factor based on the first change relationship curve and the second change relationship curve using the Paris formula.

[0236] The second determination submodule is used to determine the change relationship between the voiceprint signal characteristic parameter and the crack growth rate according to the third association relationship and the fourth association relationship.

[0237] In some optional implementations, the detection module 405 includes:

[0238] The third determination submodule is used to determine the change of the voiceprint signal characteristic parameters and multiple first crack extension states according to the voiceprint signal characteristic parameter set.

[0239] The fourth determination submodule is used to determine a plurality of second crack extension states based on the change conditions and the change relationships.

[0240] The fifth determination submodule is used to determine the crack extension state detection result according to the multiple first crack extension states and the multiple second crack extension states.

[0241] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.

[0242] The crack extension state detection device in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0243] The embodiment of the present invention also provides a computer device having the above Figure 4 The crack extension state detection device shown.

[0244] See also Figure 5 , Figure 5 is a schematic diagram of the structure of a computer device provided by an optional embodiment of the present invention, such as Figure 5 As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of the GUI on an external input / output device (such as a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 5 A processor 10 is taken as an example.

[0245] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.

[0246] The memory 20 stores instructions executable by at least one processor 10, so that at least one processor 10 executes the method shown in the above embodiment.

[0247] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0248] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.

[0249] The computer device further comprises a communication interface 30 for the computer device to communicate with other devices or a communication network.

[0250] The embodiment of the present invention also provides a computer-readable storage medium. The method according to the embodiment of the present invention can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium through a network download, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.

[0251] A part of the present invention may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. Those skilled in the art should understand that the existence of the computer program instruction in a computer-readable medium includes, but is not limited to, a source file, an executable file, an installation package file, etc., and accordingly, the way in which the computer program instruction is executed by the computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium may be any available computer-readable storage medium or communication medium accessible to the computer.

[0252] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A crack extension state detection method for fan components; characterized in that: The method comprises: Obtaining an initial acoustic emission signal set of the fan; Performing noise reduction and blind separation processing on the initial acoustic emission signal set to obtain a crack extension signal set of the fan component; Performing feature classification processing on the crack extension signal set to obtain a voiceprint signal feature parameter set reflecting the crack extension state; Using the fatigue crack growth acoustic print signal test method and Paris formula, the relationship between the characteristic parameters of the acoustic print signal and the crack growth rate is established; Based on the voiceprint signal characteristic parameter set and the change relationship, the crack propagation state of the fan component is detected to obtain a crack propagation state detection result.

2. The method according to claim 1, characterized in that The initial acoustic emission signal set is subjected to noise reduction and blind separation processing to obtain a crack extension signal set of the fan component, including: Performing noise reduction processing on the initial acoustic emission signal set using a preset signal noise reduction combination algorithm to obtain a target acoustic emission signal set; The target acoustic emission signal set is blindly separated using a nonlinear principal component analysis algorithm to obtain the crack extension signal set of the fan component.

3. The method according to claim 2, characterized in that Using a nonlinear principal component analysis algorithm to blindly separate the target acoustic emission signal set to obtain the crack extension signal set of the fan component includes: Performing time-varying linear mixing on the target acoustic emission signal set to obtain a mixed acoustic emission signal set; The objective function is constructed using the minimum mean square error approximation criterion; Based on the mixed acoustic emission signal set and the objective function, the crack extension signal set is obtained through gradient descent algorithm processing.

4. The method according to claim 1, characterized in that: The crack extension signal set is subjected to feature classification processing to obtain a voiceprint signal feature parameter set reflecting the crack extension state, including: Processing the crack extension signal set by a one-dimensional convolutional neural network algorithm to obtain a fault feature set; The fault feature set is processed by a bird flock algorithm to obtain the voiceprint signal feature parameter set reflecting the crack propagation state.

5. The method according to claim 1, characterized in that: Using the fatigue crack growth acoustic signal test method and Paris formula, the relationship between the characteristic parameters of the acoustic signal and the crack growth rate is established, including: Using the fatigue crack growth acoustic print signal test method, a first variation relationship curve between the crack growth rate and the stress intensity factor range and a second variation relationship curve between the acoustic print signal characteristic parameter and the cycle number are established; Based on the first change relationship curve and the second change relationship curve, using the Paris formula to determine a third correlation relationship between the characteristic parameter of the voiceprint signal and the range of the stress intensity factor, and a fourth correlation relationship between the crack growth rate and the range of the stress intensity factor; The change relationship between the voiceprint signal characteristic parameter and the crack growth rate is determined according to the third association relationship and the fourth association relationship.

6. The method according to claim 1, characterized in that Based on the voiceprint signal characteristic parameter set and the change relationship, the crack extension state of the fan component is detected to obtain a crack extension state detection result, including: Determining a change in a characteristic parameter of a voiceprint signal and a plurality of first crack extension states according to the characteristic parameter set of the voiceprint signal; Based on the change condition and the change relationship, determining a plurality of second crack extension states; The crack extension state detection result is determined according to the plurality of first crack extension states and the plurality of second crack extension states.

7. A crack extension state detection device for fan components; characterized in that: The device comprises: An acquisition module, used for acquiring an initial acoustic emission signal set of the fan; A first processing module, configured to perform noise reduction and blind separation processing on the initial acoustic emission signal set to obtain a crack extension signal set of the fan component; A second processing module is used to perform feature classification processing on the crack extension signal set to obtain a voiceprint signal feature parameter set reflecting the crack extension state; Establish a module for establishing the relationship between characteristic parameters of the soundprint signal and the crack growth rate by using the fatigue crack growth soundprint signal test method and the Paris formula; A detection module is used to detect the crack propagation state of the fan component based on the voiceprint signal characteristic parameter set and the change relationship to obtain a crack propagation state detection result.

8. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the crack extension state detection according to any one of claims 1 to 6 by executing the computer instructions.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the crack extension state detection according to any one of claims 1 to 6.

10. A computer program product, characterized in that The method comprises computer instructions for causing a computer to execute the crack extension state detection according to any one of claims 1 to 6.

Images