Laser shock peening quality monitoring method and device based on acousto-optic holographic correlation fusion

Abstract

The invention discloses a laser shock peening quality monitoring method and device based on acousto-optic holographic correlation fusion. The method comprises the following steps: synchronously acquiring three paths of acoustic emission signals (a target material surface, a clamp surface and an impact area back surface) and one path of nanosecond plasma optical signal; performing threshold truncation processing on the acoustic emission signal to remove redundant information; dimension reduction processing is performed on the optical signal through an automatic convolution encoder module, so that the optical signal is matched with the acoustic emission signal in time length; and inputting the processed signal into the trained acousto-optic holographic correlation fusion model, and outputting a quality label. The model is composed of an improved InResTCN network formed by connecting Temporal Blocks in series and a time correlation fusion module, holographic features can be effectively extracted, acousto-optic information can be fused, and accurate monitoring of the laser shock peening quality is achieved. The objective of the invention is to complete information fusion by using holographic correlation of acousto-optic signals, so that the laser shock peening quality can obtain acousto-optic complementary information, thereby accurately evaluating the process quality.

Description

Technical Field

[0001] The present invention relates to the field of laser shock peening quality monitoring, and particularly to a laser shock peening quality monitoring method and device based on the fusion of acousto-optic holographic correlation. Background Art

[0002] The fatigue problem of key components has become a continuous challenge in all industries, especially in applications with complex mechanical structures, such as aerospace, railways, and seagoing ships. Overcoming the challenges related to fatigue degradation, deformation, and crack initiation has become the focus of industry attention. As a new type of surface treatment technology, laser shock peening (LSP) can induce beneficial residual stress layers at vulnerable positions on the material surface using high-energy laser beams, thereby reducing material fatigue damage and extending the service life of components. However, during the laser processing process, uncontrollable abnormalities may occur, such as the rupture of the water confinement layer and the protective layer, resulting in insufficient strengthening of the material surface. Therefore, it is very necessary to realize online quality monitoring of LSP based on processing information, which helps to avoid the disadvantages of traditional offline monitoring, such as high economic costs and time consumption.

[0003] The laser-induced plasma shock wave is the direct driving force for improving the surface performance of materials. Therefore, some scholars have established the connection between signals and quality based on the analysis of the time-domain characteristics and frequency-domain characteristics of acoustic emission. However, acoustic emission signals are non-linear, non-stationary, and contain a large number of complex time-varying components, making it difficult for simple statistical features to fully characterize the information related to processing quality. On the other hand, during the laser shock peening process, plasma optical signals also contain rich information, and the research on acousto-optic holographic fusion is still not perfect at present. Summary of the Invention

[0004] Aiming at the problems existing in the prior art, the present invention provides a laser shock peening quality monitoring method and device based on the fusion of acousto-optic holographic correlation, aiming to complete information fusion using the holographic correlation of acousto-optic signals, so that the laser shock peening quality can obtain complementary information of sound and light to accurately evaluate the process quality.

[0005] To solve the above technical problems, the present invention is achieved through the following technical solutions: According to the first aspect of the present invention, a laser shock peening quality monitoring method based on the fusion of acousto-optic holographic correlation is provided, including: Synchronously acquiring three-way acoustic emission signals and one-way optical signal during the laser shock peening process, where the three-way acoustic emission signals include the acoustic emission signal AE1 on the surface of the target material, the acoustic emission signal AE2 on the surface of the fixture, and the acoustic emission signal AE3 on the back of the impact area, and the one-way optical signal is the nanosecond-level plasma optical signal PMT generated by the target element in the target material; Perform threshold truncation processing on the three-channel acoustic emission signals, delete the redundant information at the front and back ends of the three-channel acoustic emission signals, and retain the effective middle part; Perform dimensionality reduction processing on the optical signal through a pre-trained automatic convolutional encoder module to match the time length of the optical signal with that of the three-channel acoustic emission signals; Input the processed three-channel acoustic emission signals and optical signals into the trained acousto-optic holographic correlation fusion model to output the laser shock peening quality label; wherein, the acousto-optic holographic correlation fusion model includes an InResTCN network and a time correlation fusion module connected in sequence, and the InResTCN network is composed of four improved Temporal Blocks connected in series, and each improved Temporal Block realizes holographic feature extraction and dimensionality transformation through dilated causal convolution, weight normalization, activation function and residual connection.

[0006] In a possible implementation manner of the first aspect, the threshold truncation processing of the three-channel acoustic emission signals is specifically: Set the front-end truncation threshold to 0.1 and the back-end truncation threshold to 0.4 for the three-channel acoustic emission signals respectively, and delete the sampling points below the threshold.

[0007] In a possible implementation manner of the first aspect, the operation formula of the automatic convolutional encoder module is:

[0008] Wherein, is the optical signal attenuation segment sequence, is the weight of the convolutional kernel, is the bias term, is the size of the convolutional kernel, that is, the length of the convolutional kernel, is the stride that controls the sliding of the convolutional operation on the input signal, is the index of the convolutional kernel, is the output signal after dimensionality reduction processing, is the index of the optical signal attenuation segment after downsampling..

[0009] In a possible implementation manner of the first aspect, the training method of the automatic convolutional encoder module is specifically as follows: The automatic convolutional encoder module uses the results of equal-interval downsampling, maximum downsampling, average downsampling and median downsampling as the optimization target, sends the training loss of the automatic convolutional encoder module into the RMSprop optimizer for backpropagation, and the automatic convolutional encoder module can optimize its convolutional kernel parameters, so as to learn the advantages of equal-interval downsampling, maximum downsampling, average downsampling and median downsampling strategies and effectively combine them to complete the training of the automatic convolutional encoder module, specifically:

[0010]

[0011]

[0012] In the formula, is the downsampling factor; , , , are the results of equidistant downsampling, maximum downsampling, average downsampling, and median downsampling respectively; , , , are the equidistant downsampling loss, maximum downsampling loss, average downsampling loss, and median downsampling loss respectively; is the training loss of the automatic convolutional encoder module.

[0013] In a possible implementation manner of the first aspect, the four improved Temporal Blocks are Temporal Block1, Temporal Block2, Temporal Block3, and Temporal Block4, which sequentially perform the following dimensionality transformations: Temporal Block1 is used to reduce the time dimension by 1 / 2 and expand the feature dimension by 4 times; Temporal Block2 is used to reduce the time dimension by 1 / 5 and expand the feature dimension by 2 times; Temporal Block3 is used to reduce the time dimension by 1 / 2 and expand the feature dimension by 4 times; Temporal Block4 is used to reduce the time dimension by 1 / 5 and expand the feature dimension by 2 times.

[0014] In a possible implementation manner of the first aspect, the time correlation fusion module performs the following operations: (a) The processed AE1, AE2, AE3, and PMT signals are sliced into multiple segments in the time dimension, and the time-delay Pearson correlation coefficient matrices of the signal pairs AE1-PMT, AE2-PMT, AE3-PMT, AE1-AE2, AE1-AE3, and AE2-AE3 are calculated respectively; (b) The time-delay Pearson correlation coefficient matrices of each signal pair are weighted by 12 trainable parameters, the weighted matrix is multiplied by the deep features output by the InResTCN network to obtain a feature matrix, and the feature matrix is non-linearly transformed by the Tanh activation function; (c) Multiply the feature matrix processed by the Tanh activation function with the trainable vector in the InResTCN network and compress it into a one-dimensional vector; (d) Process the one-dimensional vector using the Softmax function to amplify the peak part, multiply the four-channel signals with the weight vector to achieve weighting; (e) Perform weighted summation on the four-channel signal features and output the laser shock peening quality label.

[0015] According to the second aspect of the present invention, there is provided a laser shock peening quality monitoring device based on acousto-optic holographic correlation fusion, comprising: An acquisition module for synchronously acquiring three-channel acoustic emission signals and one-channel optical signal during the laser shock peening process. The three-channel acoustic emission signals include the acoustic emission signal AE1 on the surface of the target material, the acoustic emission signal AE2 on the surface of the fixture, and the acoustic emission signal AE3 on the back of the impact area. The one-channel optical signal is the nanosecond-level plasma optical signal PMT generated by the target element in the target material; A truncation module for performing threshold truncation processing on the three-channel acoustic emission signals, deleting the redundant information at the front and back ends of the three-channel acoustic emission signals, and retaining the middle effective part; A dimensionality reduction module for performing attenuation segment dimensionality reduction processing on the optical signal through a pre-trained automatic convolutional encoder module to match the time length of the optical signal with the three-channel acoustic emission signals; An output module for inputting the processed three-channel acoustic emission signals and optical signal into the trained acousto-optic holographic correlation fusion model and outputting the laser shock peening quality label. Wherein, the acousto-optic holographic correlation fusion model includes an InResTCN network and a time correlation fusion module connected in sequence. The InResTCN network is composed of four improved Temporal Blocks connected in series. Each improved Temporal Block realizes holographic feature extraction and dimensionality transformation through dilated causal convolution, weight normalization, activation function, and residual connection.

[0016] According to the third aspect of the present invention, there is provided a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the above-mentioned method for monitoring the quality of laser shock peening based on acousto-optic holographic correlation fusion.

[0017] According to the fourth aspect of the present invention, there is provided a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, it implements the above-mentioned method for monitoring the quality of laser shock peening based on acousto-optic holographic correlation fusion.

[0018] According to a fifth aspect of the present invention, there is provided a computer program product which, when executed by a processor, implements the method for monitoring the quality of laser shock peening based on acousto-optic holographic correlation fusion described above.

[0019] Compared with the prior art, the present invention has at least the following beneficial effects: The method for monitoring the quality of laser shock peening based on acousto-optic holographic correlation fusion provided by the present invention synchronously acquires the acoustic emission signals on the surface of the target material, the acoustic emission signals on the surface of the fixture, the acoustic emission signals on the back of the impact area, and the plasma optical signals during the laser shock peening process, realizing the simultaneous acquisition of multi-source information and the fusion utilization of multi-source information. Compared with relying only on a single signal source (such as relying only on acoustic emission signals or optical signals), it can more comprehensively and accurately reflect the process state of laser shock peening, thereby improving the monitoring accuracy of the quality of laser shock peening. Aiming at the problem of large differences in the sampling rates of acousto-optic signals, the present invention uses signal threshold clipping and an automatic convolutional encoder module for processing. The redundant information at the front and back ends of the acoustic emission signals is removed by the threshold truncation method, and at the same time, the automatic convolutional encoder module is used to perform dimensionality reduction processing on the optical signals, eliminating the differences in the sampling rates of acousto-optic signals in the time dimension and ensuring the consistency of acousto-optic holographic correlation fusion. Cooperating with the InResTCN network and the time correlation fusion module of the present invention, feature extraction and fusion of multi-source information are performed. The InResTCN network realizes the efficient extraction and dimensionality transformation of holographic features through an improved Temporal Block structure; the time correlation fusion module utilizes the holographic correlation between acousto-optic signals to realize the deep fusion of information. The collaborative processing method not only improves the ability of time series analysis but also makes the monitoring results more interpretable.

[0020] To make the above objects, features, and advantages of the present invention more clearly understandable, the following specifically enumerates preferred embodiments and, in conjunction with the accompanying drawings, makes the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the specific embodiments of the present invention, the following will briefly introduce the drawings required for use in the description of the specific embodiments. Obviously, the following drawings are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0022] Figure 1 It is a flowchart of the method for monitoring the quality of laser shock peening based on acousto-optic holographic correlation fusion of the present invention; Figure 2 It is the original waveforms of three-channel acoustic emission signals; Figure 3 It is the original waveform of the optical signal; Figure 4 is the InResTCN network framework; Figure 5 is the time-correlation fusion module framework; Figure 6 is the automatic convolutional encoder module framework. Detailed implementation manners

[0023] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0024] As Figure 1 shown, the embodiments of the present invention provide a method for monitoring the quality of laser shock peening based on acousto-optic holographic correlation fusion, which specifically includes the following steps: S1. Synchronously acquire three-way acoustic emission signals and one-way optical signal during the laser shock peening process. The three-way acoustic emission signals include the acoustic emission signal AE1 on the surface of the target material, the acoustic emission signal AE2 on the surface of the fixture, and the acoustic emission signal AE3 on the back of the impact area. The one-way optical signal is the nanosecond-level plasma optical signal PMT generated by the target element in the target material.

[0025] Specifically, the acoustic emission signals AE1, AE2, and AE3 are acquired through a multi-channel acoustic emission acquisition system. The acquisition system includes a broadband acoustic emission sensor, an attenuator, an acquisition card, and acquisition software. Among them, the AE1 sensor is installed on the front of the target material and the linear distance from the impact area does not exceed 5 cm. The AE2 sensor is installed on the flat area of the fixture surface and close to the impact area. The AE3 sensor is installed directly behind the impact area on the back of the target material.

[0026] The optical signal PMT is acquired through a microchannel photomultiplier tube and a narrowband filter. The narrowband filter allows the characteristic wavelength optical signal generated by the target element in the target material to pass through.

[0027] S2. Perform threshold truncation processing on the three-way acoustic emission signals, delete the redundant information at the front and back ends of the three-way acoustic emission signals, and retain the middle effective part.

[0028] That is to say, by performing threshold truncation processing on the three-way acoustic emission signals collected, deleting the redundant information below the threshold at the front and back ends of the signals, and only retaining the middle effective part of the signals, noise interference can be reduced and the quality of the signals can be improved.

[0029] Exemplarily, as Figure 2As shown, the acoustic emission signal is a holographic signal with an amplitude less than ±0.5V at both the front and rear ends and a sampling rate of 3MHz.

[0030] In one implementable manner, threshold truncation processing is performed on the three-channel acoustic emission signals. Specifically, the front-end truncation threshold is set to 0.1 and the rear-end truncation threshold is set to 0.4 for the three-channel acoustic emission signals respectively, and the sampling points below the threshold are deleted.

[0031] S3. Perform dimensionality reduction processing on the optical signal through a pre-trained automatic convolutional encoder module to match the time lengths of the optical signal and the three-channel acoustic emission signals.

[0032] That is, several acoustic emission signals and optical signals with the same time length are obtained. Exemplarily, as Figure 3 shown, the optical signal is a holographic signal with a sampling rate of 5GHz, and its attenuation section covers approximately 7*10 6 data points. The automatic convolutional encoder module reduces its data points to 1.5*10 4 through convolutions with multiple different convolutional kernels and strides, and finally realizes that the time lengths of both the acoustic emission signal and the optical signal are 2*10 4 data points.

[0033] In one implementable manner, the operation formula of the automatic convolutional encoder module is:

[0034] where, is the optical signal attenuation section sequence, is the weight of the convolutional kernel, is the bias term, is the size of the convolutional kernel, that is, the length of the convolutional kernel, is the stride that controls the sliding of the convolutional operation on the input signal, is the index of the convolutional kernel, is the output signal after dimensionality reduction processing, is the optical signal attenuation section index after downsampling.

[0035] As Figure 6 shown, the training method of the automatic convolutional encoder module is as follows: The automatic convolutional encoder module uses the results of equal-interval downsampling, maximum downsampling, average downsampling, and median downsampling as the optimization target, sends the training loss of the automatic convolutional encoder module into the RMSprop optimizer for backpropagation, and the automatic convolutional encoder module can optimize its convolutional kernel parameters, thereby learning the advantages of equal-interval downsampling, maximum downsampling, average downsampling, and median downsampling strategies and effectively combining them to complete the training of the automatic convolutional encoder module. Specifically:

[0036]

[0037]

[0038] In the formula, is the downsampling factor (that is, one point in every sampling points is retained); , , , are the results of equally-spaced downsampling, maximum downsampling, average downsampling, and median downsampling respectively; , , , are the equally-spaced downsampling loss, maximum downsampling loss, average downsampling loss, and median downsampling loss respectively; is the training loss of the automatic convolutional encoder module.

[0039] S4. Input the processed three-channel acoustic emission signals and optical signals into the trained acousto-optic holographic correlation fusion model to output the laser shock peening quality label; wherein, the acousto-optic holographic correlation fusion model includes an InResTCN network and a temporal correlation fusion module connected in sequence, and the InResTCN network is composed of four improved Temporal Blocks connected in series. Each improved Temporal Block realizes holographic feature extraction and dimensional transformation through dilated causal convolution, weight normalization, activation function, and residual connection.

[0040] In an implementable manner, the four improved Temporal Blocks are Temporal Block1, Temporal Block2, Temporal Block3, and Temporal Block4 respectively, and the following dimensional transformations are realized in sequence: Temporal Block1 is used to reduce the time dimension by 1 / 2 and expand the feature dimension by 4 times; Temporal Block2 is used to reduce the time dimension by 1 / 5 and expand the feature dimension by 2 times; Temporal Block3 is used to reduce the time dimension by 1 / 2 and expand the feature dimension by 4 times; Temporal Block4 is used to reduce the time dimension by 1 / 5 and expand the feature dimension by 2 times.

[0041] Such as Figure 4As shown, the Temporal Block is the basic building block of the InResTCN network. The InResTCN network is composed of improved Temporal Blocks, with Dilated Causal Conv-WeightNorm-ReLU-Dropout connected in series inside, denoted as DC-WN-Re-Dp, that is, dilated causal convolution, weight normalization, activation function, and residual connection. Different from the traditional Temporal Block, the composition of the Temporal Block is improved by building in a residual connection, and its internal structure is adjusted to three DC-WN-Re-Dp connected in series, which respectively achieve the functions of holographic length reduction, feature dimension increase, and residual connection in turn, so that each Temporal Block can flexibly and gradually realize the transformation of the feature matrix towards the target dimension. The signal gradually changes from a long time series with a single feature dimension to a short time series with a rich feature dimension during the process of passing through four Temporal Blocks.

[0042] In one embodiment, the InResTCN network is composed of four Temporal Blocks. The initial dimension of the data is [1, 20000]. After the data is input into Temporal Block1, it sequentially passes through Temporal Block2, Temporal Block3, and Temporal Block4. Among them, Temporal Block1 reduces the data time dimension by 1 / 2 and expands the feature dimension by 4 times, and the dimension becomes [4, 5000] at this time; Temporal Block2 reduces the data time dimension by 1 / 5 and expands the feature dimension by 2 times, and the dimension becomes [8, 1000] at this time; Temporal Block3 reduces the data time dimension to 1 / 2 and expands the feature dimension to 4 times, and the dimension becomes [32, 500] at this time; Temporal Block4 reduces the data time dimension by 1 / 5 and expands the feature dimension by 2 times, and the dimension becomes [64, 200] at this time.

[0043] The signal gradually changes from a long time series with a feature dimension of 1 and a time dimension of 20000 to a short time series with a feature dimension of 64 and a time dimension of 200 during the process of passing through four Temporal Blocks.

[0044] In an implementable manner, the time correlation fusion module performs the following operations: (a)The processed AE1, AE2, AE3, and PMT signals are sliced into multiple segments in the time dimension, and the time-delay Pearson correlation coefficient matrices of the signal pairs AE1-PMT, AE2-PMT, AE3-PMT, AE1-AE2, AE1-AE3, and AE2-AE3 are calculated respectively; (b)The time-delay Pearson correlation coefficient matrices of each signal pair are weighted by 12 trainable parameters. The weighted matrix is multiplied by the deep features output by the InResTCN network to obtain a feature matrix, and the feature matrix is non-linearly transformed by the Tanh activation function; (c)The feature matrix processed by the Tanh activation function is multiplied by the trainable vector in the InResTCN network and compressed into a one-dimensional vector; (d)The Softmax function is used to process the one-dimensional vector to amplify the peak part, and the four-channel signals are multiplied by the weight vector to achieve weighting; (e)The four-channel signal features are weighted and summed to output the laser shock peening quality label.

[0045] Specifically, as Figure 5 shown, each channel of signal is sliced into multiple segments in the time dimension, that is, AE1, AE2, AE3, and PMT are all sliced in the same frame length and frame shift manner to obtain several signal slices. For each pair of signals, including AE1 and PMT, AE2 and PMT, AE3 and PMT, AE1 and AE2, AE1 and AE3, AE2 and AE3, the Pearson correlation coefficient between different time segments is calculated, that is, the Pearson correlation coefficient is calculated for all slices of all the above signals, so as to realize the quantification of the correlation between signals within the same time period and between different time periods. The dynamic holographic dependence relationship between signals will be reflected in the form of a time-delay Pearson correlation coefficient matrix, thereby revealing the dynamic holographic dependence relationship between signals. Taking AE1-PMT as an example, the calculated time-delay Pearson correlation coefficient matrix reflects the interaction between the two signals in different time periods.

[0046] Through the above process, a time-delay Pearson correlation coefficient matrix can be constructed for each pair of signals. The horizontal and vertical axes of the time-delay Pearson correlation coefficient matrix represent the number of signal time slices, and each element in the matrix represents the correlation between the corresponding time segments. Specifically, the time-delay Pearson correlation coefficient matrices of the following signal pairs are calculated respectively: AE1-PMT, AE2-PMT, AE3-PMT, AE1-AE2, AE1-AE3, AE2-AE3.

[0047] The weighted correlation matrices of each signal are obtained through 12 trainable parameters K1 - K12. K1 - K12 are parameters that exist during model initialization and are optimized in value during model training with gradient descent. The specific process of obtaining the weighted matrices for the four signals is as follows: K1 * AE1 - PMT + K2 * AE1 - AE2 + K3 * AE1 - AE3 = AE1 weighted matrix K4 * AE2 - PMT + K5 * AE1 - AE2 + K6 * AE2 - AE3 = AE2 weighted matrix K7 * AE3 - PMT + K8 * AE2 - AE3 + K9 * AE1 - AE3 = AE3 weighted matrix K10 * AE1 - PMT + K11 * AE2 - PMT + K12 * AE3 - PMT = PMT weighted matrix Multiply the weighted matrix by the deep features output by the InResTCN network to obtain a feature matrix, and perform a non - linear transformation through the Tanh activation function. Tanh is a smooth S - shaped curve, and the output through Tanh is from - 1 to 1.

[0048] The feature matrix after being processed by the activation function is multiplied by a trainable vector generated within the model, and its shape is adjusted to obtain a one - dimensional vector. That is to say, the feature matrix after being processed by the activation function has the same dimension as the output matrix of InResTCN, both being [64, 200]. And this trainable vector is [1, 200] with a feature dimension of 1. After multiplication, the feature dimension of the feature matrix will change from 64 to 1.

[0049] Use the Softmax function to process the obtained one - dimensional vector, amplify the peak part therein, thereby emphasizing key information, and multiply the four signals by their weight vectors to achieve weighting. That is to say, AE1, AE2, AE3, and PMT are all multiplied by their corresponding weight vectors. This weight vector only changes the eigenvalue of each frame in the time dimension and does not change the dimension size of the feature matrix.

[0050] Finally, sum the weighted features to complete the information fusion of the four types of signals and finally output a label representing the laser strengthening quality. That is to say, the feature matrices after weighting the four signals have the same dimension, and their time steps correspond one by one, enabling simple matrix summation. The process of outputting the label comes from the common method of compressing the fusion matrix into the corresponding number of classifications through a linear layer.

[0051] In another embodiment of the present invention, a laser shock peening quality monitoring device based on acousto-optic holographic correlation fusion is provided for implementing the above-mentioned laser shock peening quality monitoring method based on acousto-optic holographic correlation fusion. Specifically, it includes an acquisition module, a truncation module, a dimensionality reduction module, and an output module. The functions of each module are configured as follows: The acquisition module is used to synchronously acquire three acoustic emission signals and one optical signal during the laser shock peening process. The three acoustic emission signals include the acoustic emission signal AE1 on the surface of the target material, the acoustic emission signal AE2 on the surface of the fixture, and the acoustic emission signal AE3 on the back of the impact area. The one optical signal is the nanosecond-level plasma optical signal PMT generated by the target element in the target material.

[0052] The truncation module is used to perform threshold truncation processing on the three acoustic emission signals, delete the redundant information at the front and back of the three acoustic emission signals, and retain the middle effective part.

[0053] The dimensionality reduction module is used to perform dimensionality reduction processing on the optical signal in the attenuation section through a pre-trained automatic convolutional encoder module to make the time length of the optical signal match that of the three acoustic emission signals.

[0054] The output module is used to input the processed three acoustic emission signals and optical signal into the trained acousto-optic holographic correlation fusion model and output the laser shock peening quality label. Among them, the acousto-optic holographic correlation fusion model includes an InResTCN network and a time correlation fusion module connected in sequence. The InResTCN network is composed of four improved Temporal Blocks connected in series. Each improved Temporal Block realizes holographic feature extraction and dimensionality transformation through dilated causal convolution, weight normalization, activation function, and residual connection.

[0055] All relevant contents of each step involved in the foregoing embodiment of the laser shock peening quality monitoring method based on acousto-optic holographic correlation fusion can be cited in the function description of the corresponding functional modules of the laser shock peening quality monitoring device based on acousto-optic holographic correlation fusion in the embodiment of the present invention, and will not be elaborated here. The division of modules in the embodiment of the present invention is illustrative, only a logical function division. In actual implementation, there may be other division methods. In addition, in each embodiment of the present invention, the functional modules can be integrated in one processor, or can exist separately physically, or two or more modules can be integrated in one module. The above integrated modules can be implemented in the form of hardware or in the form of software functional modules.

[0056] In another embodiment of the present invention, a computer device is provided. The computer device includes a processor and a memory. The memory is used to store a computer program, and the computer program includes program instructions. The processor is used to execute the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or may also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, and is suitable for implementing one or more instructions. Specifically, it is suitable for loading and executing one or more instructions in the computer storage medium to implement the corresponding method flow or corresponding function. The processor described in the embodiment of the present invention can be used for the operation of a method for monitoring the quality of laser shock peening based on acousto-optic holographic correlation fusion.

[0057] In another embodiment of the present invention, a storage medium is also provided, specifically a computer-readable storage medium (Memory). The computer-readable storage medium is a memory device in a computer device and is used to store programs and data. It can be understood that the computer-readable storage medium here can include both the built-in storage medium in the computer device and, of course, the extended storage medium supported by the computer device. The computer-readable storage medium provides a storage space, and the operating system of the terminal is stored in this storage space. And, one or more instructions suitable for being loaded and executed by the processor are also stored in this storage space. These instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory. One or more instructions stored in the computer-readable storage medium can be loaded and executed by the processor to implement the corresponding steps of the method for monitoring the quality of laser shock peening based on acousto-optic holographic correlation fusion in the above embodiments.

[0058] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memory, CD-ROM, optical memory, etc.) that contain computer-usable program code.

[0059] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of the flows and / or blocks in the flowchart and / or block diagram can also be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in one Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0060] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including instruction means, and the instruction means implements the functions specified in one Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0061] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process. Therefore, the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0062] The present invention also provides a computer program product, which is used to execute any one of the above-mentioned laser shock peening quality monitoring methods based on acousto-optic holographic correlation. Since the computer program product provided by the present invention and the above-mentioned laser shock peening quality monitoring method based on acousto-optic holographic correlation belong to the same inventive concept, the computer program product provided by the present invention has all the advantages of the above-mentioned laser shock peening quality monitoring method based on acousto-optic holographic correlation. Therefore, the beneficial effects of the computer program product provided by the present invention will not be elaborated one by one here.

[0063] In the present invention, the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples", etc. mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0064] Finally, it should be noted that the above-mentioned embodiments are only specific embodiments of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit it. The protection scope of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that any person skilled in the art within the technical scope disclosed by the present invention can still modify the technical solutions described in the foregoing embodiments or easily conceive changes, or perform equivalent replacements for some of the technical features; and these modifications, changes, or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A method for monitoring the quality of laser shock peening based on the fusion of acousto-optic holographic correlation, characterized in that, Including: Synchronously acquiring three-way acoustic emission signals and one-way optical signal during the laser shock peening process. The three-way acoustic emission signals include the acoustic emission signal AE1 on the surface of the target material, the acoustic emission signal AE2 on the surface of the fixture, and the acoustic emission signal AE3 on the back of the impact area. The one-way optical signal is the nanosecond-level plasma optical signal PMT generated by the target element in the target material; Performing threshold truncation processing on the three-way acoustic emission signals, deleting the redundant information at the front end and the back end of the three-way acoustic emission signals, and retaining the middle effective part; Performing attenuation segment dimensionality reduction processing on the optical signal through a pre-trained automatic convolutional encoder module to make the time length of the optical signal match that of the three-way acoustic emission signals; Inputting the processed three-way acoustic emission signals and optical signal into the trained acoustic-optic holographic correlation fusion model to output the laser shock peening quality label; wherein, the acoustic-optic holographic correlation fusion model includes an InResTCN network and a time correlation fusion module connected in sequence. The InResTCN network is composed of four improved Temporal Blocks connected in series. Each improved Temporal Block realizes holographic feature extraction and dimensionality transformation through dilated causal convolution, weight normalization, activation function, and residual connection.

2. The method for monitoring the quality of laser shock peening based on acousto-optic holographic correlation fusion according to claim 1, wherein The specific operation of performing threshold truncation processing on the three-way acoustic emission signals is as follows: Setting the front-end truncation threshold to 0.1 and the back-end truncation threshold to 0.4 for the three-way acoustic emission signals respectively, and deleting the sampling points below the threshold.

3. A method for monitoring the quality of laser shock peening based on acousto-optic holographic correlation fusion according to claim 1, characterized in that, The operation formula of the automatic convolutional encoder module is: Among them, is the optical signal attenuation segment sequence, is the weight of the convolution kernel, is the bias term, is the size of the convolution kernel, that is, the length of the convolution kernel, is the stride that controls the sliding step of the convolution operation on the input signal, is the index of the convolution kernel, is the output signal after dimensionality reduction processing, is the optical signal attenuation segment index after downsampling.

4. A method for monitoring the quality of laser shock peening based on acousto-optic holographic correlation fusion according to claim 3, wherein The training method of the automatic convolutional encoder module is specifically as follows: The automatic convolutional encoder module takes the results of equally-spaced downsampling, maximum downsampling, average downsampling, and median downsampling as the optimization objective, sends the training loss of the automatic convolutional encoder module into the RMSprop optimizer for backpropagation. The automatic convolutional encoder module can optimize its convolution kernel parameters, thereby learning the advantages of equally-spaced downsampling, maximum downsampling, average downsampling, and median downsampling strategies and effectively combining them to complete the training of the automatic convolutional encoder module. Specifically: In the formula, is the downsampling factor; , , , are the results of equally-spaced downsampling, maximum downsampling, average downsampling, and median downsampling, respectively; , , , are the equally-spaced downsampling loss, maximum downsampling loss, average downsampling loss, and median downsampling loss, respectively; is the training loss of the automatic convolutional encoder module.

5. A method for monitoring the quality of laser shock peening based on acousto-optic holographic correlation fusion according to claim 1, characterized in that The four improved Temporal Blocks are Temporal Block1, Temporal Block2, Temporal Block3, and Temporal Block4, which sequentially perform the following dimensionality transformations: Temporal Block1 is used to reduce the time dimension by 1 / 2 and expand the feature dimension by 4 times; Temporal Block2 is used to reduce the time dimension by 1 / 5 and expand the feature dimension by 2 times; Temporal Block3 is used to reduce the time dimension by 1 / 2 and expand the feature dimension by 4 times; Temporal Block4 is used to reduce the time dimension by 1 / 5 and expand the feature dimension by 2 times.

6. A method for monitoring the quality of laser shock peening based on acousto-optic holographic correlation fusion according to claim 1, characterized in that The time correlation fusion module performs the following operations: (a)The processed AE1, AE2, AE3, and PMT signals are sliced into multiple segments in the time dimension, and the time-delay Pearson correlation coefficient matrices of the signal pairs AE1-PMT, AE2-PMT, AE3-PMT, AE1-AE2, AE1-AE3, and AE2-AE3 are calculated respectively; (b)The time-delay Pearson correlation coefficient matrices of each signal pair are weighted by 12 trainable parameters, the weighted matrix is multiplied by the deep features output by the InResTCN network to obtain a feature matrix, and the feature matrix is non-linearly transformed by the Tanh activation function; (c)The feature matrix processed by the Tanh activation function is multiplied by the trainable vector in the InResTCN network and compressed into a one-dimensional vector; (d)The Softmax function is used to process the one-dimensional vector to amplify the peak part, and the four-way signals are multiplied by the weight vector to achieve weighting; (e)The four-way signal features are weighted and summed to output the laser shock peening quality label.

7. A laser shock peening quality monitoring device based on acousto-optic holographic correlation fusion, characterized in that, Including: An acquisition module for synchronously acquiring three acoustic emission signals and one optical signal during the laser shock peening process. The three acoustic emission signals include the acoustic emission signal AE1 on the surface of the target, the acoustic emission signal AE2 on the surface of the fixture, and the acoustic emission signal AE3 on the back of the impact area. The one optical signal is the nanosecond-level plasma optical signal PMT generated by the target element in the target; A truncation module for performing threshold truncation processing on the three acoustic emission signals, deleting the redundant information at the front and back ends of the three acoustic emission signals, and retaining the middle effective part; A dimensionality reduction module for performing attenuation segment dimensionality reduction processing on the optical signal through a pre-trained auto convolutional encoder module to match the time lengths of the optical signal and the three acoustic emission signals; An output module for inputting the processed three acoustic emission signals and optical signal into the trained acousto-optic holographic correlation fusion model to output the laser shock peening quality label. Among them, the acousto-optic holographic correlation fusion model includes an InResTCN network and a time correlation fusion module connected in sequence. The InResTCN network is composed of four improved Temporal Blocks connected in series, and each improved Temporal Block realizes holographic feature extraction and dimensionality transformation through dilated causal convolution, weight normalization, activation function, and residual connection.

8. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements a laser shock peening quality monitoring method based on acousto-optic holographic correlation fusion according to any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements a laser shock peening quality monitoring method based on acousto-optic holographic correlation fusion according to any one of claims 1 to 6.

10. A computer program product, characterized in that, When the computer program product is executed by the processor, it implements a laser shock peening quality monitoring method based on acousto-optic holographic correlation fusion according to any one of claims 1 to 6.

Images