Laser vibration measurement signal optimization method, equipment and medium

Abstract

The invention discloses a laser vibration measurement signal optimization method and device and a medium, and belongs to the technical field of laser vibration measurement signal optimization. The method comprises the following steps: collecting a vibration signal of a tunnel lining surface; performing spectrum leakage suppression processing on the vibration signal based on a three-order five-item maximum sidelobe attenuation self-convolution window to obtain a suppression signal; performing variational mode decomposition noise reduction on the vibration signal through an improved chicken flock optimization algorithm to obtain a noise reduction signal; inputting the suppression signal and the noise reduction signal into a Transform model to obtain an optimized vibration signal; performing time-frequency domain feature extraction on the optimized vibration signal based on short-time Fourier transform and a wavelet packet decomposition algorithm to obtain a time-frequency feature parameter; processing the time-frequency characteristic parameters based on an adaptive filtering algorithm to obtain optimized data; and performing real-time correction on the optimization parameter through a compensation calibration algorithm to obtain a calibration vibration signal. According to the method, the technical effect of optimizing the laser vibration measurement signal is achieved.

Description

Technical Field

[0001] This application relates to the technical field of laser vibration measurement signal optimization, and particularly relates to a laser vibration measurement signal optimization method, device and medium. Background Art

[0002] Laser Doppler Vibrometry (LDV) is a non-contact vibration measurement method based on the Doppler effect, which has wide applications in the fields of mechanical structure health monitoring, precision instrument research and development, and civil engineering. This technology irradiates the surface of the object to be measured with a laser beam, and detects the Doppler frequency shift of the reflected light to obtain vibration information, avoiding the dependence on the surface state and installation conditions of the object to be measured of traditional contact sensors (such as accelerometers and velocity sensors), and is particularly suitable for vibration measurement scenarios of high-temperature, high-speed or vulnerable structures.

[0003] However, the existing laser Doppler vibrometry technology has the following problems: When performing spectral analysis on the laser Doppler vibrometry signal using the Fast Fourier Transform (FFT), if the sampling frequency is not synchronized with the signal frequency, it will cause the spectral energy to spread from the true frequency component to adjacent frequency bands, that is, the spectral leakage phenomenon. Thereby introducing measurement errors and reducing the accuracy of the measurement results. The laser Doppler vibrometry signal is often mixed with various noise interferences such as environmental noise and circuit noise. The above-mentioned noise will mask the characteristics of the true vibration signal, reduce the signal-to-noise ratio, making it difficult for a single signal processing method to effectively extract the vibration information, thus affecting the measurement accuracy.

[0004] In order to improve the measurement accuracy of the laser Doppler vibrometry signal, the existing technology usually uses a single signal processing method (such as the window function method, modal decomposition method, etc.) for noise reduction processing, but it is difficult to simultaneously meet the dual requirements of spectral leakage suppression and noise stripping.

[0005] Therefore, how to optimize the laser vibration measurement signal has become an urgent technical problem to be solved. Summary of the Invention

[0006] The embodiments of this application provide a laser vibration measurement signal optimization method, device and medium, which are used to solve the following technical problem: how to optimize the laser vibration measurement signal.

[0007] In a first aspect, an embodiment of the present application provides a method for optimizing laser vibration measurement signals, which is applied to a laser Doppler vibrometer for detecting a tunnel lining surface. The method is characterized in that it includes: collecting vibration signals of the tunnel lining surface; where the vibration signals include displacement, velocity, and acceleration; performing spectral leakage suppression processing on the vibration signals based on a preset third-order five-term maximum sidelobe attenuation self-convolution window to obtain suppressed signals; where the window length of the third-order five-term maximum sidelobe attenuation self-convolution window is associated with the signal sampling rate of the vibration signals; performing variational mode decomposition denoising on the vibration signals through a preset improved chicken swarm optimization algorithm to obtain denoised signals; where the improved chicken swarm optimization algorithm includes tent chaotic mapping, hierarchical position update algorithm, and firefly algorithm; inputting the suppressed signals and the denoised signals into a preset Transformer model for feature extraction and attention weight calculation to obtain optimized vibration signals; performing time-frequency domain feature extraction on the optimized vibration signals based on a preset short-time Fourier transform and wavelet packet decomposition algorithm to obtain time-frequency feature parameters; where the time-frequency feature parameters include band energy distribution, nonlinear complexity index, and transient feature parameters; processing the time-frequency feature parameters based on a preset adaptive filtering algorithm to obtain optimized displacement, optimized velocity, and optimized acceleration; and performing real-time correction on the optimized displacement, optimized velocity, and optimized acceleration through a preset compensation calibration algorithm to obtain calibrated vibration signals.

[0008] In an implementation manner of the present application, performing spectral leakage suppression processing on the vibration signals based on a preset third-order five-term maximum sidelobe attenuation self-convolution window to obtain suppressed signals specifically includes: generating a basic window function through three self-convolution operations based on the signal sampling rate and the window length; constructing an optimization objective function based on a preset maximum sidelobe attenuation criterion and integrating the basic window function to generate an optimized window function; where the optimization objective function is to minimize the sidelobe energy ratio; and performing a time-domain convolution operation on the vibration signals based on the optimized window function to obtain the suppressed signals. In an implementation manner of the present application, performing variational mode decomposition denoising on the vibration signals through a preset improved chicken swarm optimization algorithm to obtain denoised signals specifically includes: calling tent chaotic mapping to generate an initial population; iteratively updating the preset chicken swarm optimization algorithm based on the initial population and coupling the firefly algorithm during the iteration process to construct a brightness function based on the average envelope entropy of the variational mode components; using the average envelope entropy of the variational mode components as a fitness function and performing global optimization based on the hierarchical position update algorithm to obtain the denoised signals.

[0009] In an implementation manner of the present application, the suppression signal and the noise reduction signal are input into a preset Transformer model for feature extraction and attention weight calculation to obtain an optimized vibration signal, which specifically includes: performing sliding window slicing on the suppression signal and the noise reduction signal; wherein, the window length of the sliding window slicing is associated with the laser pulse period of the laser Doppler vibrometer; constructing an encoder structure with dynamic layers, using residual connection and layer normalization between layers, and extracting the signal features of the suppression signal and the noise reduction signal based on the encoder structure; processing the signal features based on a preset multi-head attention mechanism to convert the signal features into fusion scores; and verifying the fusion scores based on a preset error threshold to generate an optimized vibration signal.

[0010] In an implementation manner of the present application, time-frequency domain feature extraction is performed on the optimized vibration signal based on a preset short-time Fourier transform and wavelet packet decomposition algorithm to obtain time-frequency feature parameters, which specifically includes: performing frame division processing on the optimized vibration signal to obtain multiple signal frames; wherein, the frame length of the signal frame is associated with the vibration period of the vibration signal; processing the multiple signal frames based on the short-time Fourier transform, and extracting non-linear features through wavelet packet decomposition node energy ratio calculation and entropy value quantization to determine a multi-dimensional feature vector including time-domain statistics, frequency-domain energy distribution, and non-linear features.

[0011] In an implementation manner of the present application, the time-frequency feature parameters are processed based on a preset adaptive filtering algorithm to obtain an optimized displacement, an optimized velocity, and an optimized acceleration, which specifically includes: performing integral operation on the time-frequency feature parameters to obtain vibration displacement parameters; performing differential operation on the time-frequency feature parameters to obtain vibration acceleration parameters; and performing spectral analysis based on the vibration displacement parameters to obtain an optimized displacement, an optimized velocity, and an optimized acceleration.

[0012] In an implementation manner of the present application, the optimized displacement, the optimized velocity, and the optimized acceleration are corrected in real time through a preset compensation and calibration algorithm to obtain a calibrated vibration signal, which specifically includes: calibrating the optimized displacement, the optimized velocity, and the optimized acceleration based on a preset adaptive learning mechanism through historical calibration data to generate a preliminary calibrated vibration signal; and performing real-time calibration on the preliminary calibrated vibration signal based on a preset temperature and light intensity compensation model to generate a calibrated vibration signal.

[0013] In an implementation manner of the present application, a tunnel lining surface health state evaluation model is constructed based on the calibrated vibration signal; wherein, the data of the tunnel lining surface health state evaluation model includes environmental temperature and humidity, tunnel structure type, and historical detection records.

[0014] In a second aspect, the embodiments of the present application further provide a laser vibration measurement signal optimization device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to: collect vibration signals on the tunnel lining surface; wherein the vibration signals include displacement, velocity, and acceleration; perform spectral leakage suppression processing on the vibration signals based on a preset third-order five-term maximum sidelobe attenuation self-convolution window to obtain suppressed signals; wherein the window length of the third-order five-term maximum sidelobe attenuation self-convolution window is associated with the signal sampling rate of the vibration signals; perform variational mode decomposition and noise reduction on the vibration signals through a preset improved chicken swarm optimization algorithm to obtain noise-reduced signals; wherein the improved chicken swarm optimization algorithm includes tent chaotic mapping, hierarchical position update algorithm, and firefly algorithm; input the suppressed signals and the noise-reduced signals into a preset Transformer model for feature extraction and attention weight calculation to obtain optimized vibration signals; perform time-frequency domain feature extraction on the optimized vibration signals based on a preset short-time Fourier transform and wavelet packet decomposition algorithm to obtain time-frequency feature parameters; wherein the time-frequency feature parameters include frequency band energy distribution, non-linear complexity index, and transient feature parameters; process the time-frequency feature parameters based on a preset adaptive filtering algorithm to obtain optimized displacement, optimized velocity, and optimized acceleration; perform real-time correction on the optimized displacement, optimized velocity, and optimized acceleration through a preset compensation calibration algorithm to obtain calibrated vibration signals.

[0015] Thirdly, an embodiment of the present application also provides a non-volatile computer storage medium for optimizing laser vibration measurement signals, storing computer-executable instructions, and the computer-executable instructions are set as follows: collect vibration signals on the tunnel lining surface; wherein, the vibration signals include displacement, velocity, and acceleration; perform spectrum leakage suppression processing on the vibration signals based on a preset third-order five-term maximum sidelobe attenuation self-convolution window to obtain suppressed signals; wherein, the window length of the third-order five-term maximum sidelobe attenuation self-convolution window is associated with the signal sampling rate of the vibration signals; perform variational mode decomposition and noise reduction on the vibration signals through a preset improved chicken swarm optimization algorithm to obtain noise-reduced signals; wherein, the improved chicken swarm optimization algorithm includes tent chaotic mapping, hierarchical position update algorithm, and firefly algorithm; input the suppressed signals and the noise-reduced signals into a preset Transformer model for feature extraction and attention weight calculation to obtain optimized vibration signals; perform time-frequency domain feature extraction on the optimized vibration signals based on a preset short-time Fourier transform and wavelet packet decomposition algorithm to obtain time-frequency feature parameters; wherein, the time-frequency feature parameters include band energy distribution, nonlinear complexity index, and transient feature parameters; process the time-frequency feature parameters based on a preset adaptive filtering algorithm to obtain optimized displacement, optimized velocity, and optimized acceleration; perform real-time correction on the optimized displacement, optimized velocity, and optimized acceleration through a preset compensation calibration algorithm to obtain calibrated vibration signals.

[0016] A method, device, and medium for optimizing laser vibration measurement signals provided by an embodiment of the present application effectively suppress spectrum leakage through adaptive window function design, realize signal noise reduction by combining improved swarm intelligence algorithms, complete feature fusion and enhancement through a deep learning model, and innovatively introduce time-frequency joint analysis to extract multi-dimensional feature parameters. Finally, through an adaptive filtering and dynamic compensation mechanism, the influence of environmental interference on the measurement accuracy of displacement, velocity, and acceleration is eliminated. This solution constructs a complete signal optimization link, enhances the anti-interference ability and feature representation ability of the system, provides a highly robust solution for tunnel structure health monitoring, and has better fidelity and environmental adaptability compared with traditional methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the present application, and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation to the present application. In the drawings: Figure 1 It is a flowchart of a method for optimizing laser vibration measurement signals provided by an embodiment of the present application; Figure 2 It is a schematic internal structure diagram of a device for optimizing laser vibration measurement signals provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments of this application and the corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in this application without creative efforts shall fall within the scope of protection of this application.

[0019] The embodiments of this application provide a method, device, and medium for optimizing laser vibration measurement signals to solve the following technical problem: how to optimize laser vibration measurement signals.

[0020] The technical solutions proposed in the embodiments of this application will be described in detail below with reference to the drawings.

[0021] Figure 1 It is a flowchart of optimizing laser vibration measurement signals provided by the embodiments of this application. As Figure 1 shown, a method for optimizing laser vibration measurement signals provided by the embodiments of this application specifically includes the following steps: Step 1: Collect vibration signals on the tunnel lining surface; among them, the vibration signals include displacement, velocity, and acceleration.

[0022] This application obtains the vibration signals on the tunnel lining surface through a laser Doppler vibrometer. Since the laser Doppler vibrometer is a prior art, it will not be elaborated here.

[0023] Step 2: Perform spectrum leakage suppression processing on the vibration signals based on a preset third-order five-term maximum sidelobe attenuation self-convolution window to obtain suppressed signals; among them, the window length of the third-order five-term maximum sidelobe attenuation self-convolution window is associated with the signal sampling rate of the vibration signals.

[0024] Process the vibration signals on the tunnel lining surface through the third-order five-term maximum sidelobe attenuation self-convolution window function. By optimizing the characteristics of the window function, effectively suppress the spectrum leakage phenomenon caused by signal truncation, thereby improving the accuracy of spectrum analysis. Step 21: Generate a basic window function through three self-convolution operations based on the signal sampling rate and the window length.

[0025] The signal sampling rate refers to the number of vibration signal samples collected per second, with the unit of Hz.

[0026] The window length is the number of sampling points covered by the window function and has a corresponding relationship with the sampling rate. When the sampling rate is 2500 Hz, if it is necessary to analyze the vibration characteristics of a 100 ms duration, the window length should be set to 250 points.

[0027] Perform three convolution operations on the base window function with itself. The self-convolution process can be understood as time-domain expansion and characteristic optimization of the window function. After each convolution, the window length becomes 2 times the original length minus 1. After three convolutions, the total length is 8N 0 - 3, where N 0 is the initial window length.

[0028] For example, for a vibration signal with a sampling rate of 2000 Hz, set the initial window length to 200 points (corresponding to a duration of 100 ms). After the first self-convolution, the window length becomes 399 points, the second becomes 797 points, and the third becomes 1593 points. Finally, the base window function has a length of 1593 points, covering vibration data with a duration of approximately 800 ms.

[0029] Step 22: Construct an optimization objective function based on a preset maximum sidelobe attenuation criterion and integrate the base window function to generate an optimized window function; among them, the optimization objective function is to minimize the sidelobe energy ratio.

[0030] The maximum sidelobe attenuation criterion requires minimizing the ratio of the sidelobe energy of the window function to the main lobe.

[0031] Using the weighted least squares method, taking the sidelobe energy ratio as a penalty term and the main lobe width as a constraint condition, establish a multi-objective optimization model.

[0032] Take the base window function generated by three self-convolutions as the initial solution, and adjust the window function coefficients through an iterative algorithm to make the optimization objective function reach a preset threshold.

[0033] For example, through iterative calculations using the MATLAB optimization toolbox, after 200 iterations, the sidelobe peak level of the optimized window function is -74.8 dB, and the sidelobe attenuation rate is 30 dB / oct.

[0034] Step 23: Perform a time-domain convolution operation on the vibration signal based on the optimized window function to obtain a suppression signal.

[0035] In a specific example, the frequency-domain expression of the five-term maximum sidelobe attenuation window is: (1) In the formula, is the window length, .

[0036] It should be noted that the parameters (such as "0.2734375") in the frequency-domain expression of the five-term maximum sidelobe attenuation window are parameters determined according to actual work.

[0037] Step 3: Perform variational mode decomposition denoising on the vibration signal through a preset improved chicken swarm optimization algorithm to obtain a denoised signal; among them, the improved chicken swarm optimization algorithm includes tent chaotic mapping, hierarchical position update algorithm, and firefly algorithm.

[0038] The improved chicken swarm optimization algorithm aims at the non-stationary components such as mechanical noise and environmental interference mixed in the vibration signal of the tunnel lining surface. It decomposes them into multiple sub-modalities with sparse characteristics through variational mode decomposition (VMD), and then adaptively determines the decomposition parameters in combination with the improved chicken swarm optimization algorithm (ICSO), finally realizing signal denoising.

[0039] Step 31: Call tent chaotic mapping to generate an initial population.

[0040] Tent chaotic mapping is an algorithm with uniform distribution characteristics and strong ergodicity; In a specific example, N initial values are randomly generated in the interval [-1, 1] (N is the population size, and 100 is taken in this embodiment); Generate a chaotic sequence through an iterative formula, and linearly map the sequence to the parameter search spaces [αmin, αmax] and [τmin, τmax] (α is the penalty factor of VMD, and τ is the noise tolerance); Generate an initial population containing two parameters, α and τ.

[0041] For example, set the search range of α to [100, 5000], and the search range of τ to [0.1, 5]. After tent mapping, 100 groups of initial parameter combinations are obtained.

[0042] Step 32: Iteratively update the preset chicken swarm optimization algorithm based on the initial population, and couple the firefly algorithm during the iteration to construct a brightness function based on the average envelope entropy of the variational mode components.

[0043] Sort the population according to the fitness value. The top 15% are roosters, the middle 35% are hens, and the bottom 50% are chicks. Roosters perform global search: the position update formula introduces a normal distribution random number to enhance the exploration ability; Hens perform follow-up search: the position update is affected by the positions of the roosters in the group, and at the same time, the attractiveness perturbation of the firefly algorithm is introduced; Chicks perform local search: move in a small range around the hens, and the position update formula introduces the weighted influence of the hens' positions.

[0044] Firefly algorithm coupling: Based on the average envelope entropy of each modal component after VMD decomposition, the smaller the entropy value, the purer the signal. Calculate the attraction between individuals according to the brightness difference, and guide the chicken swarm to move towards the high-brightness area.

[0045] For example, set the light absorption coefficient β of the firefly to 0.2 and the step factor γ to 0.01. After each iteration of the chicken flock, calculate the brightness value of each individual in the current population, and apply the firefly attraction operation to the individuals with brightness values lower than the average level.

[0046] Step 33: Use the average envelope entropy of the variational mode components as the fitness function, and perform global optimization based on the hierarchical position update algorithm to obtain the denoised signal.

[0047] Perform VMD decomposition on each parameter combination to obtain K modal components (in this embodiment, K = 5) Calculate the envelope entropy of each component, and take the average value as the fitness value; Update of roosters: Adopt an adaptive inertia weight strategy to balance global search and local development; Update of hens: Introduce a dynamic topological structure and adjust the following object according to the number of iterations; Update of chicks: Adopt a non-linearly decreasing learning factor to improve the search accuracy; Termination conditions: Reach the maximum number of iterations (set to 200 times in this embodiment).

[0048] The improvement amount of the fitness value is less than the preset threshold for 10 consecutive generations (set to 0.01 in this embodiment); For example, after 100 iterations, the algorithm converges to the optimal parameter combination α = 2345, τ = 2.1.

[0049] In this step, the chicken swarm optimization algorithm simulates the hierarchical system of the swarm and the group activity behavior of the chicken flock. Under a special hierarchical system, there is competition among different chicken breeds in the chicken flock when searching for food. Roosters have strong food-searching abilities and small fitness values; hens are next; chicks have the weakest food-searching abilities and the largest fitness values. The chicken flock is grouped according to the number of roosters, and each group consists of one rooster, some hens, and chicks. There are as many groups as there are roosters. In the grouping, roosters have the strongest search ability and are in a dominant position with the smallest fitness value; hens with slightly worse search abilities follow closely around the roosters to search for food, with slightly larger fitness values; some of the hens also lead the chicks. Chicks have the worst search ability and only search for food around the hens, with the largest fitness value, realizing the local search function.

[0050] Under the hierarchical system, the dominance relationship among roosters and the mother - chick relationship between hens and chicks in the group will change. This hierarchical order is established through fitness values. The relationships between roosters and hens are randomly grouped, and the mother - chick relationship between hens and chicks is randomly established. In the chicken flock, the individual with a smaller fitness value has more advantages, can obtain food preferentially, and lead the individuals with larger fitness values. The individual with the smallest fitness value corresponds to the rooster in the chicken flock, the slightly larger one corresponds to the hen, and the largest one corresponds to the chick. Under this hierarchical order, they cooperate in groups, update their positions according to their respective movement laws, conduct searches, and finally search for the best foraging position, that is, obtain the optimal solution.

[0051] Roosters with good fitness can search for food in a larger range and can obtain food preferentially compared to roosters with poor fitness to achieve global search. The position update of roosters is affected by the positions of other randomly selected roosters, and the update strategy is as follows: (2) (3) Where: Randn is a random number obeying the uniform distribution on [0, 1]; the value of the j - th dimension of the position of the i - th rooster is expressed as ; S represents the current iteration number, which is a random number obeying the normal distribution with an expected value of 0 and a variance value of 2; the fitness of the i - th rooster is ; the fitness of the randomly selected rooster S is ; is the dispersion degree of the chicken flock; is the rooster update strategy. Hens follow their partner roosters to search for food, and their position updates are affected by the positions of their partner roosters. Due to the food - stealing behavior of hens, the position updates are also related to other roosters and hens. The update strategy is as follows: (4) (5) (6) Where: is a random number obeying uniform distribution; is the current iteration number; the fitness value of the partner rooster of this hen is ; the fitness value of the partner rooster of the hen is ; is the influence factor corresponding to other chickens; is the minimum constant to prevent the denominator from being 0.

[0052] The chick searches for food around its mother hen. Its search ability is the worst, and its position is affected by the mother rooster. The update strategy is as follows: (7) In the formula, the mother hen The dimensional value of the position is ; The influence factor of the mother hen's position on the chick's position is , is randomly generated by a random function, and its value range is generally (0, 2).

[0053] The population is initialized using the tent chaotic map. A chaotic sequence is generated within the interval. Combining the upper and lower limits of the search space, the initialization of the population is completed. The initialized population is shown by the following formula to replace formulas (2)-(3): (8) (9) Where represents a random number between the intervals ; is the state variable before the tent chaotic map; is the state variable after the tent chaotic map; is the upper limit of the search space; is the lower limit of the search space.

[0054] To increase the local search ability, the firefly algorithm is used to reflect the quality of the function value through the brightness of individuals. This algorithm shows good performance in local search and improves the movement efficiency and significantly improves the local optimization ability by disturbing the positions of the chicken flock.

[0055] The brightness function of the firefly is: (10) In the formula is the initial fluorescence intensity, which changes according to the objective function; is the light intensity absorption coefficient, which decreases as the distance increases; is the spatial distance of the firefly from to .

[0056] The attractiveness of the firefly is: (11) Where the firefly is attracted to move towards the firefly , $\alpha$ is the initial attractiveness of the firefly, and the position update replaces formulas (4)-(6): (12) where and are the positions of the fireflies in the current population; is the attractiveness of the firefly; is the step factor, generally a value in ; is a random number uniformly distributed in

[0057] Step 4: Input the suppression signal and the noise reduction signal into a preset Transformer model for feature extraction and attention weight calculation to obtain an optimized vibration signal.

[0058] The vibration signal of the tunnel lining surface is processed by the Transformer model in the deep neural network architecture. By fusing the multi-dimensional features of the noise reduction signal and the suppression signal (from the pre-interference suppression module), the intelligent enhancement of the signal components and the elimination of noise residues are realized through the self-attention mechanism.

[0059] Step 41: Perform sliding window slicing on the suppression signal and the noise reduction signal; among them, the window length of the sliding window slicing is associated with the laser pulse period of the laser Doppler vibrometer.

[0060] The technology of dividing the continuous vibration signal into equal-length data segments simulates the short-time Fourier analysis characteristics of the human auditory system. The window length needs to match the pulse period of the laser Doppler vibrometer system (in this embodiment, a 20 ms pulse interval is adopted) to ensure the signal phase integrity to a certain extent.

[0061] For example, set the window length to 256 sampling points (corresponding to 25.6 ms), and the sliding step size to 128 points (50% overlap). For a vibration signal with a sampling rate of 10 kHz, 12,000 training samples are generated, and each sample contains synchronous slices of the suppression signal and the noise reduction signal.

[0062] Step 42: Construct an encoder structure with dynamic layers, use residual connections and layer normalization between layers, and extract the signal features of the suppression signal and the noise reduction signal based on the encoder structure.

[0063] The dynamic layer encoder contains a neural network module with an adaptive depth adjustment mechanism, which dynamically switches between 3-8 layers according to the input signal complexity (monitored in real time through feature entropy) to balance the model capacity and computational efficiency.

[0064] The key technology to solve the problem of vanishing gradients in deep networks is to achieve identity mapping through skip connections. Normalize the input of each layer of neurons to accelerate the training convergence.

[0065] For example, the basic encoder structure contains 6 layers, with a dimension of 512 for each layer. When the signal feature entropy exceeds a preset threshold (set to 3.5 nats in this embodiment), it automatically increases to 8 layers. Use the GELU activation function to enhance the non-linear expression ability, and adopt a learning rate warm-up strategy during training.

[0066] Step 43: Process the signal features based on a preset multi-head attention mechanism to convert the signal features into fusion scores.

[0067] Compute 8 groups of independent attention weights in parallel, and each group of weights captures the associated features of different dimensions of the signal (such as time-domain waveform similarity, frequency-domain energy distribution correlation, etc.).

[0068] Concatenate the outputs of the multi-head attention and pass them through a fully connected layer to generate a fusion score in the range of [0, 1], which represents the effective information retention degree of the signal segment.

[0069] For example, set 8 attention heads, with a dimension of 64 for each head. Introduce learnable position encoding to enhance the sequence modeling ability, and the dimension of the position encoding is the same as that of the signal features. When calculating the fusion score, adopt a weighted summation strategy, and the weights are generated through the softmax function.

[0070] Step 44: Verify the fusion score based on a preset error threshold to generate an optimized vibration signal.

[0071] Error threshold verification: Set a fusion score threshold (take 0.85 in this embodiment), and the signal segments with scores higher than the threshold are determined as valid components, while those lower than the threshold are processed twice.

[0072] Generation of the optimized signal: For valid segments: Keep the original signal and apply time-frequency domain gain; For invalid segments: Use an autoregressive moving average model (ARMA) for interpolation and reconstruction; Final signal: Synthesize the complete waveform through the overlap-and-add method (OLA).

[0073] For example, in the measured signal, 80% of the valid signal segments are retained after threshold screening. The ARMA(2,1) model is used for interpolation of the invalid segments, and the interpolation coefficients are determined by the first 3 sampling points of the adjacent valid segments.

[0074] In a specific example, during the monitoring of a long tunnel A, a laser Doppler vibrometer (sampling rate 20 kHz) is deployed to collect vibration signals. After pre-processing for noise reduction, the signals are input into this Transformer model. The sliding window generates 24,000 training samples (window length 512 points). The dynamic encoder adaptively adjusts between 4 and 7 layers according to the signal complexity. The multi-head attention mechanism learns the spatio-temporal correlation features of the signal, and 88% of the effective signal segments are retained after screening by the error threshold.

[0075] Step 5: Extract time-frequency domain features from the optimized vibration signal based on the preset short-time Fourier transform and wavelet packet decomposition algorithms to obtain time-frequency feature parameters; among them, the time-frequency feature parameters include the frequency band energy distribution, the non-linear complexity index, and the transient feature parameters.

[0076] Through the joint framework of the short-time Fourier transform (STFT) and wavelet packet decomposition (WPD), feature extraction is performed on the optimized tunnel lining vibration signal. By combining the time-frequency localization advantage of STFT and the multi-scale analysis ability of wavelet packets, a refined characterization of the signal features is achieved.

[0077] Step 51: Perform frame segmentation on the optimized vibration signal to obtain multiple signal frames; among them, the frame length of the signal frame is correlated with the vibration period of the vibration signal.

[0078] The continuous vibration signal is segmented into equal-length short-time analysis units to simulate the segmentation processing mechanism of the human auditory system. The frame length needs to match the natural vibration period of the tunnel lining structure to ensure the quasi-stationarity of the signal within the frame.

[0079] For example, for the optimized vibration signal with a sampling rate of 20 kHz, the frame length is set to 512 sampling points (corresponding to 25.6 ms), and the frame shift is 256 points (50% overlap). For the measured 10-second-long signal, 390 analysis frames are generated to cover the complete vibration process before and after the train passes to a certain extent.

[0080] Step 52: Process multiple signal frames based on the short-time Fourier transform, and extract non-linear features by calculating the proportion of node energy in wavelet packet decomposition and quantifying the entropy value to determine a multi-dimensional feature vector including time-domain statistics, frequency-domain energy distribution, and non-linear features.

[0081] Use a sliding window function (in this embodiment, the Hamming window is selected) to perform Fourier transform on each frame of the signal to obtain a time-frequency spectrogram. Set the window length to be the same as the frame length.

[0082] Perform 3-layer wavelet packet decomposition on each frame of the signal. Calculate the proportion of the energy of each decomposition node in the total energy of the entire frame, and the Shannon entropy value of the node coefficients to quantify the non-linear complexity of the signal.

[0083] Integrate the following three types of features: Time-domain statistics: including the mean, variance, and peak factor of the signal within the frame; Frequency-domain energy distribution: Divide into three frequency bands of 0 - 200 Hz, 200 - 500 Hz, and 500 - 1000 Hz, and calculate the energy proportion of each band; Nonlinear features: The maximum value, minimum value of the energy ratio of wavelet packet nodes, and the quantization result of entropy value; For example, STFT uses a 512-point FFT to generate 257 frequency points (0 - 10 kHz). The third layer of wavelet packet decomposition generates 8 nodes. Calculate the energy proportion and entropy value of each node. The dimension of the feature vector is: 3 (time domain) + 3 (frequency domain) + 3 (nonlinear) = 9 dimensions.

[0084] Step 6. Process the time-frequency feature parameters based on a preset adaptive filtering algorithm to obtain the optimized displacement, optimized velocity, and optimized acceleration.

[0085] The adaptive filtering technology in the field of signal processing combines time-domain integration / differentiation operations with frequency-domain analysis to accurately extract structural motion parameters from the optimized vibration signal. By dynamically adjusting the filtering parameters, the influence of noise residue on the estimation of motion parameters is eliminated, and the reconstruction of displacement, velocity, and acceleration is realized.

[0086] Step 61. Perform an integration operation on the time-frequency feature parameters to obtain the vibration displacement parameters.

[0087] Convert the acceleration signal to a displacement signal through two numerical integrations. The first integration obtains the velocity signal, and the second integration obtains the displacement signal.

[0088] For example, for the optimized vibration signal with a sampling rate of 20 kHz, set the integration step size to 100 sampling points (corresponding to 5 ms). Use the adaptive integration algorithm to dynamically adjust the integration window length according to the signal energy (the basic length is 400 points, and the maximum extends to 1000 points) to effectively suppress the integration drift to a certain extent.

[0089] Step 62. Perform a differentiation operation on the time-frequency feature parameters to obtain the vibration acceleration parameters.

[0090] Convert the displacement signal to an acceleration signal through numerical differentiation. Use the seven-point difference method to balance the calculation accuracy and the noise amplification effect. The difference coefficients are optimized to match the signal bandwidth. Introduce median filtering preprocessing (window length 9 points) to suppress the high-frequency noise amplification in the differentiation operation.

[0091] For example, perform a second-order differentiation on the displacement signal, and set the difference coefficients to [0.0019, -0.0288, 0.1464, -0.5859, 0.5859, -0.1464, 0.0288, -0.0019], corresponding to the central difference method.

[0092] Step 63: Perform spectral analysis based on the vibration displacement parameters to obtain the optimized displacement, optimized velocity, and optimized acceleration.

[0093] Perform short-time Fourier transform (STFT) on the integrated displacement signal to analyze the time-frequency domain components. Extract the effective motion components through adaptive band-pass filtering (the center frequency tracks the main frequency of the signal), and reconstruct the optimized displacement, velocity, and acceleration.

[0094] For example, for displacement spectral analysis: Use 2048-point STFT, with a frequency resolution of 9.77 Hz, and select the Hamming window as the window function; Adaptive filtering: A band-pass filter based on the main frequency of the signal (bandwidth ±3 times the main frequency), dynamically adjust the center frequency (tracking range 10 - 1000 Hz); Motion parameter reconstruction: Perform inverse STFT on the filtered spectrum to obtain the optimized displacement; obtain the optimized velocity and acceleration through first-order and second-order differentiation.

[0095] Step 7: Real-time correct the optimized displacement, optimized velocity, and optimized acceleration through a preset compensation calibration algorithm to obtain the calibrated vibration signal.

[0096] Step 71: Based on a preset adaptive learning mechanism, calibrate the optimized displacement, optimized velocity, and optimized acceleration through historical calibration data to generate a preliminary calibrated vibration signal.

[0097] Use the recursive least squares (RLS) method to construct an adaptive filter, dynamically adjust the filter parameters through historical calibration data, and achieve intelligent calibration of motion parameters.

[0098] Establish a calibration data set including multiple environmental parameters (temperature, humidity, installation location, etc.). The adaptive filter matches the best calibration mode from historical data according to the real-time environmental parameters and performs preliminary calibration on the optimized motion parameters.

[0099] For example, at the tunnel monitoring site, collect calibration data in different seasons (35°C / 65%RH in summer, 5°C / 40%RH in winter), and construct a calibration database containing 200 groups of samples. The adaptive filter automatically selects the closest calibration mode according to the real-time temperature and humidity readings and performs preliminary calibration on the optimized displacement, velocity, and acceleration.

[0100] Step 72: Perform real-time calibration on the preliminary calibrated vibration signal based on a preset temperature and light intensity compensation model to generate the calibrated vibration signal.

[0101] Establish a multiple regression model of temperature, light intensity, and signal drift amount to quantify the influence of environmental factors on the vibration signal and achieve refined environmental compensation.

[0102] Under laboratory conditions, the temperature (10°C - 40°C) and light intensity (100 - 1000 lux) were controlled to change, and signal drift data was collected. After constructing a compensation model, it was deployed at the tunnel monitoring site to correct the signal deviation caused by environmental factors in real time.

[0103] For example, a three-dimensional compensation model of temperature - light intensity - displacement drift was established through experiments. During tunnel site monitoring, when the temperature rises from 20°C to 30°C and the light intensity increases from 300 lux to 800 lux, the compensation amounts are automatically calculated (displacement compensation +0.012 mm, velocity compensation +0.008 m / s, acceleration compensation +0.05 m / s²).

[0104] This application also includes the following method: constructing a health state assessment model for the tunnel lining surface based on calibrated vibration signals; wherein, the data of the health state assessment model for the tunnel lining surface includes environmental temperature and humidity, tunnel structure type, and historical detection records.

[0105] Using multi-source data fusion and machine learning techniques, the calibrated vibration signals are combined with tunnel environmental parameters, structural feature information, and historical detection data to construct a health state assessment model with environmental adaptability and structural specificity. This model realizes the intelligent assessment of the tunnel lining surface state through quantitative analysis.

[0106] The above is the method embodiment proposed in this application. Based on the same inventive concept, the embodiment of this application also provides a laser vibration signal optimization device, the structure of which is as Figure 2 shown.

[0107] Figure 2 This is a schematic diagram of the internal structure of a laser vibration signal optimization device provided by the embodiment of this application. As Figure 2 shown, the device includes: At least one processor 201; And, a memory 202 communicatively connected to at least one processor; Wherein, the memory 202 stores instructions executable by at least one processor, and the instructions are executed by at least one processor 201 so that at least one processor 201 can: Collect the vibration signals on the tunnel lining surface; among them, the vibration signals include displacement, velocity, and acceleration; perform spectrum leakage suppression processing on the vibration signals based on a preset third-order five-term maximum sidelobe attenuation self-convolution window to obtain suppressed signals; among them, the window length of the third-order five-term maximum sidelobe attenuation self-convolution window is associated with the signal sampling rate of the vibration signals; perform variational mode decomposition noise reduction on the vibration signals through a preset improved chicken swarm optimization algorithm to obtain noise-reduced signals; among them, the improved chicken swarm optimization algorithm includes tent chaotic mapping, hierarchical position update algorithm, and firefly algorithm; input the suppressed signals and the noise-reduced signals into a preset Transformer model for feature extraction and attention weight calculation to obtain optimized vibration signals; perform time-frequency domain feature extraction on the optimized vibration signals based on a preset short-time Fourier transform and wavelet packet decomposition algorithm to obtain time-frequency feature parameters; among them, the time-frequency feature parameters include band energy distribution, nonlinear complexity index, and transient feature parameters; process the time-frequency feature parameters based on a preset adaptive filtering algorithm to obtain optimized displacement, optimized velocity, and optimized acceleration; perform real-time correction on the optimized displacement, optimized velocity, and optimized acceleration through a preset compensation calibration algorithm to obtain calibrated vibration signals.

[0108] Some embodiments of the present application provide a corresponding Figure 1 non-volatile computer storage medium for optimizing laser vibration signals, storing computer-executable instructions, and the computer-executable instructions are set as follows: Collect the vibration signals on the tunnel lining surface; among them, the vibration signals include displacement, velocity, and acceleration; perform spectrum leakage suppression processing on the vibration signals based on a preset third-order five-term maximum sidelobe attenuation self-convolution window to obtain suppressed signals; among them, the window length of the third-order five-term maximum sidelobe attenuation self-convolution window is associated with the signal sampling rate of the vibration signals; perform variational mode decomposition noise reduction on the vibration signals through a preset improved chicken swarm optimization algorithm to obtain noise-reduced signals; among them, the improved chicken swarm optimization algorithm includes tent chaotic mapping, hierarchical position update algorithm, and firefly algorithm; input the suppressed signals and the noise-reduced signals into a preset Transformer model for feature extraction and attention weight calculation to obtain optimized vibration signals; perform time-frequency domain feature extraction on the optimized vibration signals based on a preset short-time Fourier transform and wavelet packet decomposition algorithm to obtain time-frequency feature parameters; among them, the time-frequency feature parameters include band energy distribution, nonlinear complexity index, and transient feature parameters; process the time-frequency feature parameters based on a preset adaptive filtering algorithm to obtain optimized displacement, optimized velocity, and optimized acceleration; perform real-time correction on the optimized displacement, optimized velocity, and optimized acceleration through a preset compensation calibration algorithm to obtain calibrated vibration signals.

[0109] Each embodiment in the present application is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other. Each embodiment focuses on the differences from other embodiments. In particular, for the embodiments of the Internet of Things devices and media, since they are basically similar to the method embodiments, the description is relatively simple, and reference can be made to the relevant parts of the method embodiments for the relevant content.

[0110] The systems and media provided by the embodiments of the present application correspond one-to-one with the methods. Therefore, the systems and media also have beneficial technical effects similar to those of their corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the systems and media will not be elaborated here.

[0111] Those skilled in the art should understand that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can be in the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can be in the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memories, CD-ROMs, optical memories, etc.) containing computer-usable program code.

[0112] The present application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of the processes 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 a device for realizing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0113] 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 an instruction device, and the instruction device realizes the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0114] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus, causing a series of operational steps to be performed on the computer or other programmable apparatus to generate a computer implemented process, so that the instructions executed on the computer or other programmable apparatus provide steps for implementing the functions specified in one process Figure 1 one process or a plurality of processes and / or boxes Figure 1 steps for implementing the functions specified in one box or a plurality of boxes.

[0115] In a typical configuration, a computing device includes one or more processors (CPUs), an input / output interface, a network interface, and memory.

[0116] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory such as read only memory (ROM) or flash memory (flash RAM). The memory is an example of a computer-readable medium.

[0117] Computer-readable media includes both permanent and non-permanent, removable and non-removable media implemented by any method or technology for storing information. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read only memory (ROM), electrically erasable programmable read only memory (EEPROM), flash memory or other memory technologies, compact disc read only memory (CD-ROM), digital versatile discs (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory media such as modulated data signals and carrier waves.

[0118] It should also be noted that the term "comprising", "including" or any other variation thereof is intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, commodity or device. Without further limitation, an element defined by the statement "including one..." does not exclude the presence of additional identical elements in the process, method, commodity or device including the element.

[0119] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, various changes and modifications can be made to the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included within the scope of the claims of the present application.

Claims

1. A laser vibrometer signal optimization method, applied to a laser Doppler vibrometer device, wherein the laser Doppler vibrometer device is used to detect a tunnel lining surface, characterized in that: The method comprises: Collecting vibration signals of the tunnel lining surface; wherein the vibration signals include displacement, velocity and acceleration; The vibration signal is subjected to spectrum leakage suppression processing based on a preset third-order five-term maximum sidelobe attenuation self-convolution window to obtain a suppressed signal; wherein the window length of the third-order five-term maximum sidelobe attenuation self-convolution window is associated with a signal sampling rate of the vibration signal; The vibration signal is subjected to variational mode decomposition and noise reduction by using a preset improved flock optimization algorithm to obtain a noise reduction signal; wherein the improved flock optimization algorithm includes tent chaos mapping, hierarchical position update algorithm and firefly algorithm; Inputting the suppression signal and the noise reduction signal into a preset Transformer model for feature extraction and attention weight calculation to obtain an optimized vibration signal; Based on the preset short-time Fourier transform and wavelet packet decomposition algorithm, the optimized vibration signal is subjected to time-frequency domain feature extraction to obtain time-frequency feature parameters; wherein the time-frequency feature parameters include frequency band energy distribution, nonlinear complexity index and transient feature parameters; Processing the time-frequency characteristic parameters based on a preset adaptive filtering algorithm to obtain an optimized displacement, an optimized velocity, and an optimized acceleration; The optimized displacement, optimized velocity and optimized acceleration are corrected in real time by a preset compensation calibration algorithm to obtain a calibrated vibration signal.

2. A laser vibrometer signal optimization method according to claim 1, characterized in that: The vibration signal is subjected to spectrum leakage suppression processing based on a preset third-order five-term maximum sidelobe attenuation self-convolution window to obtain a suppressed signal, specifically comprising: Based on the signal sampling rate and the window length, a basic window function is generated by three self-convolution operations; An optimization objective function is constructed based on a preset maximum sidelobe attenuation criterion, and the basic window function is integrated to generate an optimization window function; wherein the optimization objective function is to minimize the sidelobe energy ratio; A time-domain convolution operation is performed on the vibration signal based on the optimized window function to obtain a suppression signal.

3. The laser vibrometer signal optimization method according to claim 1, characterized in that: The vibration signal is subjected to variational mode decomposition and noise reduction by using a preset improved flock optimization algorithm to obtain a noise reduction signal, specifically including: Calling the tent chaotic map to generate an initial population; Iteratively updating a preset swarm optimization algorithm based on the initial population, and coupling the firefly algorithm in the iterative process to construct a brightness function based on the average envelope entropy of the variational modal component; The average envelope entropy of the variational modal component is used as a fitness function, and global optimization is performed based on the hierarchical position update algorithm to obtain the noise reduction signal.

4. The laser vibrometer signal optimization method according to claim 1, characterized in that: The suppression signal and the noise reduction signal are input into a preset Transformer model for feature extraction and attention weight calculation to obtain an optimized vibration signal, specifically including: Sliding window slicing is performed on the suppression signal and the noise reduction signal; wherein the window length of the sliding window slicing is associated with the laser pulse period of the laser Doppler vibrometer; Constructing an encoder structure with a dynamic number of layers, using residual connections and layer normalization between layers, and extracting signal features of the suppression signal and the noise reduction signal based on the encoder structure; Processing the signal features based on a preset multi-head attention mechanism to convert the signal features into a fusion score; The fusion score is validated based on a preset error threshold to generate an optimized vibration signal.

5. The laser vibrometer signal optimization method according to claim 1, characterized in that: Based on the preset short-time Fourier transform and wavelet packet decomposition algorithm, the optimized vibration signal is subjected to time-frequency domain feature extraction to obtain time-frequency feature parameters, specifically including: Performing frame processing on the optimized vibration signal to obtain a plurality of signal frames; wherein the frame length of the signal frame is associated with the vibration period of the vibration signal; Based on the short-time Fourier transform, a plurality of the signal frames are processed, and nonlinear features are extracted by calculating the energy proportion of the wavelet packet decomposition nodes and quantifying the entropy value, so as to determine a multi-dimensional feature vector including time domain statistics, frequency domain energy distribution and nonlinear features.

6. The laser vibrometer signal optimization method according to claim 1, characterized in that: The time-frequency characteristic parameters are processed based on a preset adaptive filtering algorithm to obtain optimized displacement, optimized speed and optimized acceleration, specifically including: Performing an integration operation on the time-frequency characteristic parameters to obtain vibration displacement parameters; Performing differential operation on the time-frequency characteristic parameters to obtain vibration acceleration parameters; Spectrum analysis is performed based on vibration displacement parameters to obtain optimized displacement, optimized velocity, and optimized acceleration.

7. The laser vibrometer signal optimization method according to claim 1, characterized in that: The optimized displacement, optimized velocity, and optimized acceleration are corrected in real time by a preset compensation calibration algorithm to obtain a calibration vibration signal, specifically including: Based on a preset adaptive learning mechanism, the optimized displacement, the optimized velocity, and the optimized acceleration are calibrated by historical calibration data to generate a preliminary calibration vibration signal; The preliminary calibration vibration signal is calibrated in real time based on a preset temperature and light intensity compensation model to generate a calibration vibration signal.

8. The laser vibrometer signal optimization method according to claim 1, characterized in that: The method further comprises: A tunnel lining surface health status assessment model is constructed based on the calibrated vibration signal; wherein the data of the tunnel lining surface health status assessment model includes environmental temperature and humidity, tunnel structure type and historical detection records.

9. A laser vibration measurement signal optimization device, characterized in that: The device comprises: at least one processor; and, a memory communicatively coupled to the at least one processor; The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to: Collecting vibration signals of the tunnel lining surface; wherein the vibration signals include displacement, velocity and acceleration; The vibration signal is subjected to spectrum leakage suppression processing based on a preset third-order five-term maximum sidelobe attenuation self-convolution window to obtain a suppressed signal; wherein the window length of the third-order five-term maximum sidelobe attenuation self-convolution window is associated with a signal sampling rate of the vibration signal; The vibration signal is subjected to variational mode decomposition and noise reduction by using a preset improved flock optimization algorithm to obtain a noise reduction signal; wherein the improved flock optimization algorithm includes tent chaos mapping, hierarchical position update algorithm and firefly algorithm; Inputting the suppression signal and the noise reduction signal into a preset Transformer model for feature extraction and attention weight calculation to obtain an optimized vibration signal; Based on the preset short-time Fourier transform and wavelet packet decomposition algorithm, the optimized vibration signal is subjected to time-frequency domain feature extraction to obtain time-frequency feature parameters; wherein the time-frequency feature parameters include frequency band energy distribution, nonlinear complexity index and transient feature parameters; Processing the time-frequency characteristic parameters based on a preset adaptive filtering algorithm to obtain an optimized displacement, an optimized velocity, and an optimized acceleration; The optimized displacement, optimized velocity and optimized acceleration are corrected in real time by a preset compensation calibration algorithm to obtain a calibrated vibration signal.

10. A non-volatile computer storage medium for optimizing laser vibrometer signals, storing computer executable instructions, characterized in that: The computer executable instructions are configured to: Collecting vibration signals of the tunnel lining surface; wherein the vibration signals include displacement, velocity and acceleration; The vibration signal is subjected to spectrum leakage suppression processing based on a preset third-order five-term maximum sidelobe attenuation self-convolution window to obtain a suppressed signal; wherein the window length of the third-order five-term maximum sidelobe attenuation self-convolution window is associated with a signal sampling rate of the vibration signal; The vibration signal is subjected to variational mode decomposition and noise reduction by using a preset improved flock optimization algorithm to obtain a noise reduction signal; wherein the improved flock optimization algorithm includes tent chaos mapping, hierarchical position update algorithm and firefly algorithm; Inputting the suppression signal and the noise reduction signal into a preset Transformer model for feature extraction and attention weight calculation to obtain an optimized vibration signal; Based on the preset short-time Fourier transform and wavelet packet decomposition algorithm, the optimized vibration signal is subjected to time-frequency domain feature extraction to obtain time-frequency feature parameters; wherein the time-frequency feature parameters include frequency band energy distribution, nonlinear complexity index and transient feature parameters; Processing the time-frequency characteristic parameters based on a preset adaptive filtering algorithm to obtain an optimized displacement, an optimized velocity, and an optimized acceleration; The optimized displacement, optimized velocity and optimized acceleration are corrected in real time by a preset compensation calibration algorithm to obtain a calibrated vibration signal.

Images