Tunnel lining surface vibration measurement signal noise reduction method

Abstract

The invention discloses a tunnel lining surface vibration measurement signal noise reduction method, which comprises the steps of performing sequence updating through tent mapping according to a collected vibration measurement signal to obtain a nonlinear enhanced signal; according to the non-linear enhanced signals, extracting signal features through an improved bald eagle algorithm; and according to the signal characteristics, judging the stability and low frequency property of the signal, if the signal is stable and low in frequency, carrying out noise reduction by improving a wavelet threshold, otherwise, carrying out noise reduction by improving a wavelet packet threshold. Through nonlinear enhancement of Tent mapping, dynamic feature extraction and adaptive noise reduction strategy switching of an improved bald eagle algorithm, the defects of the prior art in the aspects of noise separation capability, noise reduction strategy flexibility and calculation efficiency are overcome, efficient noise reduction and feature retention of vibration measurement signals are achieved, and the method is suitable for large-scale popularization and application. And a more reliable technical support is provided for health monitoring of the tunnel structure.

Description

Technical Field

[0001] This invention belongs to the field of signal noise reduction technology, specifically relating to a method for noise reduction of vibration measurement signals from tunnel lining surfaces. Background Technology

[0002] In tunnel engineering, vibration signal monitoring of the lining structure is a key means of assessing the structural health and safety. However, the actual vibration signals collected are usually subject to various noise interferences (such as environmental noise, equipment vibration noise, etc.), and the signals themselves have nonlinear, non-stationary, and multi-scale characteristics, making it difficult for traditional noise reduction methods to balance signal fidelity and noise suppression effectiveness.

[0003] In the process of extracting weak vibration signals, feature extraction is performed on the signal. The Bald Eagle Search algorithm (BES) is usually used for feature extraction. This algorithm has the advantages of good convergence, simple parameter setting, and strong robustness. However, it still has the disadvantages of uneven initial position distribution and easy getting trapped in local optima, which leads to inaccurate signal feature extraction.

[0004] Furthermore, current mainstream vibration signal denoising methods mainly include wavelet thresholding and wavelet packet thresholding. Wavelet thresholding decomposes the signal into sub-bands of different scales using wavelet transform, applying hard or soft thresholding to high-frequency sub-bands to suppress noise. Its advantage lies in its high computational efficiency, but its denoising effect on low-frequency trend components and non-stationary signals is limited.

[0005] Wavelet packet thresholding further subdivides the frequency band based on wavelet thresholding, improving its adaptability to complex signals. However, its computational complexity increases significantly, and it still has insufficient separation capability for mixed signals with high-frequency noise and low-frequency trends.

[0006] Existing methods typically employ a single noise reduction strategy (such as using only wavelet thresholding) without switching noise reduction modes according to signal characteristics (such as preserving trend components for low-frequency signals and enhancing noise suppression for high-frequency signals), resulting in noise reduction results that deviate from actual needs. Summary of the Invention

[0007] This invention provides a method for denoising vibration signals from tunnel lining surfaces, which solves the problem of inaccurate signal feature extraction caused by uneven initial signal position distribution in the existing Bald Eagle Search algorithm, as well as the problem of denoising results deviating from actual requirements due to the use of a single denoising strategy.

[0008] The technical solution adopted in this invention is as follows: A method for noise reduction of vibration signals from tunnel lining surfaces includes: Based on the acquired vibration measurement signals, the sequence is updated through tent mapping to obtain the nonlinear enhancement signal; Based on the nonlinear enhancement signal, signal features are extracted using the improved Bald Eagle algorithm; Based on the signal characteristics, the signal's stationarity and low-frequency characteristics are determined. If the signal is stationary and has low frequencies, noise reduction is performed by improving the wavelet threshold. Otherwise, noise reduction can be achieved by improving the wavelet packet threshold.

[0009] The method for noise reduction of vibration signals from tunnel lining surfaces disclosed in this invention also has the following additional technical features: Sequence updates are performed via tender mapping, specifically as follows: , in, For measuring vibration signals, The nonlinear enhancement signal obtained from sequence update, This is a chaos parameter, with a value range of (0, 1).

[0010] The improved vulture algorithm is as follows: The nonlinear enhancement signal is processed according to the preset vulture algorithm parameters, wherein the vulture algorithm includes a selection phase, a search phase, and a capture phase; The signal-to-noise ratio (SNR) state parameter threshold is obtained by using the power spectrum of the signal and the average power spectrum corresponding to the signal frequency, and is used to update the parameters of the Bald Eagle algorithm. During the update process, a pinhole imaging learning or Golden-sine local development strategy is determined based on the global adaptive volatility for algorithm convergence.

[0011] The parameters for the Bald Eagle algorithm are as follows: The parameters of the bald eagle algorithm include population size. In the selection phase, based on the nonlinear enhancement signal and combined with population data, the current position of each population is determined, the average position of the population is obtained, and the optimal position of the population is selected to update the current position of each population and determine the search area. During the search phase, the target location is determined based on various group search areas; During the acquisition phase, signal characteristics are obtained based on the target location.

[0012] The signal-to-noise ratio (SNR) state parameters are obtained and used to update the parameters of the Bald Eagle algorithm, specifically as follows: The signal-to-noise ratio is obtained from the signal processed by the Bald Eagle algorithm; When the signal-to-noise ratio is less than the signal-to-noise ratio state parameter threshold, the current bald eagle algorithm parameters are updated through a stochastic resonance system.

[0013] Based on the global adaptive volatility, determine the pinhole imaging learning or Golden-sine local development strategy, specifically as follows: When the global adaptive volatility is less than a preset threshold, the Golden-sine strategy is used for processing by the Bald Eagle algorithm. Otherwise, a pinhole imaging learning strategy is used for processing in the Bald Eagle algorithm.

[0014] To determine signal stability and low-frequency characteristics, the following steps are taken: Based on the signal features extracted by the Bald Eagle algorithm, when the stationarity parameter is less than or equal to the stationarity threshold, the signal is determined to be stationary. When the main frequency of a signal is less than or equal to the low-frequency threshold, it is determined to be a low-frequency signal.

[0015] The wavelet threshold is improved as follows: The dynamic wavelet threshold is obtained based on the preset wavelet coefficients, wavelet threshold, and adjustment factor. When the dynamic wavelet threshold is less than the wavelet threshold, the signal value is assigned to 0.

[0016] The wavelet packet threshold is improved as follows: Based on the preset original wavelet packet coefficients, wavelet packet decomposition level, and wavelet packet threshold, combined with the signal component length, the dynamic wavelet packet threshold is obtained. When the dynamic wavelet packet threshold is less than the wavelet packet threshold, the signal value is assigned to 0.

[0017] The present invention also provides a processing apparatus, comprising: Memory, used to store computer programs; A processor is used to implement the steps of the method for noise reduction of vibration signals from tunnel lining surfaces when executing the computer program.

[0018] Due to the adoption of the above technical solution, the beneficial effects achieved by this invention are as follows: 1. In this invention, the original vibration measurement signal is updated sequentially through Tent mapping, which enhances the nonlinear characteristics of the signal, effectively expands the dynamic range of the signal, makes the spectral distribution of noise and effective signal easier to distinguish, and significantly improves the separation capability of noise and effective signal. Especially in low signal-to-noise ratio scenarios, it reduces the false suppression of useful signal and improves the signal fidelity after noise reduction.

[0019] By utilizing an improved bald eagle algorithm for feature extraction of nonlinear enhanced signals, the significant impact of uneven initial population location distribution on optimization results can be avoided in traditional search algorithms. After introducing the Tent mapping, the distribution of the bald eagle population within the search space becomes more uniform, and population diversity is significantly increased. This addresses the shortcomings of uneven initial location distribution and the tendency to get trapped in local optima, which leads to inaccurate signal feature extraction.

[0020] By combining the stationarity and low-frequency characteristics of the signal, a dynamic denoising method is selected. For stationary and low-frequency signals, an improved wavelet threshold is used to preserve low-frequency trend components; for non-stationary or high-frequency signals, an improved wavelet packet threshold is used to enhance noise suppression. This achieves adaptive switching of denoising strategies, taking into account the multi-scale characteristics and non-stationarity of the signal, and avoiding signal distortion or insufficient denoising caused by traditional fixed threshold methods. It also improves the robustness of denoising under complex conditions (such as noise interference in tunnel environments), ensuring that the physical meaning of the signal is not destroyed after denoising.

[0021] In summary, this invention addresses the shortcomings of existing technologies in terms of noise separation capability, noise reduction strategy flexibility, and computational efficiency by enhancing the nonlinearity of the Tent mapping, improving the dynamic feature extraction of the Bald Eagle algorithm, and switching adaptive noise reduction strategies. It achieves efficient noise reduction and feature preservation of vibration measurement signals, providing more reliable technical support for tunnel structure health monitoring. Attached Figure Description

[0022] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this invention, illustrate exemplary embodiments of the invention and are used to explain the invention, but do not constitute an undue limitation of the invention. In the drawings: Figure 1 This is a flowchart illustrating the noise reduction method for vibration measurement signals on tunnel lining surfaces according to one embodiment of the present invention. Detailed Implementation

[0023] To more clearly illustrate the overall concept of the present invention, a detailed description will be provided below with reference to the accompanying drawings and examples.

[0024] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0025] like Figure 1 As shown, a method for noise reduction of vibration signals from tunnel lining surfaces includes: S100: Based on the acquired vibration measurement signal, the sequence is updated through tent mapping to obtain the nonlinear enhancement signal.

[0026] The core purpose of this step is to perform nonlinear enhancement processing on the original vibration measurement signal through Tent mapping, expand the dynamic range of the signal and enhance its nonlinear characteristics, so as to provide clearer input data for subsequent feature extraction and noise reduction, thereby improving the overall noise reduction effect.

[0027] It should be noted that the input signal preprocessing involves standardizing the acquired vibration measurement signal (usually time series data) to the interval (0,1] to ensure that it conforms to the mathematical properties of the Tent mapping.

[0028] By applying the Tent mapping, the standardized signal sequence is iteratively updated according to the piecewise function rules of the Tent mapping. After multiple iterations, a nonlinear enhanced signal is output, whose dynamic range is significantly expanded compared to the original signal, and the spectral distribution of high-frequency noise and low-frequency trends is more easily distinguishable.

[0029] Understandably, traditional linear processing methods (such as Fourier transform or conventional wavelet transform) are not very sensitive to the nonlinear components in vibration measurement signals and are difficult to effectively separate noise from valid signals. However, this invention, through the nonlinear enhancement of Tent mapping, can significantly improve the local abrupt change characteristics of the signal (such as impact noise or transient vibration), providing a clearer spectral boundary for subsequent noise reduction.

[0030] Specifically, the Tent map amplifies minute fluctuations in the signal through a piecewise scaling mechanism, making the difference between low-amplitude noise and high-amplitude effective signals more significant. The chaotic nature of the Tent map can introduce controllable nonlinear perturbations, preventing the loss of useful information due to over-smoothing in subsequent signal processing.

[0031] In summary, this invention overcomes the limitations of traditional linear denoising methods by introducing Tent mapping into the preprocessing stage of tunnel vibration measurement signals. By directly enhancing signal features through chaotic mapping, it provides higher-quality input data for subsequent feature extraction and adaptive denoising strategies, thus solving the technical bottlenecks of traditional vibration signal denoising methods in terms of dynamic range expansion, nonlinear feature extraction, and denoising robustness.

[0032] S200: Based on the nonlinear enhancement signal, extract signal features using the improved Bald Eagle algorithm.

[0033] The core objective of this step is to extract features from the nonlinear enhancement signal using the improved Bald Eagle Algorithm (BES), dynamically identify the key spectral components and energy distribution characteristics of the signal, and provide a decision-making basis for the adaptive selection of subsequent noise reduction strategies, thereby improving the targeting and effectiveness of noise reduction.

[0034] Using the nonlinear enhancement signal output from the previous step as input, initialize parameters such as population size, number of iterations, and search range for the Bald Eagle algorithm.

[0035] The improved vulture algorithm simulates the "reconnaissance-search" behavior of vulture flocks. It uses chaotic perturbations (such as sequences generated by Tent mappings) to disrupt the initial population position, preventing the algorithm from getting trapped in local optima. It performs local target acquisition and search, and evaluates feature saliency by combining the signal's spectral energy distribution.

[0036] When the number of iterations reaches a preset value or the feature convergence threshold is met, the optimal feature set is output. Based on the output features, stationarity indices (such as the Hurst exponent) and low-frequency indices (such as the proportion of low-frequency energy) are obtained, which are used to select and judge subsequent noise reduction strategies.

[0037] In this step, the improved Bald Eagle algorithm introduces a chaotic perturbation mechanism (such as the nonlinear sequence of the Tent map) to enhance the algorithm's ability to escape local optima and ensure the comprehensiveness of feature extraction. It can adapt to the stationary and low-frequency characteristics of vibration measurement signals, thus enabling accurate selection of noise reduction methods.

[0038] In summary, this step uses the improved Bald Eagle algorithm to extract features from the nonlinear enhancement signal, which solves the technical bottleneck of traditional methods in dynamic feature recognition and provides key decision support for the adaptive selection of subsequent noise reduction strategies.

[0039] S300: Based on the signal characteristics, determine the signal's stationarity and low-frequency characteristics. If the signal is stationary and low-frequency, then perform noise reduction by improving the wavelet threshold. Otherwise, noise reduction can be achieved by improving the wavelet packet threshold.

[0040] The core purpose of this step is to dynamically determine the signal characteristics (stationarity and low frequency) based on the signal features (stationarity and low frequency) extracted in the previous step, and adaptively select the improved wavelet threshold or improved wavelet packet threshold as the noise reduction strategy, so as to maximize the noise suppression effect while preserving the effective signal and solve the noise reduction mismatch problem caused by the difference in signal characteristics in the traditional fixed noise reduction method.

[0041] For stationarity assessment, the time-domain stability of a signal is evaluated using the Hurst exponent, autocorrelation function, or Hjorth parameter. If the Hurst exponent... This is determined to be a stable signal.

[0042] For low-frequency determination, the proportion of signal energy in the low-frequency sub-band (such as the low-frequency coefficients of wavelet decomposition) is calculated. If the proportion exceeds a preset threshold (such as 60%), it is determined to be a low-frequency signal.

[0043] If the signal is stable and low-frequency, an improved wavelet threshold is used to retain high-energy components in the low-frequency subbands, while only applying soft / hard thresholding to the high-frequency subbands to avoid disrupting the signal trend.

[0044] If the signal is not stationary or not low frequency, an improved wavelet packet threshold is adopted. The frequency band division is refined by wavelet packet decomposition, and multi-scale threshold optimization is performed on the high-frequency noise sub-band, while preserving the low-frequency trend and transient characteristics.

[0045] Understandably, traditional wavelet thresholding methods (such as Donoho thresholding) employ a fixed denoising strategy, which has limited effectiveness in denoising non-stationary signals (such as tunnel blasting vibrations) or multi-band mixed signals (such as vehicle traffic noise), easily leading to the loss of useful signals or noise residue. While wavelet packet thresholding can refine the frequency band, it does not adaptively select based on signal characteristics, resulting in wasted computational resources or over-denoising.

[0046] This step employs an adaptive noise reduction strategy, dynamically selecting the noise reduction method through signal feature classification. This approach balances low-frequency trend preservation with high-frequency noise suppression, avoiding the limitations of a single strategy. Improving the wavelet threshold increases the retention rate of stationary low-frequency signals, while improving the wavelet packet threshold enhances the fidelity of transient features in non-stationary signals, thereby improving overall signal fidelity.

[0047] This step, through adaptive switching of noise reduction strategies driven by signal characteristics, solves the technical bottlenecks of traditional methods in terms of signal characteristic adaptability and noise reduction accuracy, and significantly improves the noise reduction effect and signal fidelity of vibration measurement signals.

[0048] In a preferred embodiment of the present invention, sequence updates are performed through tent mapping, specifically as follows: , in, For measuring vibration signals, The nonlinear enhancement signal obtained from sequence update, This is a chaos parameter, with a value range of (0, 1], used to control the nonlinearity of the mapping. It should be noted that this can be achieved through... The value of (e.g., 0.4 <) <0.7), balancing signal enhancement and fidelity, suitable for vibration measurement scenarios with different noise levels.

[0049] For the standardized vibration measurement signal By iteratively applying the above mapping rules, a nonlinear enhancement signal is generated. This makes it easier to distinguish between minute fluctuations in the signal and its spectral distribution.

[0050] Understandably, the uneven initial position distribution of the population in the traditional vulture algorithm can significantly impact the optimization results. Tent mapping offers advantages such as faster iteration speed, better distribution, and stronger randomness. Through Tent mapping, nonlinear enhancement can significantly improve the local abrupt changes in the signal (such as impulse noise or transient vibrations), amplify minute signal fluctuations, and make the difference between low-amplitude noise and high-amplitude effective signals more significant. This results in a more uniform distribution of the vulture population within the search space, while also significantly increasing population diversity.

[0051] This implementation method overcomes the limitations of traditional linear noise reduction methods by introducing tent mapping into the preprocessing stage of tunnel vibration measurement signals. Through nonlinear enhancement processing, it solves the technical bottleneck of uneven initial search space in the Bald Eagle algorithm.

[0052] As a preferred embodiment of the present invention, the improved bald eagle algorithm is specifically as follows: The nonlinear enhancement signal is processed according to the preset vulture algorithm parameters, wherein the vulture algorithm includes a selection phase, a search phase, and a capture phase; The signal-to-noise ratio (SNR) state parameter threshold is obtained by using the power spectrum of the signal and the average power spectrum corresponding to the signal frequency, and is used to update the parameters of the Bald Eagle algorithm. During the update process, a pinhole imaging learning or Golden-sine local development strategy is determined based on the global adaptive volatility for algorithm convergence.

[0053] The core objective of this implementation is to improve the global search capability and local exploitation efficiency of the improved Bald Eagle Algorithm (BES) by combining signal power spectrum analysis with an algorithm convergence strategy switching mechanism. This optimizes feature extraction accuracy and accelerates algorithm convergence, solving the problems of traditional BES algorithms being prone to getting trapped in local optima, slow convergence speed, and insufficient parameter adaptability in complex signal processing.

[0054] It should be noted that the parameters of the Bald Eagle algorithm are as follows: The parameters of the vulture algorithm include population size. In the selection phase, based on the nonlinear enhancement signal and combined with population data, the current position of each population is determined, the average position of the population is obtained, and the optimal position of the population is selected to update the current position of each population and determine the search area. During the search phase, the target location is determined based on various group search areas; During the acquisition phase, signal characteristics are obtained based on the target location.

[0055] The core objective of this embodiment is to improve the global search capability and local development efficiency of the Bald Eagle Algorithm (BES) in feature extraction by optimizing the parameter settings and phased processing mechanism of the BES algorithm and combining it with the characteristics of nonlinear enhancement signals, thereby more accurately identifying the key features of vibration measurement signals.

[0056] In this process, the population size is set according to the signal complexity (e.g., 50-100) to ensure that the algorithm covers a sufficient number of candidate solutions in the search space and avoids premature convergence.

[0057] Based on the power spectrum distribution of the nonlinear enhancement signal, the location of the bald eagle population is randomly initialized. Prioritize coverage of the dominant signal frequency band (such as low-frequency trends and high-frequency noise regions). Calculate the population's average location. with optimal position .

[0058] Update the vulture's position according to the formula: , in, Control the search step size.

[0059] During the selection phase, regions with significant signal characteristics (such as low-frequency trends or impulse noise bands) are quickly located through global exploration.

[0060] Within the selected search area, a spiral search strategy is employed: , in, From polar angle and polar diameter calculate.

[0061] During the search phase, the search range is refined to accelerate convergence to sub-bands with significant signal characteristics (such as specific frequency components).

[0062] During the capture phase, the vulture captures information from the determined optimal location: , in, and This represents the intensity of the vulture's movement towards the optimal position and the center position, with values ​​of [1, 2].

[0063] Based on the target location, extract key features of the signal, including: power spectrum distribution, calculated by FFT to determine the energy proportion of the signal in the corresponding frequency band at the target location; and frequency distribution, identifying the dominant frequency components of the signal (such as low-frequency trend frequencies). With high frequency noise The signal-to-noise ratio (SNR) state parameter, combined with the power spectrum and average power spectrum at the target location, is used to calculate the SNR threshold. This provides accurate input for subsequent noise reduction strategies (such as wavelet packet decomposition).

[0064] This embodiment dynamically adjusts the population initialization position and search step size by analyzing the power spectrum distribution of the nonlinearly enhanced signal, thereby improving the algorithm's adaptability to signal features. Combining global exploration in the selection phase with local refinement in the search phase avoids premature convergence. By optimizing the bald eagle algorithm parameter settings and implementing a phased dynamic adjustment mechanism, the feature extraction efficiency and noise reduction effect of the vibration measurement signal are significantly improved.

[0065] As another embodiment of this implementation, the signal-to-noise ratio state parameter is obtained and used to update the parameters of the Bald Eagle algorithm, specifically as follows: The signal-to-noise ratio is obtained from the signal processed by the Bald Eagle algorithm; When the signal-to-noise ratio is less than the signal-to-noise ratio state parameter threshold, the current bald eagle algorithm parameters are updated through a stochastic resonance system.

[0066] The core objective of this embodiment is to optimize the performance of the Bald Eagle Algorithm (BES) by dynamically adjusting the parameters of the BES algorithm and combining it with the characteristics of the stochastic resonance system when the signal-to-noise ratio (SNR) is lower than a preset threshold. This solves the problem of slow convergence speed and easy getting trapped in local optima in the traditional BES algorithm under low SNR environment, thereby improving the robustness and accuracy of vibration signal feature extraction.

[0067] The signal-to-noise ratio (SNR) of the signal processed by the Bald Eagle algorithm (such as the power spectrum reconstructed signal) is calculated using the following formula:

[0068] in, For the main frequency band energy of the signal, This refers to the energy in the noise frequency band. It should be noted that the SNR state parameter threshold is set according to the vibration signal scenario (such as tunnel blasting vibration or vehicle traffic noise). (e.g., 10dB).

[0069] If the calculation yields If this occurs, the parameter update mechanism of the stochastic resonance system is activated. Thus, in low signal-to-noise ratio scenarios, by enhancing the algorithm's global search capability, local optima traps caused by noise interference are avoided.

[0070] In the nonlinear stochastic resonance system, the resonance phenomenon is simulated using the following formula to optimize the BES parameters: , Where x is the system response, corresponding to the parameters of the BES algorithm (such as the population step size). α or search radius r γ is the damping coefficient.a The restoring force coefficient is nonlinear. External periodic excitation (analog signal characteristic changes). This is for the noise term (simulating a low signal-to-noise ratio environment).

[0071] System output x Mapped to dynamic parameters of the BES algorithm, for example: , in, As the reference step size, k The proportionality coefficient (e.g.) k =0.1).

[0072] By dynamically adjusting the global search step size of the BES through the nonlinear response of the stochastic resonance system. α The search scope is expanded to cover more candidate solutions, enhancing the global search. High-frequency components of the system output are used to refine the local search direction, improving convergence accuracy.

[0073] Understandably, traditional BES algorithms rely on experience to set the step size. α and search radius r It is difficult to adapt to the dynamic changes of signal characteristics in low signal-to-noise ratio scenarios. Under noise interference, traditional BES algorithms are prone to getting trapped in local optima (such as misjudging noise frequency bands as signal features), leading to feature extraction failure.

[0074] This embodiment uses a stochastic resonance system to adjust the BES parameters in real time, enabling the algorithm to maintain global search capability in low signal-to-noise ratio scenarios and avoiding premature convergence. The stochastic resonance system, by introducing nonlinear excitation and noise terms, simulates a real signal environment, improving the algorithm's stability under noise interference.

[0075] Furthermore, the local search direction is refined by using high-frequency components in the system output, thereby improving the accuracy of feature extraction. Dynamic parameter updates are triggered only when the signal-to-noise ratio is below a threshold, avoiding redundant computation and reducing overall energy consumption.

[0076] In summary, this embodiment solves the technical bottleneck of parameter rigidity and low convergence efficiency of the traditional BES algorithm in low signal-to-noise ratio environments by using a stochastic resonance system update mechanism driven by signal-to-noise ratio state parameters, and significantly improves the accuracy and robustness of vibration signal feature extraction.

[0077] As another embodiment of this implementation, the pinhole imaging learning or Golden-sine local development strategy is determined based on the global adaptive volatility, specifically as follows: When the global adaptive volatility is less than a preset threshold, the Golden-sine strategy is used for processing by the Bald Eagle algorithm. Otherwise, a pinhole imaging learning strategy is used for processing in the Bald Eagle algorithm.

[0078] The core objective of this embodiment is to dynamically switch the strategy mode of the Bald Eagle Algorithm (BES) and combine it with the evaluation results of global adaptive volatility to intelligently select either the Golden-sine local development strategy or the pinhole imaging learning strategy during the algorithm execution process. This balances the algorithm's global exploration capability with local development efficiency, and solves the problems of traditional BES algorithms being prone to getting trapped in local optima, having slow convergence speed, and insufficient parameter adaptability in complex signal processing.

[0079] Wherein, the global fitness volatility ρ characterizes the degree of dynamic change in the population fitness function, and the calculation formula is: , in, CR represents the standard deviation of the current population fitness, reflecting population diversity; CR represents the convergence rate of the current iteration (e.g., the difference rate between optimal solutions of adjacent generations).

[0080] It should be noted that the threshold (e.g., 0.5) is set according to the signal processing scenario (e.g., the dynamic signal-to-noise ratio change of tunnel vibration measurement signals).

[0081] Example 1: In a low volatility scenario, the population fitness tends to stabilize, and the convergence rate is relatively fast, indicating that the algorithm has entered the local development stage. A Golden-sine local development strategy is adopted, using a sine function to optimize the neighborhood search direction and improve convergence accuracy.

[0082] Specifically, , in, By periodically adjusting the search direction using a sine function, the diversity of local development is enhanced. By controlling the search step size, the algorithm can achieve a balance between convergence speed and accuracy.

[0083] This embodiment utilizes The periodic perturbation at the globally optimal position g Guided by this, a high-frequency, small-step local search is performed on the current individual to improve convergence speed and accuracy.

[0084] Example 2: The population fitness fluctuates wildly, and the convergence rate is slow, indicating that the algorithm is in the global exploration stage. A pinhole imaging learning strategy is adopted, using a "focus-diffusion" mechanism similar to pinhole imaging to expand the search range and avoid local optima traps.

[0085] Specifically, , in, To ensure uniformly distributed random numbers ∈ [0,1], random perturbation is introduced to prevent the population from prematurely converging to a local optimum. θ is an adaptive angle adjustment factor that dynamically adjusts the search direction (e.g., θ∈[0,1]). π / 2] enhances exploration, θ∈[ π / 2, π [Enhance development].

[0086] This embodiment utilizes the periodic fluctuations of cos(θ) combined with... The random perturbation causes the solution to be searched nonlinearly near the historical best position, balancing global exploration and local development.

[0087] By using global adaptive volatility to determine the algorithm's current stage (exploration / development) in real time, intelligent strategy switching is implemented. The Golden-sine strategy optimizes local development efficiency, while the pinhole imaging strategy enhances global exploration capabilities, preventing the algorithm from getting trapped in local optima and improving its robustness.

[0088] In summary, this embodiment solves the technical bottlenecks of the traditional BES algorithm in terms of dynamic parameter adaptability, global-local search balance, and computational efficiency by adopting a strategy switching mechanism based on global adaptive volatility, and significantly improves the feature extraction accuracy and noise reduction effect of nonlinear enhancement signals.

[0089] In a preferred embodiment of the present invention, the determination of signal stability and low-frequency characteristics specifically involves: Based on the signal features extracted by the Bald Eagle algorithm, when the stationarity parameter is less than or equal to the stationarity threshold, the signal is determined to be stationary. When the main frequency of a signal is less than or equal to the low-frequency threshold, it is determined to be a low-frequency signal.

[0090] The core objective of this implementation is to use the signal features extracted by the Bald Eagle algorithm, combined with stationarity parameters and low-frequency thresholds, to quickly determine the stationarity and low-frequency characteristics of a signal, thereby optimizing subsequent signal processing strategies.

[0091] For stationarity assessment, the time-domain stability of a signal is evaluated using the Hurst exponent, autocorrelation function, or Hjorth parameter. If the Hurst exponent... This is determined to be a stable signal.

[0092] For low-frequency determination, the proportion of signal energy in the low-frequency sub-band (such as the low-frequency coefficients of wavelet decomposition) is calculated. If the proportion exceeds a preset threshold (such as 60%), it is determined to be a low-frequency signal.

[0093] Specifically, if the signal is stable and low-frequency, noise reduction is performed by improving the wavelet threshold, wherein improving the wavelet threshold specifically involves: The dynamic wavelet threshold is obtained based on the preset wavelet coefficients, wavelet threshold, and adjustment factor. When the dynamic wavelet threshold is less than the wavelet threshold, the signal value is assigned to 0.

[0094] The core objective of this embodiment is to optimize the signal denoising effect through a dynamic wavelet threshold function, thereby overcoming the limitations of traditional soft / hard thresholding methods. By using an adjustable parameter adaptive threshold function, it balances continuity and detail preservation, thus improving denoising accuracy.

[0095] Among them, wavelet coefficients These are the signal coefficients obtained through wavelet decomposition. Wavelet threshold. A preset threshold (e.g., based on noise estimation or a general threshold) is used. An adjustment factor p∈(0,1] is used to control the shape of the threshold function, determining the soft / hard threshold characteristics. Specifically, .

[0096] If η(ω,λ) < λ, the signal value is set to 0 to suppress noise. The dynamic threshold η(ω,λ) is usually greater than λ, preserving the signal (retaining high-frequency details). The processed dynamic threshold coefficient η(ω,λ) is substituted into the inverse wavelet transform to reconstruct the denoised signal.

[0097] This embodiment achieves a balance between the continuity of signal denoising and the preservation of details by using an adjustable parameter adaptive threshold function combined with dynamic condition judgment (the signal is set to zero when the dynamic threshold is less than the preset threshold).

[0098] Otherwise, noise reduction is achieved by improving the wavelet packet threshold, specifically by: Based on the preset original wavelet packet coefficients, wavelet packet decomposition level, and wavelet packet threshold, combined with the signal component length, the dynamic wavelet packet threshold is obtained. When the dynamic wavelet packet threshold is less than the wavelet packet threshold, the signal value is assigned to 0.

[0099] The core objective of this embodiment is to optimize the denoising effect by dynamically adjusting the wavelet packet threshold through an improved wavelet packet thresholding method, under the premise that the signal is non-stationary or high-frequency.

[0100] Among them, the original wavelet packet coefficients These are the signal coefficients obtained through wavelet packet decomposition, used for subsequent thresholding. The number of wavelet packet decomposition layers... The fineness of wavelet packet decomposition determines the noise suppression effect. Wavelet packet threshold Specifically: , in, is the median of the absolute values ​​of the wavelet packet coefficients; M: the length of the IMF (Intrinsic Mode Function) component.

[0101] Combined signal component length M Number of decomposition layers j and preset threshold The threshold is dynamically adjusted to adapt to the signal characteristics.

[0102] In addition, regulatory factors Specifically: , By decomposing the number of layers j Dynamic adjustment Balance hard / soft threshold characteristics.

[0103] , This embodiment uses adjustment factors Achieve a smooth transition and reduce the discontinuities of traditional methods. When Preserve high-frequency details; when Suppress noise (assign a value of 0).

[0104] This embodiment is illustrated by... Dynamically balance soft / hard threshold characteristics to reduce signal distortion. Combined with signal component length. M and decomposition layer number j It is suitable for non-stationary / high-frequency signals. It effectively solves the limitations of traditional soft / hard thresholds and achieves better noise reduction in non-stationary / high-frequency signal scenarios.

[0105] The present invention further provides a processing apparatus, comprising: Memory, used to store computer programs; A processor is used to implement the steps of the method for noise reduction of vibration signals from tunnel lining surfaces when executing the computer program.

[0106] Therefore, any effect that can be achieved by the method of noise reduction of vibration signals on tunnel lining surfaces will not be elaborated here.

[0107] For any parts not mentioned in this invention, existing technologies can be used or referenced.

[0108] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

[0109] The above description is merely an embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the present invention should be included within the scope of the claims of the present invention.

Claims

1. A method for noise reduction of vibration signals from tunnel lining surfaces, characterized in that, include: Based on the acquired vibration measurement signals, the sequence is updated through tent mapping to obtain the nonlinear enhancement signal; Based on the nonlinear enhancement signal, signal features are extracted using the improved Bald Eagle algorithm; Based on the signal characteristics, the signal's stationarity and low-frequency characteristics are determined. If the signal is stationary and has low frequencies, noise reduction is performed by improving the wavelet threshold. Otherwise, noise reduction can be achieved by improving the wavelet packet threshold.

2. The method for noise reduction of vibration signals from tunnel lining surfaces according to claim 1, characterized in that, Sequence updates are performed via tender mapping, specifically as follows: ， in, For measuring vibration signals, The nonlinear enhancement signal obtained from sequence update, This is a chaos parameter, with a value range of (0, 1).

3. The method for noise reduction of vibration signals from tunnel lining surfaces according to claim 1, characterized in that, The improved vulture algorithm is as follows: The nonlinear enhancement signal is processed according to the preset vulture algorithm parameters, wherein the vulture algorithm includes a selection phase, a search phase, and a capture phase; The signal-to-noise ratio (SNR) state parameter threshold is obtained by using the power spectrum of the signal and the average power spectrum corresponding to the signal frequency, and is used to update the parameters of the Bald Eagle algorithm. During the update process, a pinhole imaging learning or Golden-sine local development strategy is determined based on the global adaptive volatility for algorithm convergence.

4. The method for noise reduction of vibration signals from tunnel lining surfaces according to claim 3, characterized in that, The parameters for the Bald Eagle algorithm are as follows: The parameters of the vulture algorithm include population size. In the selection phase, based on the nonlinear enhancement signal and combined with population data, the current position of each population is determined, the average position of the population is obtained, and the optimal position of the population is selected to update the current position of each population and determine the search area. During the search phase, the target location is determined based on various group search areas; During the acquisition phase, signal characteristics are obtained based on the target location.

5. The method for noise reduction of vibration signals from tunnel lining surfaces according to claim 3, characterized in that, The signal-to-noise ratio (SNR) state parameters are obtained and used to update the parameters of the Bald Eagle algorithm, specifically as follows: The signal-to-noise ratio is obtained from the signal processed by the Bald Eagle algorithm; When the signal-to-noise ratio is less than the signal-to-noise ratio state parameter threshold, the current bald eagle algorithm parameters are updated through a stochastic resonance system.

6. The method for noise reduction of vibration signals from tunnel lining surfaces according to claim 3, characterized in that, Based on the global adaptive volatility, determine the pinhole imaging learning or Golden-sine local development strategy, specifically as follows: When the global adaptive volatility is less than a preset threshold, the Golden-sine strategy is used for processing by the Bald Eagle algorithm. Otherwise, a pinhole imaging learning strategy is used for processing in the Bald Eagle algorithm.

7. The method for noise reduction of vibration signals from tunnel lining surfaces according to claim 1, characterized in that, To determine signal stability and low-frequency characteristics, the following steps are taken: Based on the signal features extracted by the Bald Eagle algorithm, when the stationarity parameter is less than or equal to the stationarity threshold, the signal is determined to be stationary. When the main frequency of a signal is less than or equal to the low-frequency threshold, it is determined to be a low-frequency signal.

8. The method for noise reduction of vibration signals from tunnel lining surfaces according to claim 7, characterized in that, The wavelet threshold is improved as follows: The dynamic wavelet threshold is obtained based on the preset wavelet coefficients, wavelet threshold, and adjustment factor. When the dynamic wavelet threshold is less than the wavelet threshold, the signal value is assigned to 0.

9. The method for noise reduction of vibration signals from tunnel lining surfaces according to claim 7, characterized in that, The wavelet packet threshold is improved as follows: Based on the preset original wavelet packet coefficients, wavelet packet decomposition level, and wavelet packet threshold, combined with the signal component length, the dynamic wavelet packet threshold is obtained. When the dynamic wavelet packet threshold is less than the wavelet packet threshold, the signal value is assigned to 0.

10. A processing apparatus, characterized in that, include: Memory, used to store computer programs; A processor, configured to execute the computer program to implement the steps of the method for noise reduction of vibration signals from tunnel lining surfaces as described in any one of claims 1 to 9.

Images