Road vibration signal analysis method based on multi-scale wavelet transform

Abstract

The invention discloses a road vibration signal analysis method based on multi-scale wavelet transform, and relates to the technical field of signal analysis. The method comprises the following steps of: 1, performing convolution filtering on an original road time sequence vibration signal by adopting a Gaussian band-pass filtering kernel so as to capture local changes of a standardized signal in different scales and time domains; step 2, according to the time domain scale coefficient, calculating a noise standard deviation and setting a threshold value in each local time window based on median statistics, and normalizing the noise standard deviation into probability distribution; 3, calculating a first spectral moment and a second spectral moment to obtain a comprehensive spectral feature; under adjacent scales, correlation coefficients are calculated through local cross integration; and 4, according to the comprehensive spectrum characteristics and the correlation coefficient, constructing nonlinear mapping by adopting a hyperbolic tangent function, and calculating the road damage probability. According to the invention, high-precision and high-robustness road health monitoring can be realized.

Description

Technical Field

[0001] The present invention relates to the technical field of signal analysis, and in particular to a road vibration signal analysis method based on multi-scale wavelet transform. Background Art

[0002] At present, the methods for road structure health monitoring at home and abroad mainly include those based on visual inspection, strain monitoring, acceleration sensor analysis and signal processing technology. Among them, the methods based on visual inspection mainly rely on drone aerial photography, high-definition camera monitoring or laser scanners to identify road surface defects. This method has certain advantages in detecting visible damage such as cracks and potholes on the road surface, but it is difficult to detect hidden damage (such as internal road structure degradation or foundation anomalies), and is greatly affected by ambient lighting, weather conditions and shooting angles, and has limited reliability. In addition, visual inspection methods usually require complex image processing and deep learning algorithms, which are computationally intensive and are not suitable for real-time monitoring.

[0003] Strain monitoring technology obtains strain data and analyzes the deformation of roads under load by deploying strain sensors on the road surface or inside. However, the deployment cost of strain sensors is high, and they are easily affected by changes in ambient temperature and humidity, and have poor long-term stability. At the same time, since strain monitoring mainly reflects the static response of the structure, it is difficult to effectively capture short-term transient impact signals (such as dynamic vibration responses caused by rapid vehicle passage), which limits its application in early damage detection.

[0004] Acceleration sensing analysis technology is an important development direction in the field of road health monitoring in recent years. This method collects road vibration signals through sensors on the road surface or embedded in the road, and analyzes the road structure status based on the acceleration response characteristics. Compared with strain monitoring, acceleration signals can capture short-term dynamic characteristics more sensitively and are suitable for the detection of various damage modes. However, traditional acceleration signal analysis methods still have certain limitations. For example, although the traditional Fourier transform (FFT) can analyze the frequency components of the vibration signal, it cannot provide the time evolution information of the signal, so it is difficult to detect short-term transient damage. Although the short-time Fourier transform (STFT) can combine time and frequency information, its window function is fixed and it is difficult to adapt to the changes of multi-scale vibration signals, resulting in limited time-frequency resolution. In the actual acquisition process, road vibration signals are easily affected by factors such as environmental noise, vehicle load changes, and sensor measurement errors, resulting in a large amount of random noise in the signal. Existing signal denoising methods mainly include low-pass filtering, wavelet denoising, and empirical mode decomposition (EMD), but these methods may lose some useful signal features while suppressing noise, affecting the accuracy of damage detection. Summary of the invention

[0005] In view of this, the present invention provides a road vibration signal analysis method based on multi-scale wavelet transform, which uses Gaussian bandpass filtering combined with normalization processing to improve signal quality, and extracts time domain scale coefficients through attenuation weighted continuous wavelet transform to ensure that signal features in different frequency ranges can be accurately captured. In addition, an adaptive denoising method based on local energy statistics is used to improve the signal-to-noise ratio, and at the same time, comprehensive spectrum features and correlation coefficients are constructed to quantify road damage characteristics, and combined with mechanical parameters such as elastic modulus, fatigue limit stress, and foundation stiffness of materials, baseline anomaly indicators and critical anomaly indicators are constructed. Finally, the road damage probability is calculated through nonlinear mapping to form a complete damage quantification assessment system. The present invention can effectively overcome the problems of insufficient multi-scale feature extraction capability, serious noise interference, lack of physical basis for damage assessment, and unclear quantitative standards in the prior art, and realize high-precision and high-robustness road health monitoring, providing a scientific decision-making basis for road maintenance. .

[0006] The technical solution adopted by the present invention is as follows: A road vibration signal analysis method based on multi-scale wavelet transform, the method comprising: Step 1: Perform convolution filtering on the original road time series vibration signal using a Gaussian bandpass filter kernel, obtain the filtered signal and then normalize it to form a standardized signal; use the attenuation weighted continuous wavelet transform on the standardized signal to obtain the time domain scale coefficient to capture the local changes of the standardized signal at different scales and time domains; Step 2: According to the time domain scale coefficient, the noise standard deviation is calculated and the threshold is set based on the median statistics in each local time window, and the denoised time domain scale coefficient is obtained through hard threshold processing; the local energy is calculated using the denoised time domain scale coefficient and normalized to a probability distribution; Step 3: Perform spectral moment statistics on local energy in the local time window, calculate the first spectral moment and the second spectral moment, and obtain comprehensive spectral features; calculate the correlation coefficient by local cross integration at adjacent scales; Step 4: Based on the comprehensive spectrum characteristics and correlation coefficients, a global abnormality index is constructed to reflect the abnormal road vibration. The predefined baseline abnormality index and critical abnormality index are combined, and the hyperbolic tangent function is used to construct a nonlinear mapping to calculate the road damage probability.

[0007] Furthermore, in step 1, sensors are deployed on the target road at a set sampling interval. , collect the original road time series vibration signal of the target road; suppose the collected original road time series vibration signal is , using a bandpass filter for preprocessing, bandpass filter is a Gaussian bandpass kernel, expressed as follows: ; in, is the filter width parameter; is the center frequency, corresponding to the main frequency band of road vibration; is a time variable; the filtered signal after filtering is ;in is the total signal acquisition time; is the time integral variable; through the following formula, Perform normalization to obtain a standardized signal: ; in, is a standardized signal.

[0008] Furthermore, in step 1, the standardized signal is subjected to attenuated weighted continuous wavelet transform to obtain the time domain scaling coefficient : ; in, , is the scale parameter; , is the translation parameter; is the mother wavelet function, ; is the imaginary number symbol; is the time domain scaling factor.

[0009] Furthermore, in step 2, in order to adaptively suppress noise interference in the time domain, at each scale and local time window The local energy statistics are used to estimate the noise level and calculate the local noise standard deviation. for: ; in, is the local time window width; is the median function; the local adaptive threshold is defined by the following formula: ; in, is the sampling interval, reflecting the number of local samples, and the hard threshold strategy is used for denoising to obtain the denoised coefficient : ; in, It is an absolute value operation.

[0010] Furthermore, in step 2, the local energy distribution is calculated using the denoised wavelet coefficients: ; To eliminate the amplitude effect caused by scale and time domain integration, the energy is normalized to a probability distribution : ; in, is the scale integral variable; is the translation integral variable.

[0011] Furthermore, in step 3, the spectral moment statistics of the denoised energy distribution are performed in the local time window to characterize the center and expansion of the energy distribution, thereby reflecting the local structural state; Next, in the time window The first spectral moment is defined internally and the second spectral moment for: ; in, is the Poisson’s ratio of the road material of the target road; is the road material density of the target road; is the road material damping coefficient of the target road; is the pavement layer thickness of the target road; ; in, is the foundation rebound modulus of the target road; is the foundation damping coefficient of the target road; is the foundation stiffness of the target road.

[0012] Furthermore, in step 3, the comprehensive spectrum feature is defined by the following formula: : ; in, is the fatigue limit stress of the road material of the target road; is the elastic modulus of the road material of the target road.

[0013] Furthermore, in step 3, the correlation coefficient is calculated by local cross integration at adjacent scales using the following formula: ; in, is the correlation coefficient.

[0014] Furthermore, the global anomaly index is constructed through the following formula: ; in, It is the global abnormality indicator.

[0015] Furthermore, the baseline abnormality index is defined by the following formula: : ; in, for The mean of for The critical abnormality index is defined by the following formula: : ; in, is the local settlement of the target road; the road damage probability is calculated by the following formula: ; in, is the road damage probability.

[0016] By adopting the above technical scheme, the present invention produces the following beneficial effects: Based on the multi-scale wavelet transform method, the present invention can simultaneously analyze the time-frequency characteristics of road vibration signals in different frequency ranges, which has significant advantages over traditional time domain or frequency domain methods. Traditional signal analysis methods are usually based on fixed windows or single scales for analysis, which makes it difficult to take into account the characteristics of different vibration modes. The present invention utilizes the time-frequency localization characteristics of multi-scale wavelet transform to decompose the signal in different scales, so that low-frequency and high-frequency vibration information can be accurately extracted at the same time. This method is particularly suitable for road health monitoring scenarios, because road damage often manifests as changes in different frequency components, such as changes in the overall vibration mode of the structure usually appear in the low-frequency range, while local damage, crack extension or uneven foundation settlement will cause abnormal fluctuations in high-frequency signals. By constructing the time-frequency distribution of vibration signals at different scales, the present invention can more accurately capture the detailed characteristics of road damage and improve the sensitivity of damage detection. The present invention adopts an adaptive denoising strategy to optimize the denoising and feature extraction process of the signal. In the actual acquisition process, the road vibration signal is often affected by the external environment, vehicle load changes and sensor measurement errors, resulting in a large amount of random noise in the signal. Existing denoising methods often lead to the loss of key vibration information, thus affecting the accuracy of subsequent analysis. The present invention estimates the noise level through a local energy statistics method and adopts an adaptive threshold denoising strategy, so that the noise component in the signal can be effectively suppressed while retaining key vibration characteristics. Compared with the traditional fixed threshold method, the adaptive denoising strategy of the present invention can be dynamically adjusted according to the noise characteristics of different scales and time windows, ensuring that the signal can still maintain a high signal-to-noise ratio under different environmental conditions. This method not only improves the robustness of signal processing, but also enhances the distinguishability of damage features, laying a solid foundation for subsequent damage assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a flow chart of the application method of intelligent electromagnetic simulation data in an embodiment of the present invention. DETAILED DESCRIPTION

[0018] All features disclosed in this specification, or steps in all methods or processes disclosed, except mutually exclusive features and / or steps, can be combined in any manner.

[0019] Any feature disclosed in this specification (including any additional claims and abstract), unless otherwise stated, may be replaced by other alternative features that are equivalent or have similar purposes. That is, unless otherwise stated, each feature is only an example of a series of equivalent or similar features.

[0020] Example 1, reference Figure 1 :A road vibration signal analysis method based on multi-scale wavelet transform, the method comprising: Step 1: Perform convolution filtering on the original road time series vibration signal using a Gaussian bandpass filter kernel, obtain the filtered signal and then normalize it to form a standardized signal; use the attenuation weighted continuous wavelet transform on the standardized signal to obtain the time domain scale coefficient to capture the local changes of the standardized signal at different scales and time domains; Road vibration signals usually contain components in different frequency ranges, including low-frequency components, which are often derived from the overall load and slow deformation effect of the vehicle, and high-frequency components, which are mainly generated by minor damage, cracks, potholes on the road surface, and instantaneous impacts between the vehicle tires and the road surface. Direct analysis of the original signal may be affected by low-frequency background noise or high-frequency random interference. Therefore, the present invention first uses a Gaussian bandpass filter kernel to filter the original road vibration signal to enhance the ability to extract vibration information in the target frequency band. The role of Gaussian bandpass filtering is to retain the main features of the road vibration signal while suppressing low-frequency drift and high-frequency interference, ensuring that subsequent wavelet transforms can be analyzed within a suitable signal range. Due to the smoothness of the Gaussian bandpass filter kernel, it can effectively reduce the influence of mutations and burr signals in the time domain, thereby improving the stability of signal processing. In addition, considering that different road environments and vehicle types may cause changes in the amplitude of the vibration signal, in order to eliminate the sensitivity differences of the measuring equipment and the uneven amplitude of the vibration signal, the present invention normalizes the signal after filtering. Normalization uses the method of mean removal and standard deviation normalization to make the processed signal have zero mean and unit standard deviation, thereby ensuring the scale consistency of the signal under different measurement conditions and improving the applicability and stability of subsequent wavelet analysis. The normalized signal, as a standardized signal, can more accurately reflect the road vibration characteristics and avoid errors caused by different measurement equipment or changes in signal amplitude.

[0021] Based on the standardized signal, the present invention uses the attenuation weighted continuous wavelet transform to extract the time domain scale coefficient, so as to capture the time-frequency characteristics of the road vibration signal at different scales. The core idea of ​​the wavelet transform is to use a set of mother wavelet functions with scaling and translation characteristics to decompose the signal, so as to analyze the local characteristics of the signal at different scales and time positions at the same time. Since the road vibration signal usually contains multiple components of different frequencies, the direct use of the traditional Fourier transform cannot capture its instantaneous changes. Therefore, the present invention adopts the method of continuous wavelet transform, uses the mother wavelet function and the standardized signal to perform convolution operations, and calculates the time domain scale coefficients at different scales. The time domain scale coefficient can not only characterize the instantaneous characteristics of the signal at different time points, but also characterize the distribution of signal energy in different frequency bands through scale parameters. In the present invention, the attenuation weighted continuous wavelet transform introduces a scale-dependent weight factor, so that the signal components of the high-scale (low-frequency) part are appropriately suppressed during calculation, while the signal components of the low-scale (high-frequency) part are enhanced, thereby improving the ability to analyze the road surface detail information. This weight adjustment mechanism helps to highlight the high-frequency vibration information caused by the pavement microstructure, while not over-amplifying the influence of low-frequency background signals, so that abnormal features such as road damage can be more clearly reflected in the changes in wavelet coefficients.

[0022] Step 2: According to the time domain scale coefficient, the noise standard deviation is calculated and the threshold is set based on the median statistics in each local time window, and the denoised time domain scale coefficient is obtained through hard threshold processing; the local energy is calculated using the denoised time domain scale coefficient and normalized to a probability distribution; Based on the time domain scale coefficient extracted by wavelet transform, the present invention adopts a noise standard deviation estimation method based on median statistics to calculate the noise level in the local time window, and uses a hard threshold method to denoise the time domain scale coefficient to retain the main characteristics of the road vibration signal while effectively suppressing the influence of random noise. Traditional wavelet threshold denoising methods usually rely on global standard deviations or empirical threshold settings, which are difficult to adapt to changes in noise characteristics under different environments. The present invention dynamically calculates the noise standard deviation in the local time window through the median statistics method, making the threshold setting more robust, thereby ensuring that good denoising effects can be obtained under different road conditions and vibration environments. Specifically, in the local time window, the noise standard deviation is first calculated based on the absolute median of the time domain scale coefficient. The advantage of this method is that it is more robust to noise and can avoid statistical deviations caused by individual abnormal points, thereby improving the stability of the denoising method. After obtaining the noise standard deviation, the present invention uses a hard threshold processing method to denoise the time domain scale coefficient, that is, the signal components with amplitudes below the threshold are directly set to zero, while the signal components with amplitudes above the threshold remain unchanged. This denoising method can remove small-amplitude noise interference while retaining the main vibration characteristics. The denoised time domain scale coefficient is smoother than the original signal, can more accurately reflect the true vibration characteristics of the road surface, and reduce the impact of random noise on subsequent feature extraction. After completing the noise removal, the present invention further uses the denoised time domain scale coefficient to calculate the local energy and normalize it to a probability distribution to construct the statistical characteristics of the vibration signal.

[0023] The calculation method of local energy is based on the sum of squares of time domain scale coefficients after denoising. This energy feature can effectively reflect the intensity changes of vibration signals on different time scales. In practical applications, road damage is usually manifested as abnormal increase or violent fluctuation of local vibration energy. Therefore, the present invention can effectively capture abnormal areas in vibration signals through the local energy calculation method, and provide quantitative feature description for subsequent abnormal detection. In order to further enhance the statistical characteristics of road vibration signals, the present invention normalizes the local energy and converts it into a probability distribution, so that vibration signals at different scales and different time windows can be compared and analyzed under a unified statistical framework. The normalized probability distribution not only eliminates the difference in amplitude of vibration signals, but also enhances the discriminability of road damage, so that different types of vibration modes can be distinguished by probability features. Compared with traditional vibration signal analysis methods, the present invention adopts the method of local energy normalization, which can effectively avoid feature distortion caused by changes in measurement conditions and improve the ability to identify road vibration anomalies. In addition, since the local energy feature is calculated based on the time domain scale coefficient after wavelet transformation, it has good time resolution and frequency resolution, and can simultaneously describe the local change characteristics of the vibration signal at different scales, thereby enhancing the detection capability of road damage. In combination with the method framework of the present invention, through the joint calculation of denoising and local energy normalization, it is not only possible to effectively suppress measurement noise and improve the signal-to-noise ratio of the signal, but also to construct a robust statistical feature of the road vibration signal, providing a high-precision feature description for subsequent anomaly detection.

[0024] Step 3: Perform spectral moment statistics on local energy in the local time window, calculate the first spectral moment and the second spectral moment, and obtain comprehensive spectral features; calculate the correlation coefficient by local cross integration at adjacent scales; Since road vibration signals usually have non-stationary and multi-scale characteristics, it is difficult to fully characterize their complexity with features at a single scale. The present invention calculates the distribution characteristics of local energy through spectral moment statistics, thereby capturing the frequency components of the signal and its change pattern at multiple scales, providing a more robust characterization method for road damage detection. In the local time window, the calculated local energy is first analyzed by spectral moment, wherein the first spectral moment describes the central position of the local energy, which can reflect the average energy distribution of the road vibration signal at different scales, while the second spectral moment describes the discrete degree of the local energy distribution, which can characterize the volatility and mutation degree of the vibration signal. The anomaly in the road vibration signal is often manifested as a drastic change in local energy in time or scale, so the joint calculation of the first spectral moment and the second spectral moment can effectively capture the change trend of the vibration signal and highlight the characteristics of abnormal vibration. Compared with the traditional time domain or frequency domain analysis method, the spectral moment statistics method can extract high-order statistical features of the signal based on wavelet transform, thereby more accurately reflecting the vibration mode changes caused by road damage and improving the sensitivity of damage detection.

[0025] In order to further enhance the characteristic analysis of road vibration signals, the present invention calculates local cross integrals at adjacent scales to quantify the correlation between different scales. Since wavelet transform provides the ability of multi-scale decomposition, the time domain scale coefficients at different scales characterize the vibration characteristics of different frequency components, and the correlation between adjacent scales can reveal the change pattern of vibration signals in multi-scale space. Abnormalities in road vibration signals often manifest as local mismatches or mutations between different scales. Therefore, calculating local cross integrals and analyzing the trend of changes in their correlation coefficients can effectively detect the existence and severity of road damage. The calculation method of local cross integral is based on the time domain scale coefficients after denoising. By calculating the dot product relationship of signals of different scales in the local time window, the similarity index between scales is obtained, thereby reflecting the dynamic change characteristics of vibration signals at different scales. Compared with traditional autocorrelation or cross-correlation methods, the local cross integral method of the present invention is calculated in the wavelet transform domain, which can make full use of the time localization characteristics of wavelet transform and improve the detection ability of transient abnormal signals. In addition, by performing cross-integration calculations at different scales, the ability to capture multi-frequency component vibration modes can be enhanced, making the method of the present invention more suitable for the detection of complex road conditions and various damage types in road vibration signal analysis.

[0026] Step 4: Based on the comprehensive spectrum characteristics and correlation coefficients, a global abnormality index is constructed to reflect the abnormal road vibration. The predefined baseline abnormality index and critical abnormality index are combined, and the hyperbolic tangent function is used to construct a nonlinear mapping to calculate the road damage probability.

[0027] Specifically, since the anomaly of road vibration signals is often closely related to factors such as structural damage, material fatigue, crack expansion or road deformation, it is crucial to accurately evaluate the degree of abnormality of vibration signals for road damage detection. The present invention first constructs a global abnormality index based on multi-scale analysis to quantify the overall abnormality of vibration signals in different time windows. The construction of the global abnormality index comprehensively considers multiple factors, including comprehensive spectral features calculated based on spectral moment statistics, correlation coefficients calculated at adjacent scales, and the change trend of local cross-integral. Through the combination of these features, the present invention can accurately characterize the abnormal mode of road vibration signals and distinguish between normal vibration and abnormal vibration. Traditional anomaly detection methods usually use a single feature for analysis, which is difficult to take into account the multi-scale characteristics and local change patterns of signals. The present invention realizes the fusion of different feature information through the construction of the global abnormality index, which can more comprehensively characterize the degree of abnormality of road vibration signals and improve the accuracy and adaptability of anomaly detection. After the global abnormality index is constructed, the present invention further combines the predefined baseline abnormality index and critical abnormality index to establish a hierarchical evaluation system for abnormal signals.

[0028] The baseline anomaly index is used to characterize the vibration characteristics under normal road conditions and serves as a benchmark for normal vibration signals, while the critical anomaly index is used to define the vibration mode with obvious abnormalities to identify situations where damage or structural defects may exist. In practical applications, the baseline anomaly index is usually determined based on the statistical characteristics of a large amount of normal road vibration data, while the critical anomaly index is set based on laboratory test data or engineering experience to ensure that it can effectively distinguish between damage signals and normal signals. Based on the determination of the baseline anomaly index and the critical anomaly index, the present invention uses a hyperbolic tangent function for nonlinear mapping to calculate the probability of road damage. Since the relationship between the abnormality degree of the road vibration signal and the probability of damage is often nonlinear, directly using a linear method for calculation may result in a low degree of differentiation of the probability of damage, and it is difficult to accurately reflect the changes in road conditions. Therefore, the present invention introduces a hyperbolic tangent function to enhance the sensitivity of anomaly detection through nonlinear transformation. The characteristics of the hyperbolic tangent function make the abnormal index change more slowly when it is close to the baseline anomaly index, and change steeply when it is close to the critical anomaly index. The advantage of this nonlinear mapping is that it can more clearly divide the normal state, the slightly abnormal state and the seriously abnormal state, thereby improving the accuracy of damage assessment. In addition, the smoothness of the hyperbolic tangent function can ensure the continuity of the calculation of the damage probability under different abnormality levels, avoiding the drastic change of the damage probability due to the slight fluctuation of the abnormal index, thereby improving the stability of the damage assessment. Through this method, the present invention makes the calculation of the road damage probability more in line with engineering practice and can maintain high reliability and adaptability under different types of road environments.

[0029] Example 2: In step 1, sensors are deployed on the target road at a set sampling interval. , collect the original road time series vibration signal of the target road; suppose the collected original road time series vibration signal is , using a bandpass filter for preprocessing, bandpass filter is a Gaussian bandpass kernel, expressed as follows: ; in, is the filter width parameter; is the center frequency, corresponding to the main frequency band of road vibration; is a time variable; the filtered signal after filtering is ;in is the total signal acquisition time; is the time integral variable; through the following formula, Perform normalization to obtain a standardized signal: ; in, is a standardized signal.

[0030] Specifically, the essence of the road vibration signal is a time-varying random process, which is affected by vehicle driving, road surface characteristics, tire contact force and environmental noise. It contains both low-frequency overall vibration components and high-frequency vibration components caused by factors such as minor road damage, cracks, potholes, etc. Since the signal contains multiple frequency components, direct analysis of the original signal may be interfered by low-frequency trend terms or high-frequency noise. Therefore, the present invention uses a bandpass filter for preprocessing to enhance the signal within the target frequency band while suppressing irrelevant low-frequency or high-frequency components. The design of the bandpass filter adopts the form of a Gaussian bandpass kernel, that is, the mathematical expression of the filter includes the product of a Gaussian function and a sine term, where the Gaussian function partially determines the distribution shape of the filter in the time domain, and the sine term is used to select a specific center frequency component. The standard deviation of the Gaussian function Determines the time expansion range of the filter, the smaller The value makes the filter have a higher time resolution in the time domain, while a larger The value enhances the selectivity of specific frequency components, thereby ensuring that the filtered signal can accurately retain the target vibration frequency information. Corresponding to the main frequency band of road vibration, it ensures that the filtered signal mainly contains frequency components related to the health of the road structure without being affected by low-frequency trends or high-frequency noise. By the original signal With bandpass kernel It is obtained by integrating calculations over time, and the integration operation is completed within the entire signal time range, thereby realizing bandpass filtering processing of the road vibration signal. After filtering, the signal may still have amplitude drift or energy level fluctuations introduced by the measuring equipment. Therefore, the present invention further normalizes the filtered signal to improve the numerical stability of the signal and ensure the scale consistency of the signal under different measurement conditions. The normalization process first calculates the mean of the filtered signal and subtracts it from the signal to remove the DC component of the signal so that the mean of the normalized signal is zero. The standard deviation of the signal is then calculated and normalized using the standard deviation to ensure that the amplitude distribution of the signal has a fixed standard scale, thereby enhancing the comparability and processing stability of the signal. The normalized signal With zero mean and unit standard deviation, subsequent wavelet analysis can be calculated on a consistent numerical scale, avoiding instability in feature extraction caused by differences in the sensitivity of the measuring equipment or changes in the vibration environment. Compared with traditional signal preprocessing methods, the present invention combines the optimization design of the bandpass filter with normalization processing to retain the effective information of the road vibration signal to the greatest extent, and can still maintain high accuracy and stability under different measurement environments, thereby providing reliable input data for subsequent analysis based on multi-scale wavelet transform. In addition, the filter parameters and The adjustability enables the present invention to adapt to different types of road structures and vibration environments, and achieve more accurate vibration signal extraction.

[0031] Embodiment 3: In step 1, the standardized signal is subjected to attenuated weighted continuous wavelet transform to obtain the time domain scaling coefficient : ; in, , is the scale parameter; , is the translation parameter; is the mother wavelet function, ; is the imaginary number symbol; is the time domain scaling factor.

[0032] Specifically, wavelet transform is a mathematical tool that can provide both time and frequency information. Compared with the traditional Fourier transform, it can analyze the instantaneous characteristics of signals at different time scales, and is particularly suitable for data with non-stationary and multi-scale characteristics such as road vibration signals. After wavelet transform, the time domain scaling coefficient is obtained Describes the signal at different scales and different time locations The local characteristics under the vibration signal are obtained, thereby providing an in-depth analysis of the multi-scale dynamics of the vibration signal. A complex Gaussian wavelet is used, which consists of the product of two exponential functions, one of which is used to control the center frequency of the wavelet, while the other is used to ensure the localization characteristics of the wavelet. The complex Gaussian wavelet can maintain high temporal resolution at different scales and has good frequency selectivity, making it particularly suitable for analyzing short-term transient features and complex patterns in road vibration signals. Scale parameter Control the expansion and contraction of the wavelet, that is, when When the value is small, the wavelet function appears as a narrow wave packet in the time domain, which is suitable for extracting high-frequency components in the signal. When the value is large, the wavelet function expands in the time domain and can extract the low-frequency components of the signal. To ensure that different vibration modes from high-frequency details to low-frequency trends can be covered, adapting to signal characteristics under different road conditions.

[0033] In addition, in order to further optimize the characteristics of wavelet transform, the present invention introduces an exponential decay factor To enhance the sensitivity of the time domain scale coefficient to local changes in the signal. The role of this item is to calculate the wavelet transform for the signal far away from the current time position. The signal components are exponentially decayed, so that the wavelet transform pays more attention to local features, reduces the interference of long-distance signals on the current scale calculation, and improves the time localization ability of the signal. The parameters in the exponential decay factor are Controls the strength of the attenuation, larger The value means stronger suppression of signals far from the center point, so that the wavelet transform focuses more on the short-term variation characteristics of the road vibration signal. This design is of great significance for road damage detection, because road damage usually manifests as short-term sudden vibration abnormalities, and exponential decay can effectively highlight these instantaneous characteristics and improve the distinguishability of damage characteristics. The continuous wavelet transform calculation formula of the present invention also includes a normalization factor , which is used to maintain the energy consistency of the time domain scale coefficient at different scales, prevent numerical imbalance caused by scale changes, and ensure that the features calculated at different scales are comparable. In traditional wavelet analysis methods, the signal energy at different scales often changes with the The present invention uses normalization processing to make the wavelet coefficients maintain a relatively balanced amplitude level within different scale ranges, thereby improving the robustness and consistency of the vibration characteristics.

[0034] Embodiment 4: In step 2, in order to adaptively suppress noise interference in the time domain, at each scale and local time window The local energy statistics are used to estimate the noise level and calculate the local noise standard deviation. for: ; in, is the local time window width; is the median function; the local adaptive threshold is defined by the following formula: ; in, is the sampling interval, reflecting the number of local samples, and the hard threshold strategy is used for denoising to obtain the denoised coefficient : ; in, It is an absolute value operation.

[0035] Specifically, in the method of the present invention, in order to perform adaptive noise suppression in the time domain, it is first necessary to and local time window The noise level is estimated by local energy statistics estimation. The local noise standard deviation is determined by calculating the median of the time domain scaling coefficient within the time window. . Different from the traditional noise estimation method based on mean and variance, the present invention uses the median function to calculate the noise level. Its main advantage is that it is more robust to extreme values ​​and can effectively avoid the influence of outliers. Since noise usually follows a symmetrical distribution, and the signal part may have a large amplitude variation, the use of the median to estimate the noise standard deviation can more accurately reflect the true level of local noise without being disturbed by individual outliers. In addition, in order to ensure that the calculation of the noise standard deviation has good numerical stability, the present invention uses a normalization factor Scale adjustment is performed. This factor is derived from the statistical characteristics of the Gaussian distribution and can ensure that the calculated noise standard deviation remains consistent across different scales and time windows, thereby improving the robustness and generalization ability of the denoising algorithm.

[0036] After completing the estimation of the noise standard deviation, the present invention further defines a local adaptive threshold , the threshold is dynamically adjusted based on the noise level to ensure that the optimal denoising threshold can be set according to the noise characteristics of the signal itself at different scales and time windows. The calculation formula of the adaptive threshold contains a logarithmic term, which is used to describe the time sampling relationship of the signal at different scales, so that the threshold can be adjusted as the scale changes, thereby enhancing the adaptability of the denoising strategy to signals of different scales. Since different frequency components of the road vibration signal have different energy levels, a fixed global threshold may cause signal loss at certain scales. Therefore, the present invention adopts an adaptive threshold calculation method, so that the denoising strategy can dynamically adapt to the signal characteristics of different scales and improve the accuracy of signal denoising. Specifically, the threshold calculation method not only takes into account the local noise standard deviation , combined with the temporal resolution of the signal sampling And the total duration , thus ensuring that the denoising method still has stable performance under different sampling rates and signal durations.

[0037] After determining the local adaptive threshold, the present invention adopts the hard threshold strategy to denoise the time domain scale coefficients. The basic idea of ​​the hard threshold method is to directly set the coefficients with amplitudes lower than the threshold to zero, while retaining the coefficients with larger amplitudes. This strategy can effectively remove noise interference while retaining the main features of the signal as much as possible. Specifically, if Less than the local threshold , then the coefficient is set to zero to remove low-energy noise components, and if If it is greater than or equal to the threshold, the original value is maintained to ensure that the important features of the signal are not lost. Compared with the traditional soft threshold method, the hard threshold denoising strategy does not scale the high-energy signal components during the denoising process, so it can better retain the amplitude information of the signal, so that the denoised signal still maintains the integrity of the original vibration characteristics. In addition, the hard threshold denoising method has a small amount of calculation, is suitable for real-time processing of large-scale signal data, and can meet the engineering application requirements of the present invention in road vibration signal analysis. Through the above method, the present invention realizes adaptive noise suppression in the time domain scale domain, can accurately remove noise interference within different scales and time windows, and ensures that the key features of the signal are effectively retained.

[0038] Example 5: In step 2, the local energy distribution is calculated using the denoised wavelet coefficients: ; To eliminate the amplitude effect caused by scale and time domain integration, the energy is normalized to a probability distribution : ; in, is the scale integral variable; is the translation integral variable.

[0039] Specifically, the present invention calculates the local energy distribution through the denoised wavelet coefficients, that is, by taking the square of the wavelet coefficients, the signal at different scales is obtained. and time location The local energy on the road vibration signal. This calculation method can effectively characterize the local intensity of the signal and highlight the abnormal patterns in the road vibration signal. The calculation principle of local energy is that the wavelet transform provides a time-frequency representation of the signal at different scales, and the square of the wavelet coefficient reflects the energy density of the signal in a specific scale and time window, so it can be used to quantify the local characteristics of the road vibration signal. Larger local energy usually corresponds to a mutation point in the signal, and the mutation point is often closely related to road damage, so the calculation of local energy is crucial for road damage detection. However, since the amplitudes of wavelet coefficients at different scales may vary greatly, if the unnormalized energy value is directly used for analysis, it may cause the features at certain scales to be amplified or compressed, thereby affecting the accuracy of anomaly detection. Therefore, the present invention further adopts a normalization method to convert the local energy into a probability distribution to eliminate the amplitude influence caused by scale and time domain integration. The normalization process calculates the sum of energy in all scales and time ranges and divides the local energy by the sum, so that the normalized energy features numerically meet the constraints of probability distribution, that is, the sum of normalized energy in all scales and time windows is 1, thereby ensuring the comparability of signal features at different scales and in different measurement environments. The advantage of this normalization method is that it can not only balance the energy distribution at different scales, making the energy features at all scales comparable within the same numerical range, but also enhance the robustness of the signal and avoid feature distortion caused by changes in measurement conditions or uneven signal amplitudes. In addition, the normalized energy distribution can be used as a probability density function for subsequent statistical analysis and anomaly detection to improve the accuracy and stability of road damage identification.

[0040] Embodiment 6: In step 3, the spectral moment statistics of the denoised energy distribution are performed in the local time window to characterize the center and expansion of the energy distribution, thereby reflecting the local structural state; set the scale Next, in the time window The first spectral moment is defined internally and the second spectral moment for: ; in, is the Poisson’s ratio of the road material of the target road; is the road material density of the target road; is the road material damping coefficient of the target road; is the pavement layer thickness of the target road; ; in, is the foundation rebound modulus of the target road; is the foundation damping coefficient of the target road; is the foundation stiffness of the target road.

[0041] Specifically, the present invention is in scale Time window Within the range, the energy distribution A weighted integration is performed, and the physical parameters of the road material and foundation are incorporated into the integration process, so that the spectral moment not only reflects the statistical properties of the signal itself, but also reflects the influence of the target road structure characteristics. It can be understood as the central position of the local energy distribution. Its calculation method includes the difference between two groups of molecules and two groups of denominators, where some molecules are , which is used to measure the energy over time. The center of gravity of the change; while the other part of the molecules These expressions convert the damping coefficient and pavement layer thickness Taking into account, when weighting the energy center, attention is paid to the influence of road surface attenuation and road material thickness on vibration propagation. The denominator contains Arithmetic structure, the reason for introducing Poisson's ratio and density , because these two items will affect the deformation and mass distribution characteristics of road materials after being loaded by vehicles. When there is a significant difference between the two, it indicates that there may be signs of structural abnormality or damage in the time window. It is not only a simple geometric center of energy, but can also be regarded as a comprehensive indicator that integrates material mechanics and dynamics factors, thus providing a more physically meaningful numerical measurement of the local center of road vibration response.

[0042] At the same time, the second spectral moment It is used to describe the discreteness of the energy distribution relative to the first spectral moment, which is included in the integral of the numerator The structure is used to measure the fluctuation amplitude of the signal in the local window. If the energy distribution is Highly concentrated within the range, The integral result of one item will be small, indicating that the signal has no significant diffusion or multi-peak distribution; but if the energy distribution is very wide, or there are multiple energy peaks locally, then the integral result will increase accordingly, indicating that the local vibration mode is more complex. More noteworthy is that The denominator of Projects such as , foundation damping coefficient and foundation stiffness , these parameters together reflect the interaction characteristics between the target road and its foundation: on the one hand, the energy transfer of the road structure when subjected to vehicle loads or environmental disturbances will be restricted by the foundation stiffness and damping; on the other hand, an excessively large rebound modulus or an unmatched stiffness coefficient may also cause the road surface to produce a large amplitude vibration response. By combining and The difference between and can numerically balance the influence of foundation characteristics on the spectral moment. A sudden change or abnormal increase in this balance relationship often means that there is an abnormality in the interaction between the road and the foundation, such as insufficient support of the local foundation, local hardening or loosening of the road material, etc., which allows the present invention to have a keener insight in identifying road damage. In other words, It is not just a statistical variance measurement, but also coupled with the mechanical properties of the foundation-pavement system to map the numerical change of the discreteness to the real road health status. It is worth emphasizing that the local time window used in the present invention With scale The combination of is based on the multi-scale wavelet analysis framework to select the appropriate time length and frequency scale, so that the energy distribution can be statistically analyzed in a reasonable local range without losing key features, and also avoids the interference of noise or irrelevant vibration in a too wide range. and The present invention can capture the evolution of road vibration energy center and dispersion in multi-scale and local time domain space, and incorporate the dynamic parameters of materials and foundation into the spectral moment calculation, which greatly improves the sensitivity of the method to potential damage and anomalies. When the present invention is deployed in actual projects, by performing time series analysis and comparison of spectral moment results of different scales and time windows, it can quickly find that the spectral moment values ​​of the road surface or foundation in certain areas have significantly deviated. Once and Abnormal increases or large fluctuations often indicate damage to the road structure or potential risk of destruction, thus providing maintenance departments with quantitative and detailed early warning information.

[0043] Example 7: In step 3, the comprehensive spectrum feature is defined by the following formula: : ; in, is the fatigue limit stress of the road material of the target road; is the elastic modulus of the road material of the target road.

[0044] Specifically, the first It directly inherits the definition of the spectral moment statistics mentioned above and describes the center of gravity of the vibration energy in the local time window. It is calculated based on the energy distribution after denoising, so it can not only reflect the instantaneous energy peak position of the signal, but also combine factors such as the Poisson's ratio, density, damping coefficient and foundation stiffness of the road material to provide a more physically meaningful signal center feature. It may correspond to the location of the main peaks of the local vibration signal, and the drift or abnormal fluctuation of these peaks is often closely related to the evolution of road damage. For example, if the elastic deformation of the local road structure increases due to fatigue effects, the energy center may shift systematically. There will be a trend change within a certain range, so the present invention uses it as the basic item of the comprehensive spectrum feature. It is mainly used to quantify the fatigue effect of road materials in vibration signals. First, As the second spectral moment, it measures the expansion of local energy distribution and reflects the degree of discreteness or unevenness of the vibration signal in the local time window. When the road structure is in normal operation, the distribution of vibration energy is often more concentrated, that is, the time-scale joint distribution of the signal will not show too much deviation. However, when the road microstructure changes due to external forces or material aging, the vibration response on the local scale tends to diffuse, which is manifested as Therefore, the present invention is based on The comprehensive spectrum features are calculated so that the subtle changes in road damage can be reflected through the variance effect of the spectral moment.

[0045] In order to map this statistical property to the physical properties of road materials, the present invention introduces This ratio. Fatigue limit stress It is the maximum stress level at which a material can withstand cyclic loading without breaking, while the elastic modulus describes the stress-strain relationship of the material within a small deformation range. It can be regarded as the deformation sensitivity of the material under fatigue limit conditions, which describes the material's response ability to external vibration loading. A higher ratio means that the material will deform more at the fatigue limit, while a lower ratio means that the material is more rigid at the same stress level. Used as a weighting factor to adjust The influence of makes the comprehensive spectrum feature not only characterize the energy discreteness of the vibration signal, but also quantify the dynamic response of the road material under fatigue. The physical meaning of this formula is that when fatigue damage occurs in the road structure, will increase, indicating that the energy distribution of the local vibration signal is more discrete, which leads to Increase; at the same time, if the fatigue limit stress of the road material is low or the elastic modulus is high, If the ratio is small, the change range of the comprehensive spectrum characteristics is relatively small, which means that the material still has a certain fatigue resistance; but if the fatigue limit of the material is low and the elastic modulus is also low, then The ratio is large, The influence of is amplified, indicating that the material is prone to large deformation under fatigue conditions, which may cause road damage. The construction makes the road vibration signal analysis not only limited to the time-frequency characteristics of the signal itself, but also can combine the fatigue characteristics of the material to form a multi-level fusion analysis from signal characteristics to physical and mechanical characteristics. Compared with the traditional vibration signal analysis method, the present invention fully utilizes the multi-scale advantages of wavelet transform when constructing comprehensive spectral characteristics, and combines the key mechanical parameters of the material, so that the road damage analysis is no longer solely dependent on statistical signal characteristics, but further considers the fatigue behavior and mechanical response characteristics of the material. This method can identify the fatigue damage trend of the road at an early stage and provide a more scientific basis for road maintenance decisions. For example, in practical applications, if a certain scale and local time window are observed If the damage is gradually increasing, it may mean that the material in this area has entered the fatigue damage stage. At this time, the development trend of the damage can be further confirmed by performing time series analysis on the spectral features of multiple scales, and maintenance measures can be taken in time. In addition, since the comprehensive spectral features are calculated based on wavelet coefficients, they have a high time resolution and can track early changes in road damage within a short time window, thus improving the real-time performance of road health monitoring.

[0046] Embodiment 8: In step 3, the correlation coefficient is calculated by local cross integration at adjacent scales using the following formula: ; in, is the correlation coefficient.

[0047] Specifically, the correlation coefficient The calculation is based on the time domain scale coefficient after denoising and The local cross-integral between adjacent scales and The numerator of the calculation formula is Through the time window The signal within is integrated to calculate the joint energy contribution of adjacent scales, thereby measuring the intrinsic correlation of signals at different scales. If the integral result is large, it indicates that the signal characteristics at the two scales have strong similarities, that is, in the multi-scale space, the local structure of the vibration signal has stable transmission in different frequency bands. When the integral approaches zero or even a negative value, it means that the energy characteristics of the signal at different scales show large differences, and there may be mutations in the local structure, such as nonlinear effects caused by road damage, crack extension or foundation anomalies. In order to ensure that the correlation of signals at different scales has a standardized measurement method, the present invention introduces a normalization factor in the denominator of the correlation coefficient calculation. , which consists of the square root of the product of the local energies of the two scaled signals and acts like covariance normalization to ensure that the calculated correlation coefficient is This normalization method can effectively eliminate the influence of signal amplitude, so that the calculation result of correlation coefficient depends only on the similarity of signal structure and will not be affected by the change of local signal energy.

[0048] In practical applications, It can be used to detect the stability of road vibration signals at different scales and provide important auxiliary information for anomaly detection. Generally speaking, a healthy road structure should maintain a high correlation at multiple scales, that is, the characteristics of the vibration signal should be consistent within different frequency component ranges. However, when the road structure is damaged, due to local cracks, potholes or material fatigue, the energy distribution of the vibration signal at adjacent scales may mutate, causing This downward trend can be regarded as one of the important indicators for early detection of damage. For example, within a certain scale range, the vibration signal may remain stable at a lower frequency, but show greater discontinuity at a higher frequency scale, which often means that the road structure has already suffered microscopic cracks or local damage, but has not yet developed to the macroscopic visible destruction stage. Therefore, the present invention calculates the correlation coefficient through local cross-integration, which can track the dynamic evolution of the road structure in a multi-scale space, providing a scientific basis for early damage warning. In addition, The calculation method is applicable to various types of road structures, whether it is concrete pavement or asphalt pavement. The method can adjust the scale range according to the vibration characteristics of different materials. Compared with the traditional single-scale signal analysis method, the present invention utilizes the multi-scale characteristics of wavelet transform and calculates the correlation between adjacent scales, which can not only improve the sensitivity to signal pattern changes, but also enhance the adaptability of the algorithm in complex environments, making the road damage detection results more robust.

[0049] Example 9: Construct a global abnormality index using the following formula: ; in, It is the global abnormality indicator.

[0050] Specifically, in this formula, the integral range covers the scale and time window , where the integral variable represents the spectral characteristics of the signal at different scales, and Represents the evolution trajectory of the vibration signal in the time domain. Global abnormality index The molecular part contains ,in is the abnormal sensitivity factor, which is used as a weighting coefficient to enhance the contribution of the local abnormal area of ​​the signal. Adaptive settings can be made through data-driven methods, such as training anomaly detection models based on historical vibration data, or optimizing them in combination with prior knowledge of road structure damage to improve the adaptability to different types of road damage. As the main feature input, the energy distribution characteristics of the vibration signal in the local scale and time window are quantified, and the ratio of the material fatigue limit stress to the elastic modulus is combined to reflect the dynamic response of the road material under fatigue limit. The larger the value, the more abnormal the distribution of local vibration energy of the road, which may cause structural damage, crack extension or uneven foundation settlement. Therefore, it plays a key role in the calculation of global abnormal indicators.

[0051] Exponential decay term Mainly used to adjust the time sensitivity of abnormal indicators, where Represents the main frequency characteristics of the vibration signal, Reflects the current time window The deviation from the main frequency position of the signal. near When , the exponential term approaches 1, indicating that the vibration signal contributes more at this time; when keep away When , the exponential term tends to zero, reducing the impact of this time window on the global abnormal index. This attenuation mechanism can effectively eliminate the interference of short-term abnormal fluctuations on the overall abnormal evaluation, ensuring that the abnormal index can accurately reflect the long-term trend. In addition, the denominator of the exponential term As fatigue limit stress, it plays a normalization role, so that road structures of different materials can maintain stable calculation results under the same formula. The denominator of the global abnormal index contains ,in is the correlation coefficient between adjacent scales. The value range is Therefore, the denominator The value range is , ensuring that the denominator does not approach zero and avoiding the problem of unstable calculation. The introduction of the correlation coefficient is mainly to constrain the multi-scale consistency of the vibration signal. If the vibration signals in a certain area have high similarity at different scales, then If the correlation between adjacent scales is low, it means that there may be mutation signals or structural abnormalities in the region, and the denominator tends to 1, which amplifies the abnormal contribution of the region. Therefore, this normalization strategy can adaptively adjust the abnormal contribution of signals at different scales, avoid the excessive influence of a single scale signal on the overall abnormal index, and improve the sensitivity of the global abnormal index to road damage. In a physical sense, the global abnormal index It reflects the comprehensive abnormality of the entire road vibration signal at different scales and time windows. The larger the value, the more serious damage or degradation the road structure may have. Compared with the traditional anomaly detection method, the method of the present invention not only combines the time-frequency characteristics of multi-scale wavelet analysis, but also introduces material mechanical parameters and fatigue characteristics, so that the anomaly index has a stronger physical interpretation. For example, in practical applications, if Abnormal rise in certain time periods may mean that the road has experienced greater dynamic stress during that period, leading to fatigue accumulation and micro crack growth; If the value remains high for a long time in a certain local area, it may indicate that the structure in that area has been significantly damaged, such as local collapse, crack expansion or uneven foundation settlement. Analyzing the evolution trend in time and space can achieve early warning of road damage and provide a scientific basis for road maintenance decisions.

[0052] Example 10: Define the baseline abnormality index by the following formula : ; in, for The mean of for The critical abnormality index is defined by the following formula: : ; in, is the local settlement of the target road; the road damage probability is calculated by the following formula: ; in, is the road damage probability.

[0053] Specifically, baseline abnormal indicators The construction of is to provide a normal state benchmark for road vibration signals for comparison in subsequent anomaly detection. Its calculation formula is similar to the global anomaly index With a similar integral structure, still on the scale and time window However, unlike the global anomaly index, the baseline anomaly index uses local energy and comprehensive spectral features The mean value of and These means represent the long-term average energy distribution and characteristic trend of the road vibration signal, thus constructing a reference signal to measure the normal vibration state. Since the vibration state of a normal road is usually affected by the road material, foundation conditions and traffic load, the present invention extracts the steady-state characteristics of the long-term operation of the road through mean calculation, avoiding the influence of short-term fluctuations or local anomalies on the baseline anomaly index. In the calculation formula of the baseline anomaly index, the exponential decay term is still included. , which acts like the exponential term in the global anomaly index, but its frequency center position is set to , which is half of the main frequency of the vibration signal. This adjustment is mainly based on experience in engineering practice. The normal vibration modes of many road structures often have lower frequency components, while damage signals usually appear in a higher frequency range. Therefore, the construction of the baseline abnormality index pays more attention to the low-frequency vibration characteristics to ensure that it can truly reflect the vibration mode under normal operating conditions and avoid the contamination of the baseline value by abnormal vibration. In addition, the denominator of the index term The decay rate is further reduced, making the baseline anomaly index smoother in the time domain and reducing the impact of noise on the calculation results.

[0054] After determining the baseline abnormality index, the present invention further defines the critical abnormality index , which is used to characterize the critical point at which the road enters a damaged state. Its calculation method is based on the baseline abnormality index A correction term is added, which is determined by the fatigue limit stress of the road material , elastic modulus and local settlement Jointly determine. Local settlement It represents the deformation degree of the road structure under long-term load, which is usually related to uneven settlement of the foundation, material fatigue or structural damage. When it is larger, it means that the road may have a more serious deformation, which will increase the overall abnormality of the vibration signal. This correction term directly maps the mechanical deformation state of the road to the anomaly index, so that the critical anomaly index can flexibly adapt to different road structures and damage conditions. Finally, the present invention calculates the road damage probability based on the baseline anomaly index and the critical anomaly index. , to quantify the health status of the road structure. The damage probability is calculated using the hyperbolic tangent function ( ) to enhance the resolution of damage extent. Represents the deviation of the global abnormal index relative to the baseline abnormal index. The larger the deviation, the higher the degree of abnormal vibration of the road. is used for normalization so that the damage probability is always The characteristic of the hyperbolic tangent function is that when the value is small, the output changes slowly, and when it is close to the critical value, the change is steep, thus ensuring that the damage probability can rise rapidly when it is close to the damage threshold, and improving the sensitivity to the damage state. Compared with the traditional linear method, this nonlinear mapping method has stronger classification ability, can reduce false alarms when the damage probability is small, and enhance the recognition ability when the damage probability is close to 1. In a physical sense, the damage probability It reflects the current health status of the road structure and combines the abnormality of the vibration signal, the fatigue characteristics of the material and the deformation of the foundation. If the value is below a certain threshold (e.g. 0.3), the road can be considered to be in a healthy state and does not require additional maintenance. When it is at a medium level (such as between 0.3 and 0.7), it indicates that the road may be slightly damaged and needs further monitoring; When it is higher than 0.7, it may mean that the road structure has entered a serious damage stage, and it is recommended to repair or reinforce it. The method of the present invention can effectively quantify the health status of the road and provide a damage assessment mechanism based on physical parameters, providing scientific and reliable decision support for road structure health monitoring.

[0055] The present invention is not limited to the above-mentioned specific embodiments, but extends to any new features or any new combination disclosed in this specification, as well as any new method or process steps or any new combination disclosed.

Claims

1. A road vibration signal analysis method based on multi-scale wavelet transform, characterized in that: The method comprises: Step 1: Perform convolution filtering on the original road time series vibration signal using a Gaussian bandpass filter kernel, obtain the filtered signal and then normalize it to form a standardized signal; use the attenuation weighted continuous wavelet transform on the standardized signal to obtain the time domain scale coefficient to capture the local changes of the standardized signal at different scales and time domains; Step 2: According to the time domain scale coefficient, the noise standard deviation is calculated and the threshold is set based on the median statistics in each local time window, and the denoised time domain scale coefficient is obtained through hard threshold processing; the local energy is calculated using the denoised time domain scale coefficient and normalized to a probability distribution; Step 3: Perform spectral moment statistics on local energy in the local time window, calculate the first spectral moment and the second spectral moment, and obtain comprehensive spectral features; calculate the correlation coefficient by local cross integration at adjacent scales; Step 4: Based on the comprehensive spectrum characteristics and correlation coefficients, a global abnormality index is constructed to reflect the abnormal road vibration. The predefined baseline abnormality index and critical abnormality index are combined, and the hyperbolic tangent function is used to construct a nonlinear mapping to calculate the road damage probability.

2. The road vibration signal analysis method based on multi-scale wavelet transform as claimed in claim 1, characterized in that: In step 1, sensors are deployed on the target road to collect data at a set sampling interval. , collect the original road time series vibration signal of the target road; suppose the collected original road time series vibration signal is , using a bandpass filter for preprocessing, bandpass filter is a Gaussian bandpass kernel, expressed as follows: ； in, is the filter width parameter; is the center frequency, corresponding to the main frequency band of road vibration; is a time variable; the filtered signal after filtering is ;in is the total signal acquisition time; is the time integral variable; through the following formula, Perform normalization to obtain a standardized signal: ； in, is a standardized signal.

3. The road vibration signal analysis method based on multi-scale wavelet transform as claimed in claim 2, characterized in that: In step 1, the time domain scaling coefficient is obtained by using the attenuation weighted continuous wavelet transform on the standardized signal : ； in, , is the scale parameter; , is the translation parameter; is the mother wavelet function, ; is the imaginary number symbol; is the time domain scaling factor.

4. The road vibration signal analysis method based on multi-scale wavelet transform as claimed in claim 3, characterized in that: In step 2, in order to adaptively suppress noise interference in the time domain, at each scale and local time window The local energy statistics are used to estimate the noise level and calculate the local noise standard deviation. for: ； in, is the local time window width; is the median function; the local adaptive threshold is defined by the following formula: ； in, is the sampling interval, reflecting the number of local samples, and the hard threshold strategy is used for denoising to obtain the denoised coefficient : ； in, It is an absolute value operation.

5. The road vibration signal analysis method based on multi-scale wavelet transform as claimed in claim 4, characterized in that: In step 2, the local energy distribution is calculated using the denoised wavelet coefficients: ; To eliminate the amplitude effect caused by scale and time domain integration, the energy is normalized to a probability distribution : ； in, is the scale integral variable; is the translation integral variable.

6. The road vibration signal analysis method based on multi-scale wavelet transform as claimed in claim 5, characterized in that: In step 3, the spectral moment statistics of the denoised energy distribution are performed in the local time window to characterize the center and extension of the energy distribution, thereby reflecting the local structural state; Set on scale Next, in the time window The first spectral moment is defined internally and the second spectral moment for: ； in, is the Poisson’s ratio of the road material of the target road; is the road material density of the target road; is the road material damping coefficient of the target road; is the pavement layer thickness of the target road; ； in, is the foundation rebound modulus of the target road; is the foundation damping coefficient of the target road; is the foundation stiffness of the target road.

7. The road vibration signal analysis method based on multi-scale wavelet transform as claimed in claim 6, characterized in that: In step 3, the comprehensive spectrum characteristics are defined by the following formula: : ； in, is the fatigue limit stress of the road material of the target road; is the elastic modulus of the road material of the target road.

8. The road vibration signal analysis method based on multi-scale wavelet transform as claimed in claim 7, characterized in that: In step 3, the correlation coefficient is calculated by local cross integration at adjacent scales using the following formula: ； in, is the correlation coefficient.

9. The road vibration signal analysis method based on multi-scale wavelet transform as claimed in claim 8, characterized in that: The global anomaly index is constructed using the following formula: ； in, It is the global abnormality indicator.

10. The road vibration signal analysis method based on multi-scale wavelet transform as claimed in claim 9, characterized in that: The baseline abnormality index is defined by the following formula: : ； in, for The mean of for The critical abnormality index is defined by the following formula: : ； in, is the local settlement of the target road; the road damage probability is calculated by the following formula: ； in, is the road damage probability.

Images