Non-destructive testing method and apparatus for internal voids in structures based on acoustic spectrum analysis

Abstract

This invention relates to the field of nondestructive testing (NDT) technology and provides a method and apparatus for NDT of internal voids in structures based on acoustic spectrum analysis. The method includes the following steps: time-frequency decomposition of multi-channel time-domain response signals to extract spectral characteristic parameters of each channel, and acoustic compensation correction based on coating thickness; extraction of abnormal features from surface temperature field data to obtain temperature anomaly characteristic parameters, and thermal compensation correction based on coating thickness; prediction of the geometric parameters of void defects, and simultaneous orthogonal joint analysis to generate a true defect confidence score. This invention eliminates coating interference by introducing coating thickness compensation and incorporates a nonlinear inversion model improved based on the physical laws of bridge structures, achieving a correlation between spectral characteristics and the geometric parameters of void defects. Simultaneously, the orthogonal joint analysis effectively suppresses false defect misjudgments, significantly improving the accuracy and reliability of NDT of internal voids in bridge structures.

Description

Technical Field

[0001] This invention belongs to the field of non-destructive testing technology, and particularly relates to a method and apparatus for non-destructive testing of internal voids in structures based on acoustic spectrum analysis. Background Technology

[0002] During long-term service, long-span tied-arch bridges are prone to developing voids within their steel box girder structures due to the combined effects of welding defects, cyclic fatigue loads, and environmental corrosion. These voids significantly weaken the effective load-bearing cross-section of the structure and may lead to fracture accidents under extreme conditions. Therefore, high-precision and high-reliability non-destructive testing of internal voids in steel box girder structures has significant engineering value and is of urgent practical importance.

[0003] Currently, commonly used non-destructive testing methods in engineering practice include ultrasonic transmission, ground-penetrating radar, and acoustic spectrum analysis. Among these, acoustic spectrum analysis has received widespread attention due to its ease of operation and relatively controllable cost. For example, CN120870345A discloses a non-destructive testing method and system for underground cavities using a mobile terminal. This method generates an acoustic signal of a preset frequency via the mobile terminal, analyzes the spectrum of the acoustic signal, extracts spectral features, and performs resonance frequency shift detection and cavity depth estimation based on these spectral features.

[0004] However, existing acoustic spectrum analysis methods face several unresolved technical problems in practical applications: First, there is a lack of a physical model that quantitatively correlates spectral characteristics with the physical size of the cavity. Most existing methods only make qualitative judgments by comparing the spectral frequency shifts with and without defects, failing to establish a quantitative correlation model between spectral characteristic parameters and the physical size of the defect. For example, CN120870345A only involves estimating the cavity depth based on the resonant frequency shift, without addressing the quantitative inversion of equivalent diameter and shape parameters, nor establishing a nonlinear physical mapping relationship between spectral characteristics and cavity geometric parameters.

[0005] Secondly, acoustic analysis alone is severely inadequate in identifying false defects. The surface conditions of long-span tied-arch steel box girders are extremely complex, with localized variations in steel plate thickness, residual stress concentration zones, and even surface coating peeling being common. These false defects, under acoustic excitation, may produce spectral response characteristics similar to real voids. However, traditional acoustic spectrum analysis lacks cross-validation with other physical field information, making it highly susceptible to false positives. This deficiency severely restricts the practical application of acoustic spectrum analysis in bridge engineering. Summary of the Invention

[0006] The purpose of this invention is to provide a method and apparatus for non-destructive testing of internal voids in structures based on acoustic spectrum analysis, in order to solve the aforementioned technical problems.

[0007] The present invention is implemented as follows: a non-destructive testing method for internal voids of a structure based on acoustic spectrum analysis, comprising the following steps: acquiring multi-channel time-domain response signals and surface temperature field data of the structure under test, and determining the coating thickness of the structure under test based on the multi-channel time-domain response signals.

[0008] The multi-channel time-domain response signal is decomposed into time and frequency components to extract the spectral feature parameters of each channel. The spectral feature parameters are then acoustically compensated and corrected according to the coating thickness to obtain the corrected spectral feature parameters.

[0009] Anomaly features are extracted from the surface temperature field data to obtain temperature anomaly feature parameters. The temperature anomaly feature parameters are then thermally compensated and corrected based on the coating thickness to obtain corrected temperature anomaly feature parameters.

[0010] The corrected spectral feature parameters are input into a preset nonlinear inversion model based on the physical laws of bridge structures to predict the geometric parameters of void defects. At the same time, the corrected temperature anomaly feature parameters and the corrected spectral feature parameters are orthogonally combined to generate the confidence level of the true defects.

[0011] Based on the geometric parameters of the void defect and the confidence level of the true defect, the non-destructive testing results of the void are obtained.

[0012] Furthermore, the step of determining the coating thickness of the structure under test based on the multi-channel time-domain response signal specifically includes: performing continuous wavelet transform on the multi-channel time-domain response signal using the wavelet transform modulus maxima method to construct a time-scale map.

[0013] Extract the time coordinates of the first mode maximum point representing the reflected wave from the upper surface of the coating, and the time coordinates of the second mode maximum point representing the reflected wave from the coating-steel substrate interface from the time-scale diagram.

[0014] The time delay difference is calculated based on the time coordinates of the first and second modulus maxima, and the coating thickness of the structure under test is calculated by combining the preset sound velocity value in the coating.

[0015] Furthermore, the spectral characteristic parameters include the main frequency amplitude, peak-to-width ratio, and energy concentration. The steps of performing time-frequency decomposition on the multi-channel time-domain response signal, extracting the spectral characteristic parameters of each channel, and performing acoustic compensation correction on the spectral characteristic parameters according to the coating thickness to obtain the corrected spectral characteristic parameters specifically include: performing time-frequency decomposition on the multi-channel time-domain response signal to obtain multiple frequency bands.

[0016] Calculate the energy percentage of each frequency band, take the frequency band with the highest energy percentage as the main frequency band, and determine the main frequency amplitude based on the main frequency band.

[0017] On the power spectrum of the main frequency band, determine the half-power point, calculate the half-power bandwidth, and determine the peak-to-width ratio based on the half-power bandwidth.

[0018] The sum of the energy proportions of the top-ranking frequency bands is used as the energy concentration.

[0019] Based on the pre-calibrated acoustic compensation coefficient function, the main frequency amplitude is corrected according to the coating thickness to obtain the corrected main frequency amplitude, which, together with the peak-to-width ratio and energy concentration, constitutes the corrected spectral characteristic parameters.

[0020] Furthermore, the surface temperature field data includes temperature field images before and after acoustic excitation; the temperature anomaly feature parameters include the local temperature rise amplitude, thermal diffusion gradient, and morphological feature parameters of the abnormal hot spot; the step of extracting anomaly features from the surface temperature field data to obtain temperature anomaly feature parameters, and performing thermal compensation correction on the temperature anomaly feature parameters according to the coating thickness to obtain corrected temperature anomaly feature parameters specifically includes: performing differential processing on the temperature field images before and after acoustic excitation to obtain a differential thermal map.

[0021] In the differential thermal map, connected regions whose temperature amplitude exceeds a preset background threshold are identified, and these connected regions are defined as abnormal hot spots.

[0022] The local temperature rise amplitude is extracted from the abnormal hot spot as the peak temperature, and the thermal diffusion gradient centered on the peak temperature point is calculated, that is, the rate of change of temperature with radial distance.

[0023] The morphological feature parameters of the abnormal hot spot are extracted using an edge detection algorithm; the morphological feature parameters include hot spot area, eccentricity, and compactness.

[0024] Based on the pre-calibrated thermal compensation coefficient function, the local temperature rise amplitude is corrected according to the coating thickness to obtain the corrected local temperature rise amplitude, which, together with the rate of change of temperature with radial distance and morphological characteristic parameters, constitutes the corrected temperature anomaly characteristic parameters.

[0025] Furthermore, the geometric parameters of the void defect include equivalent diameter, depth, and shape. The construction method of the nonlinear inversion model improved based on the physical laws of bridge structures is as follows: a finite element model of the structure to be tested is established, and tie-arch constraints are applied to the finite element model; the tie-arch constraints include tie preload, arch rib elastic modulus, and hanger equivalent stiffness; voids with different equivalent diameters, depths, and shapes are set in the finite element model, and the corresponding corrected spectral feature parameter sets are calculated. Then, the mapping relationship between the geometric parameters of the void defect and the corrected spectral feature parameter sets is fitted by neural network regression to obtain the nonlinear inversion model improved based on the physical laws of bridge structures.

[0026] Furthermore, the method for orthogonally combining the corrected temperature anomaly characteristic parameters and the corrected spectral characteristic parameters is as follows: a two-dimensional discriminant space is constructed, where the horizontal axis represents the predicted equivalent diameter of the void defect calculated based on the corrected spectral characteristic parameters, and the vertical axis represents the acoustic-induced temperature rise response intensity index calculated based on the corrected temperature anomaly characteristic parameters. The measured samples are projected onto this two-dimensional discriminant space to obtain projection points. Then, based on the distances between the projection points and the prior cluster centers of the real voids and the prior cluster centers of the pseudo-defects, the confidence level of the real defects is determined. The measured samples include the corrected temperature anomaly characteristic parameters and the corrected spectral characteristic parameters corresponding to the current structure under test.

[0027] Another objective of this invention is to provide a non-destructive testing device for internal voids in structures based on acoustic spectrum analysis, used to implement the aforementioned non-destructive testing method for internal voids in structures based on acoustic spectrum analysis. Specifically, it includes: a data acquisition module, used to acquire multi-channel time-domain response signals and surface temperature field data of the structure under test, and to determine the coating thickness of the structure under test based on the multi-channel time-domain response signals.

[0028] The spectral feature extraction and correction module is used to perform time-frequency decomposition on the multi-channel time-domain response signal, extract the spectral feature parameters of each channel, and perform acoustic compensation correction on the spectral feature parameters according to the coating thickness to obtain the corrected spectral feature parameters.

[0029] The abnormal feature extraction and correction module is used to extract abnormal features from the surface temperature field data to obtain temperature abnormal feature parameters, and to perform thermal compensation correction on the temperature abnormal feature parameters according to the coating thickness to obtain corrected temperature abnormal feature parameters.

[0030] The void defect prediction module is used to input the corrected spectral feature parameters into a preset nonlinear inversion model based on the physical laws of bridge structures to predict the geometric parameters of void defects. At the same time, the corrected temperature anomaly feature parameters and the corrected spectral feature parameters are orthogonally combined to generate the confidence level of the true defects.

[0031] The test result output module is used to obtain the non-destructive testing results of voids based on the geometric parameters of the void defect and the confidence level of the true defect.

[0032] Furthermore, the data acquisition module specifically includes: a data acquisition unit, used to acquire multi-channel time-domain response signals and surface temperature field data of the structure under test.

[0033] The time-scale plot construction unit is used to perform continuous wavelet transform on the multi-channel time-domain response signal using the wavelet transform modulus maxima method to construct a time-scale plot.

[0034] The time coordinate extraction unit is used to extract the time coordinates of the first mode maximum point representing the reflected wave on the upper surface of the coating and the second mode maximum point representing the reflected wave at the coating-steel substrate interface in the time-scale diagram.

[0035] The coating thickness calculation unit is used to calculate the time delay difference based on the time coordinates of the first modulus maximum point and the second modulus maximum point, and to calculate the coating thickness of the structure under test by combining the preset sound velocity value in the coating.

[0036] Furthermore, the spectral feature extraction and correction module specifically includes a time-frequency decomposition unit, used to perform time-frequency decomposition on the multi-channel time-domain response signal to obtain multiple frequency bands.

[0037] The main frequency amplitude determination unit is used to calculate the energy proportion of each frequency band, take the frequency band with the highest energy proportion as the main frequency band, and determine the main frequency amplitude based on the main frequency band.

[0038] The peak-to-width ratio calculation unit is used to determine the half-power point on the power spectrum of the main frequency band, calculate the half-power bandwidth, and determine the peak-to-width ratio based on the half-power bandwidth.

[0039] The energy concentration calculation unit is used to calculate the sum of the energy proportions of several frequency bands with the highest energy proportions, which is then used as the energy concentration.

[0040] The main frequency amplitude correction unit is used to correct the main frequency amplitude based on the coating thickness according to the pre-calibrated acoustic compensation coefficient function, so as to obtain the corrected main frequency amplitude, and together with the peak width ratio and energy concentration, constitute the corrected spectral characteristic parameters.

[0041] Furthermore, the abnormal feature extraction and correction module specifically includes: a differential processing unit, used to perform differential processing on the temperature field images before and after acoustic excitation to obtain a differential thermogram.

[0042] An abnormal hot spot determination unit is used to identify connected regions in the differential thermal map whose temperature amplitude exceeds a preset background threshold, and define the connected regions as abnormal hot spots.

[0043] The thermal diffusion gradient calculation unit is used to extract the local temperature rise value as the peak temperature from the abnormal hot spot and calculate the thermal diffusion gradient centered on the peak temperature point, that is, the rate of change of temperature with radial distance.

[0044] The morphological feature extraction unit is used to extract the morphological feature parameters of the abnormal hot spot using an edge detection algorithm; the morphological feature parameters include hot spot area, eccentricity, and compactness.

[0045] The local temperature rise amplitude correction unit is used to correct the local temperature rise amplitude based on the coating thickness according to the pre-calibrated thermal compensation coefficient function, so as to obtain the corrected local temperature rise amplitude, and together with the rate of change of temperature with radial distance and morphological characteristic parameters, constitute the corrected temperature anomaly characteristic parameters.

[0046] This invention provides a non-destructive testing method for internal voids in structures based on acoustic spectrum analysis. By acquiring acoustic and thermal dual-physical field information, introducing coating thickness compensation to eliminate coating interference, and introducing a nonlinear inversion model improved based on the physical laws of bridge structures, the method realizes the correlation between spectral characteristics and the geometric parameters of void defects. At the same time, combined with orthogonal joint analysis, it effectively suppresses false defect misjudgments, significantly improving the accuracy and reliability of non-destructive testing of internal voids in long-span tied arch bridge structures. Attached Figure Description

[0047] Figure 1 This is a flowchart illustrating the non-destructive testing method for internal voids in structures based on acoustic spectrum analysis, provided in an embodiment of the present invention.

[0048] Figure 2 This is a flowchart illustrating the steps for determining the coating thickness of the structure to be tested, as provided in an embodiment of the present invention.

[0049] Figure 3 This is a flowchart illustrating step S200 in the non-destructive testing method for internal voids in a structure based on acoustic spectrum analysis provided in an embodiment of the present invention.

[0050] Figure 4 This is a flowchart illustrating step S300 in the non-destructive testing method for internal voids in a structure based on acoustic spectrum analysis provided in an embodiment of the present invention.

[0051] Figure 5 This is a schematic diagram of the structure of the non-destructive testing device for internal voids based on acoustic spectrum analysis provided in an embodiment of the present invention.

[0052] Figure 6 This is a schematic diagram of the data acquisition module provided in an embodiment of the present invention.

[0053] Figure 7 This is a schematic diagram of the structure of the spectral feature extraction and correction module provided in an embodiment of the present invention.

[0054] Figure 8 This is a schematic diagram of the structure of the abnormal feature extraction and correction module provided in an embodiment of the present invention. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0056] like Figure 1 As shown, in one embodiment of the present invention, a non-destructive testing method for internal voids of a structure based on acoustic spectrum analysis is provided, including the following steps: S100, acquiring multi-channel time-domain response signals and surface temperature field data of the structure under test, and determining the coating thickness of the structure under test based on the multi-channel time-domain response signals.

[0057] S200. Perform time-frequency decomposition on the multi-channel time-domain response signal, extract the spectral feature parameters of each channel, and perform acoustic compensation correction on the spectral feature parameters according to the coating thickness to obtain the corrected spectral feature parameters.

[0058] S300. Extract abnormal features from the surface temperature field data to obtain temperature abnormal feature parameters, and perform thermal compensation correction on the temperature abnormal feature parameters according to the coating thickness to obtain corrected temperature abnormal feature parameters.

[0059] S400. Input the corrected spectral characteristic parameters into the preset nonlinear inversion model based on the physical laws of bridge structures to predict the geometric parameters of void defects. At the same time, perform orthogonal joint analysis of the corrected temperature anomaly characteristic parameters and the corrected spectral characteristic parameters to generate the confidence level of the true defects.

[0060] S500, based on the geometric parameters of the void defect and the confidence level of the true defect, obtains the non-destructive testing results of the void.

[0061] In practical applications, firstly, an acoustic wave sensor array (e.g., 15 accelerometers arranged in 3 rows and 5 columns) and an infrared thermal imager (e.g., FLIR A655sc) are deployed on the surface of the structure under test (e.g., a steel box girder). An electromagnetic exciter is used to generate broadband acoustic waves (e.g., linear sweep pulses with a frequency range of 0.5kHz-20kHz) at the geometric center of the acoustic wave sensor array. Simultaneously with the acoustic wave excitation, a synchronous clock triggers the acoustic wave sensor array and the infrared thermal imager to acquire multi-channel time-domain response signals and surface temperature field data. The surface temperature field data includes one frame of surface temperature field image before and after acoustic wave excitation. Simultaneously, the coating thickness of the structure under test is determined from the multi-channel time-domain response signal.

[0062] Next, time-frequency decomposition was performed on the time-domain response signals of each channel in the multi-channel time-domain response signal to extract the spectral characteristic parameters of each channel (including the main frequency amplitude, peak width ratio, energy concentration, etc.), and acoustic compensation correction was performed according to the coating thickness to obtain the corrected spectral characteristic parameters. At the same time, anomaly features were extracted from the surface temperature field data to obtain temperature anomaly characteristic parameters (including local temperature rise amplitude, thermal diffusion gradient, hot spot area, eccentricity, compactness), and thermal compensation correction was performed according to the coating thickness to obtain the corrected temperature anomaly characteristic parameters.

[0063] Then, the corrected spectral characteristic parameters are input into a pre-constructed nonlinear inversion model based on the physical laws of bridge structures to predict the geometric parameters of void defects (such as equivalent diameter, depth, and shape). At the same time, an orthogonal joint analysis method is used to fuse the corrected temperature anomaly characteristic parameters with the corrected spectral characteristic parameters to generate the confidence level of the true defect (value range 0-1).

[0064] Finally, based on the predicted void defect geometric parameters and the confidence level of the actual defect, the void non-destructive testing results are output: if the confidence level of the actual defect is greater than the preset first threshold (e.g., 0.8), it is determined to be a real void defect, and the corresponding equivalent diameter, depth, and shape parameters are output; if the confidence level of the actual defect is less than the preset second threshold (e.g., 0.3), it is determined to be a pseudo defect or a non-void defect, and "non-void" and possible pseudo defect types are output; if the confidence level of the actual defect is greater than or equal to the second threshold and less than or equal to the first threshold, "uncertain" and the inverted geometric parameters (for reference only) are output.

[0065] like Figure 2 As shown, in a preferred embodiment of the present invention, the step of determining the coating thickness of the structure under test based on the multi-channel time-domain response signal specifically includes: S110, performing continuous wavelet transform on the multi-channel time-domain response signal using the wavelet transform modulus maxima method to construct a time-scale map.

[0066] S120. Extract the time coordinates of the first mode maximum point representing the reflected wave on the upper surface of the coating and the second mode maximum point representing the reflected wave at the coating-steel substrate interface from the time-scale diagram.

[0067] S130. Calculate the time delay difference based on the time coordinates of the first modulus maximum point and the second modulus maximum point, and calculate the coating thickness of the structure to be tested by combining the preset sound velocity value in the coating.

[0068] In this embodiment of the invention, the coating thickness is determined based on the principle of ultrasonic pulse echo. Due to the significant difference in acoustic impedance between the coating (such as epoxy zinc-rich primer) and the steel substrate, ultrasonic waves will generate two reflected echoes at the upper surface of the coating and the coating-steel substrate interface. When the coating thickness is thin, these two echoes may overlap significantly in the time domain, making them difficult to read directly. Therefore, this embodiment of the invention employs the wavelet transform modulus maxima method.

[0069] Specifically, first, a continuous wavelet transform is performed on the time-domain response signal x(t) of a single channel: In the formula, The result of the transformation is a time-scale plot; 'a' is the scale factor (inversely proportional to the frequency, i.e., the actual frequency corresponding to the scale factor 'a'). ,in, The center frequency of the mother wavelet. (where b is the sampling frequency of the signal), b is the shift factor (time), and t is the time variable; The mother wavelet function (in this embodiment of the invention, the real-valued Db4 wavelet can be selected); Indicates complex conjugation.

[0070] Next, on the time-scale plot, the local maxima of the wavelet coefficients' modulus correspond to the abrupt change points of the signal (i.e., the starting points of the reflected wave); the first significant modulus maxima corresponds to the reflected wave from the upper surface of the coating, and its time coordinate is the time coordinate t of the first modulus maxima. S The second significant modulus maximum corresponds to the reflected wave at the coating-steel substrate interface, and its time coordinate is the time coordinate t of the second modulus maximum. I Then the time delay difference The coating thickness d is calculated by the following formula: In the formula, v c The longitudinal wave velocity in the coating can be obtained from a material handbook or pre-calibrated.

[0071] It should be noted that when the coating is extremely thin (e.g., <0.2mm), the reflected waves from the upper surface of the coating and the reflected waves from the coating-steel substrate interface highly overlap in the time domain. In this case, the interface reflected wave components can be extracted by combining multi-channel data with existing signal separation methods based on K-SVD dictionary learning, and then the time delay difference can be calculated.

[0072] like Figure 3 As shown, in a preferred embodiment of the present invention, the spectral characteristic parameters include the main frequency amplitude, peak width ratio, and energy concentration; the step of performing time-frequency decomposition on the multi-channel time-domain response signal, extracting the spectral characteristic parameters of each channel, and performing acoustic compensation correction on the spectral characteristic parameters according to the coating thickness to obtain the corrected spectral characteristic parameters, namely step S200, specifically includes: S210, performing time-frequency decomposition on the multi-channel time-domain response signal to obtain multiple frequency bands.

[0073] Specifically, time-frequency decomposition can be performed using wavelet packet transform to decompose the signal into multiple (e.g., 4 layers) frequency bands of equal bandwidth; for each frequency band, its energy is calculated based on the wavelet packet coefficients, and then the total energy is obtained by summing them up.

[0074] S220. Calculate the energy percentage of each frequency band, take the frequency band with the highest energy percentage as the main frequency band, and determine the main frequency amplitude based on the main frequency band.

[0075] Specifically, the energy of each frequency band is divided by the total energy to obtain the energy percentage of that frequency band. The frequency band with the highest energy percentage is called the main frequency band, and its center frequency is called the main frequency. The amplitude corresponding to the energy of the main frequency band is called the main frequency amplitude.

[0076] S230. On the power spectrum of the main frequency band, determine the half-power point, calculate the half-power bandwidth, and determine the peak-to-width ratio based on the half-power bandwidth.

[0077] Specifically, on the power spectrum of the main frequency band, find the left and right frequencies corresponding to the point where the peak amplitude drops to half power. The half-power bandwidth is equal to the difference between the left and right frequencies. Dividing the half-power bandwidth by the center frequency gives the peak-to-width ratio (PVR). The PVR reflects the sharpness of the resonance peak. For void defects, due to the reduced local stiffness, the resonance peak is usually sharper, resulting in a smaller PVR; for homogeneous structures or pseudo-defects, the PVR is larger.

[0078] S240. Calculate the sum of the energy proportions of several frequency bands with the highest energy proportions, and use this as the energy concentration.

[0079] Specifically, each frequency band is sorted from largest to smallest according to its energy proportion, and the sum of the energy proportions of the first k frequency bands (such as the first three) is the energy concentration. The higher the value, the more concentrated the energy is in a few main frequency bands. Void defects usually have a high energy concentration.

[0080] S250. Based on the pre-calibrated acoustic compensation coefficient function, the main frequency amplitude is corrected according to the coating thickness to obtain the corrected main frequency amplitude, which, together with the peak-to-width ratio and energy concentration, constitutes the corrected spectral characteristic parameters.

[0081] Since the coating thickness attenuates sound wave energy, this embodiment of the invention uses an exponentially decaying acoustic compensation coefficient function to adjust the original dominant frequency amplitude. Compensation corrections are made; specifically, the acoustic compensation coefficient function. The expression is as follows: In the formula, e is the natural constant. The acoustic attenuation coefficient can be pre-calibrated as follows: Prepare a set of standard steel box girder test blocks with different coating thicknesses. Machine standard voids with known equivalent diameters, depths, and shapes on each standard test block. Use the same detection system to collect the original dominant frequency amplitude values ​​under different coating thicknesses. Fit the acoustic attenuation coefficient using exponential regression analysis; then the corrected dominant frequency amplitude value... .

[0082] In this embodiment of the invention, multidimensional spectral features are extracted using wavelet packet transform, and the dominant frequency amplitude is exponentially compensated and corrected based on the coating thickness, eliminating the deviation caused by coating attenuation. Furthermore, by introducing peak-to-width ratio and energy concentration as supplementary features, the sensitivity to the geometric parameters of void defects is enhanced. Experiments show that after compensation and correction, the consistency of the dominant frequency amplitude of the same void under different coating thicknesses is improved to over 95%.

[0083] like Figure 4 As shown, in a preferred embodiment of the present invention, the surface temperature field data includes temperature field images before and after acoustic excitation; the temperature anomaly feature parameters include the local temperature rise amplitude, thermal diffusion gradient, and morphological feature parameters of the abnormal hot spot; the step of extracting anomaly features from the surface temperature field data to obtain temperature anomaly feature parameters, and performing thermal compensation correction on the temperature anomaly feature parameters according to the coating thickness to obtain the corrected temperature anomaly feature parameters, i.e., step S300 specifically includes: S310, performing differential processing on the temperature field images before and after acoustic excitation to obtain a differential thermal map.

[0084] In practical applications, an infrared thermal imager is used to acquire one frame of temperature field image before acoustic excitation (at time t0) and after excitation (at time t1, usually delayed by 1-2 seconds to allow the acoustically induced temperature rise to be fully conducted), forming surface temperature field data; by performing pixel-by-pixel difference between the two temperature field images, a differential thermal map can be obtained.

[0085] S320. Identify connected regions in the differential thermal map whose temperature amplitude exceeds a preset background threshold, and define the connected regions as abnormal hot spots.

[0086] Specifically, let the average temperature of the background region of the differential heatmap (usually a flat area far from the excitation point) be μ. b The standard deviation is σ b ; Temperature amplitude exceeding μ b +2σ b Pixels are marked as candidate anomalies; connected component analysis (eight-neighborhood) is performed on the candidate anomalies, and connected regions with an area greater than a minimum threshold (e.g., 10 pixels) are defined as anomalous hot spots. Each anomalous hot spot corresponds to a possible real void defect or pseudo-defect.

[0087] S330. Extract the local temperature rise value from the abnormal hot spot as the peak temperature, and calculate the thermal diffusion gradient centered on the peak temperature point, i.e., the rate of change of temperature with radial distance.

[0088] Specifically, the local temperature rise amplitude is defined as the maximum value of the pixel temperature within the abnormal hot spot (i.e., the maximum temperature difference); with the peak temperature point as the center, the average temperature at each radial distance is calculated, and then the first derivative is obtained to get the rate of change of temperature with radial distance, which is the thermal diffusion gradient.

[0089] S340. The morphological feature parameters of the abnormal hot spot are extracted using an edge detection algorithm; the morphological feature parameters include hot spot area, eccentricity, and compactness.

[0090] Specifically, the existing Canny edge detection algorithm can be used to extract the contours of abnormal hot spots; the area of ​​the hot spot is equal to the number of pixels within the contour multiplied by the area of ​​a single pixel; the eccentricity is obtained by fitting the ratio of the major and minor axes of the ellipse, where the square of the eccentricity is equal to 1 minus the square of the ratio of the major and minor axes; the compactness is 4πS / P. 2 Where P is the perimeter of the profile and S is the area of ​​the hot spot.

[0091] S350. Based on the pre-calibrated thermal compensation coefficient function, the local temperature rise amplitude is corrected according to the coating thickness to obtain the corrected local temperature rise amplitude, and together with the rate of change of temperature with radial distance and morphological characteristic parameters, the corrected temperature anomaly characteristic parameters are formed.

[0092] Because coating thickness attenuates heat conduction, causing the measured temperature rise on the surface to be lower than the true value, this embodiment of the invention uses an exponentially decaying thermal compensation coefficient function to adjust the original local temperature rise value. Compensation and correction are performed; specifically, the thermal compensation coefficient function. The expression is as follows: In the formula, e is the natural constant. The thermal conductivity attenuation coefficient can be pre-calibrated as follows: Prepare a set of standard steel box girder test blocks with different coating thicknesses. Machine standard voids with known equivalent diameters, depths, and shapes on each standard test block. Use the same detection system to collect the original local temperature rise amplitude values ​​under different coating thicknesses. Fit the thermal conductivity attenuation coefficient using exponential regression analysis; then the corrected local temperature rise amplitude value... .

[0093] In this embodiment of the invention, differential thermal mapping effectively eliminates environmental background interference, and the spatial characteristics (temperature rise, gradient, morphology) of abnormal hot spots are used to describe the defect response from a thermal perspective. Furthermore, thermal compensation correction eliminates the attenuation effect of coating thickness on the temperature rise amplitude, making the thermal anomaly characteristics under different coating thicknesses comparable. In actual testing, hot spots generated by real void defects typically exhibit concentrated, high eccentricity, and low compactness characteristics, while hot spots in the residual stress zone are uniformly diffused, have low eccentricity, and high compactness; the two can be clearly distinguished.

[0094] In a preferred embodiment of the present invention, the geometric parameters of the void defect include equivalent diameter, depth, and shape; the construction method of the nonlinear inversion model based on the physical laws of bridge structures is as follows: a finite element model of the structure to be tested is established, and tie-arch constraints are applied to the finite element model; the tie-arch constraints include tie preload, arch rib elastic modulus, and hanger equivalent stiffness; voids with different equivalent diameters, depths, and shapes are set in the finite element model, the corresponding corrected spectral feature parameter sets are calculated, and then the mapping relationship between the geometric parameters of the void defect and the corrected spectral feature parameter sets is fitted by neural network regression to obtain the nonlinear inversion model based on the physical laws of bridge structures.

[0095] In practical applications, firstly, a refined finite element model of the structure to be tested (such as the steel box girder to be tested) is established using Abaqus or ANSYS for simulation. The dimensions and material parameters (such as steel density, elastic modulus, Poisson's ratio, etc.) of the finite element model are determined according to the bridge design drawings. The tie rod arch constraint conditions are applied as follows: tie rod pretension: axial tension is applied at both ends of the tie rod element.

[0096] Arch rib elastic modulus: Input according to the actual material.

[0097] Equivalent stiffness of the suspension rod: The suspension rod is simplified as a spring element, and the stiffness is calculated based on the actual cross-section.

[0098] Next, in the aforementioned finite element model, void defects are set at different locations inside the steel box girder (avoiding welds and stiffeners) to generate training data. Specifically, the geometric parameters of the void defects are set as follows: equivalent diameter: set in increments of 3-5 mm, such as setting 6 increments.

[0099] Depth (distance from the upper surface of the cavity to the outer surface of the steel box girder): set at intervals of 1-3mm, such as 5 intervals.

[0100] Shape: such as sphere, ellipsoid (major axis ratio 2:1, 3:1), irregular shape, etc.

[0101] The aforementioned geometric parameters together constitute the structural modification factor vector.

[0102] For each set of geometric parameters (equivalent diameter, depth, shape), the same broadband acoustic excitation as in field detection is applied, and the time-domain response signals of each measurement point of the acoustic sensor array are extracted to obtain multi-channel time-domain response signals. Then, the corrected spectral characteristic parameter set is calculated according to the above method (Note: In the finite element model simulation, the coating thickness can be set to 0, but in actual applications, the coating thickness is determined and compensated for using the above method, which is equivalent to applying reverse attenuation to the uncoated simulation data). Approximately 6000 sets of samples are generated, of which 80% are used for the training set and 20% for the validation set.

[0103] Then, a feedforward deep neural network is constructed with an input layer dimension of 3 (corresponding to the corrected main frequency amplitude, peak width ratio, and energy concentration), four hidden layers, each with 128 neurons, and ReLU activation function. The output layer has three branches: equivalent diameter regression (linear activation), depth regression (linear activation), and shape classification (based on the Softmax function to output the probability distribution of spherical, ellipsoidal, and irregular shapes). The mean squared error loss function can be used for training. During training, the structural modification factor vector is embedded as a fixed parameter into the regularization term of the network: the structural modification factor vector corresponding to each sample in the training set is consistent with the actual parameters of the bridge, and the gradient update direction of the network is modified by the eigenvalues ​​pre-calculated by the finite element method. In specific implementation, a physical constraint layer can be added to the middle layer of the network. This physical constraint layer calculates the theoretical Green's function of the sound wave in the steel box girder structure based on the structural modification factor vector and compares it with the spectral characteristics currently predicted by the network. The difference is backpropagated to force the network to learn the feature mapping that conforms to the physical laws, and finally the trained network is obtained, which serves as the final nonlinear inversion model based on the physical laws of the bridge structure. This nonlinear inversion model can output the equivalent diameter prediction value, depth prediction value, and probability distribution of the shape of the void defect based on the currently obtained corrected main frequency amplitude, peak width ratio, and energy concentration.

[0104] Finally, the model performance was evaluated using a validation set. In practical applications, on an independent test set (containing 200 sets of untrained void defect geometric parameters), the average relative error for the equivalent diameter prediction was 6.8%, the average absolute error for the depth prediction was 0.7 mm, and the shape classification accuracy was 89.3%.

[0105] In this embodiment of the invention, a large-scale training data covering various geometric parameters of void defects and various tie-arch constraints is generated through finite element simulation. The nonlinear mapping between spectral features and geometric parameters of void defects is established by using neural network regression, which can significantly improve the prediction accuracy of void defects.

[0106] In a preferred embodiment of the present invention, the method for orthogonally combining the corrected temperature anomaly characteristic parameters and the corrected spectral characteristic parameters is as follows: a two-dimensional discriminant space is constructed, where the horizontal axis is the predicted equivalent diameter of the void defect calculated based on the corrected spectral characteristic parameters, and the vertical axis is the acoustic-induced temperature rise response intensity index calculated based on the corrected temperature anomaly characteristic parameters. The measured samples are projected onto this two-dimensional discriminant space to obtain projection points. Then, based on the distances between the projection points and the prior cluster centers of the real voids and the prior cluster centers of the pseudo-defects, the confidence level of the real defects is determined. The measured samples include the corrected temperature anomaly characteristic parameters and the corrected spectral characteristic parameters corresponding to the current structure under test.

[0107] Specifically, the horizontal axis X of the two-dimensional discrimination space represents the predicted equivalent diameter D of the void defect output by the aforementioned nonlinear inversion model. pred The vertical axis Y represents the intensity index of the acoustic-induced temperature rise response, defined as: In the formula, G is the thermal diffusion gradient of the abnormal hot spot; C is the compactness of the abnormal hot spot; E is the eccentricity of the abnormal hot spot. This acoustic-induced temperature rise response intensity index comprehensively reflects the intensity, diffusion degree and morphological characteristics of the abnormal hot spot: real void defects usually have a high local temperature rise amplitude, a small thermal diffusion gradient (heat accumulation), a large hot spot area, a low compactness and a high eccentricity, so the acoustic-induced temperature rise response intensity index Y value is large; while the acoustic-induced temperature rise response intensity index Y value of pseudo defects (such as residual stress zone) is small.

[0108] Before testing, calibration experiments were conducted using standard test blocks (containing known real void defects and known pseudo-defects, such as areas of localized thickness variation, coating peeling, and stress concentration); samples of each standard test block were collected (D...). pred For each pair of points (Y), calculate the true hollow prior cluster centers. and pseudo-defect prior cluster centers ;in, The mean of the predicted equivalent diameters for all known real void defects; The average value of the acoustic-induced temperature rise response intensity index corresponding to all known real void defects; The mean of the predicted equivalent diameters for all known pseudo-defects; The mean of the acoustic-induced temperature rise response intensity indices corresponding to all known pseudo-defects is given; the covariance matrix is ​​also calculated to perform Mahalanobis distance calculation.

[0109] For the projection point obtained from the current measured sample, calculate the Mahalanobis distance M from the projection point to the aforementioned real void prior cluster center. t And the Mahalanobis distance M from the projection point to the aforementioned pseudo-defect prior cluster center. f The confidence level of the true defect Defined as: Among them, when The larger the value, the farther the measured sample is from the prior cluster center of the spurious defect and the closer it is to the cluster of the real void, and the more likely it is to be a real void defect; otherwise, it is more likely to be a spurious defect.

[0110] like Figure 5 As shown, in another embodiment of the present invention, a non-destructive testing device for internal voids of a structure based on acoustic spectrum analysis is also provided to implement the above-mentioned non-destructive testing method for internal voids of a structure based on acoustic spectrum analysis. Specifically, it includes: a data acquisition module 10, used to acquire multi-channel time-domain response signals and surface temperature field data of the structure under test, and to determine the coating thickness of the structure under test based on the multi-channel time-domain response signals.

[0111] The spectral feature extraction and correction module 20 is used to perform time-frequency decomposition on the multi-channel time-domain response signal, extract the spectral feature parameters of each channel, and perform acoustic compensation correction on the spectral feature parameters according to the coating thickness to obtain the corrected spectral feature parameters.

[0112] The abnormal feature extraction and correction module 30 is used to extract abnormal features from the surface temperature field data to obtain temperature abnormal feature parameters, and to perform thermal compensation correction on the temperature abnormal feature parameters according to the coating thickness to obtain corrected temperature abnormal feature parameters.

[0113] The void defect prediction module 40 is used to input the corrected spectral feature parameters into a preset nonlinear inversion model based on the physical laws of bridge structures to predict the geometric parameters of void defects. At the same time, it performs orthogonal joint analysis on the corrected temperature anomaly feature parameters and the corrected spectral feature parameters to generate the confidence level of the true defect.

[0114] The test result output module 50 is used to obtain the non-destructive testing results of the void based on the geometric parameters of the void defect and the confidence level of the true defect.

[0115] like Figure 6 As shown, in a preferred embodiment of the present invention, the data acquisition module 10 specifically includes: a data acquisition unit 11, used to acquire multi-channel time-domain response signals and surface temperature field data of the structure under test.

[0116] The time-scale plot construction unit 12 is used to perform continuous wavelet transform on the multi-channel time-domain response signal using the wavelet transform modulus maxima method to construct a time-scale plot.

[0117] The time coordinate extraction unit 13 is used to extract the time coordinates of the first mode maximum point representing the reflected wave on the upper surface of the coating and the second mode maximum point representing the reflected wave at the coating-steel substrate interface in the time-scale diagram.

[0118] The coating thickness calculation unit 14 is used to calculate the time delay difference based on the time coordinates of the first modulus maximum point and the second modulus maximum point, and to calculate the coating thickness of the structure to be tested by combining the preset sound velocity value in the coating.

[0119] like Figure 7 As shown, in a preferred embodiment of the present invention, the spectral feature extraction and correction module 20 specifically includes: a time-frequency decomposition unit 21, used to perform time-frequency decomposition on the multi-channel time-domain response signal to obtain multiple frequency bands.

[0120] The main frequency amplitude determination unit 22 is used to calculate the energy proportion of each frequency band, take the frequency band with the highest energy proportion as the main frequency band, and determine the main frequency amplitude based on the main frequency band.

[0121] The peak-to-width ratio calculation unit 23 is used to determine the half-power point on the power spectrum of the main frequency band, calculate the half-power bandwidth, and determine the peak-to-width ratio based on the half-power bandwidth.

[0122] The energy concentration calculation unit 24 is used to calculate the sum of the energy proportions of several frequency bands with the highest energy proportions, which is used as the energy concentration.

[0123] The main frequency amplitude correction unit 25 is used to correct the main frequency amplitude based on the coating thickness according to the pre-calibrated acoustic compensation coefficient function, so as to obtain the corrected main frequency amplitude, and together with the peak width ratio and energy concentration, constitute the corrected spectral characteristic parameters.

[0124] like Figure 8 As shown, in a preferred embodiment of the present invention, the abnormal feature extraction and correction module 30 specifically includes: a differential processing unit 31, used to perform differential processing on the temperature field images before and after acoustic excitation to obtain a differential thermogram.

[0125] The abnormal hot spot determination unit 32 is used to identify connected regions in the differential thermal map whose temperature amplitude exceeds a preset background threshold, and define the connected regions as abnormal hot spots.

[0126] The thermal diffusion gradient calculation unit 33 is used to extract the local temperature rise value as the peak temperature from the abnormal hot spot and calculate the thermal diffusion gradient centered on the peak temperature point, that is, the rate of change of temperature with radial distance.

[0127] The morphological feature extraction unit 34 is used to extract the morphological feature parameters of the abnormal hot spot using an edge detection algorithm; the morphological feature parameters include hot spot area, eccentricity and compactness.

[0128] The local temperature rise amplitude correction unit 35 is used to correct the local temperature rise amplitude based on the coating thickness according to the pre-calibrated thermal compensation coefficient function, so as to obtain the corrected local temperature rise amplitude, and together with the rate of change of temperature with radial distance and morphological characteristic parameters, constitute the corrected temperature anomaly characteristic parameters.

[0129] It should be noted that the above modules and units can be implemented as a computer program, which can run on a computer device. The computer device's memory can store the computer program that makes up the modules or units, enabling the processor to execute the various steps of the above method.

[0130] It should be understood that although the steps in the flowcharts of the embodiments of the present invention are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in each embodiment may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps.

[0131] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods.

[0132] The above embodiments merely illustrate several implementation methods of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this patent should be determined by the appended claims.

Claims

1. A non-destructive testing method for internal voids in structures based on acoustic spectrum analysis, characterized in that, Includes the following steps: The multi-channel time-domain response signal and surface temperature field data of the structure under test are collected, and the coating thickness of the structure under test is determined based on the multi-channel time-domain response signal. The multi-channel time-domain response signal is decomposed into time and frequency, the spectral feature parameters of each channel are extracted, and the spectral feature parameters are acoustically compensated and corrected according to the coating thickness to obtain the corrected spectral feature parameters. Anomaly features are extracted from the surface temperature field data to obtain temperature anomaly feature parameters. The temperature anomaly feature parameters are then thermally compensated and corrected based on the coating thickness to obtain corrected temperature anomaly feature parameters. The corrected spectral feature parameters are input into a preset nonlinear inversion model based on the physical laws of bridge structures to predict the geometric parameters of void defects. At the same time, the corrected temperature anomaly feature parameters and the corrected spectral feature parameters are orthogonally combined to generate the confidence level of the true defects. Based on the geometric parameters of the void defect and the confidence level of the true defect, the non-destructive testing results of the void are obtained.

2. The non-destructive testing method for internal voids in structures based on acoustic spectrum analysis according to claim 1, characterized in that, The steps for determining the coating thickness of the structure under test based on the multi-channel time-domain response signal specifically include: For the multi-channel time-domain response signal, continuous wavelet transform is performed using the wavelet transform modulus maxima method to construct a time-scale map; Extract the time coordinates of the first mode maximum point representing the reflected wave from the upper surface of the coating and the time coordinates of the second mode maximum point representing the reflected wave from the coating-steel substrate interface from the time-scale diagram. The time delay difference is calculated based on the time coordinates of the first and second modulus maxima, and the coating thickness of the structure under test is calculated by combining the preset sound velocity value in the coating.

3. The non-destructive testing method for internal voids in structures based on acoustic spectrum analysis according to claim 1, characterized in that, The spectral characteristic parameters include the dominant frequency amplitude, peak-to-width ratio, and energy concentration. The steps of performing time-frequency decomposition on the multi-channel time-domain response signal, extracting the spectral characteristic parameters of each channel, and then performing acoustic compensation correction on the spectral characteristic parameters based on the coating thickness to obtain the corrected spectral characteristic parameters specifically include: The multi-channel time-domain response signal is decomposed into multiple frequency bands by performing time-frequency decomposition. Calculate the energy percentage of each frequency band, take the frequency band with the highest energy percentage as the main frequency band, and determine the main frequency amplitude based on the main frequency band; On the power spectrum of the main frequency band, determine the half-power point, calculate the half-power bandwidth, and determine the peak-to-width ratio based on the half-power bandwidth; The sum of the energy proportions of the top-ranking frequency bands is used as the energy concentration. Based on the pre-calibrated acoustic compensation coefficient function, the main frequency amplitude is corrected according to the coating thickness to obtain the corrected main frequency amplitude, which, together with the peak-to-width ratio and energy concentration, constitutes the corrected spectral characteristic parameters.

4. The non-destructive testing method for internal voids in structures based on acoustic spectrum analysis according to claim 1, characterized in that, The surface temperature field data includes temperature field images before and after acoustic excitation; the temperature anomaly characteristic parameters include the local temperature rise amplitude, thermal diffusion gradient, and morphological characteristic parameters of the abnormal hot spots; the steps of extracting anomaly features from the surface temperature field data to obtain temperature anomaly characteristic parameters, and performing thermal compensation correction on the temperature anomaly characteristic parameters according to the coating thickness to obtain the corrected temperature anomaly characteristic parameters, specifically include: Differential processing is performed on the temperature field images before and after acoustic excitation to obtain differential thermograms; In the differential thermal map, connected regions with temperature amplitudes exceeding a preset background threshold are identified, and these connected regions are defined as abnormal hot spots. The local temperature rise amplitude is extracted from the abnormal hot spot as the peak temperature, and the thermal diffusion gradient centered on the peak temperature point is calculated, that is, the rate of change of temperature with radial distance. The morphological feature parameters of the abnormal hot spot are extracted using an edge detection algorithm; the morphological feature parameters include hot spot area, eccentricity, and compactness. Based on the pre-calibrated thermal compensation coefficient function, the local temperature rise amplitude is corrected according to the coating thickness to obtain the corrected local temperature rise amplitude, which, together with the rate of change of temperature with radial distance and morphological characteristic parameters, constitutes the corrected temperature anomaly characteristic parameters.

5. The non-destructive testing method for internal voids in structures based on acoustic spectrum analysis according to claim 1, characterized in that, The geometric parameters of the void defect include equivalent diameter, depth, and shape. The construction method of the nonlinear inversion model based on the physical laws of bridge structures is as follows: a finite element model of the structure to be tested is established, and tie-arch constraints are applied to the finite element model. The tie-arch constraints include tie preload, arch rib elastic modulus, and hanger equivalent stiffness. Voids with different equivalent diameters, depths, and shapes are set in the finite element model, and the corresponding corrected spectral feature parameter sets are calculated. Then, the mapping relationship between the geometric parameters of the void defect and the corrected spectral feature parameter sets is fitted by neural network regression to obtain the nonlinear inversion model based on the physical laws of bridge structures.

6. The non-destructive testing method for internal voids in structures based on acoustic spectrum analysis according to claim 5, characterized in that, The method for orthogonal joint analysis of the corrected temperature anomaly characteristic parameters and the corrected spectral characteristic parameters is as follows: a two-dimensional discriminant space is constructed, where the horizontal axis represents the predicted equivalent diameter of the void defect calculated based on the corrected spectral characteristic parameters, and the vertical axis represents the acoustic temperature rise response intensity index calculated based on the corrected temperature anomaly characteristic parameters. The measured samples are projected onto this two-dimensional discriminant space to obtain projection points. Then, based on the distances of the projection points to the prior cluster centers of the real voids and the prior cluster centers of the pseudo-defects, the confidence level of the real defects is determined. The measured samples include the corrected temperature anomaly characteristic parameters and the corrected spectral characteristic parameters corresponding to the current structure under test.

7. A non-destructive testing device for internal voids in structures based on acoustic spectrum analysis, used to implement the non-destructive testing method for internal voids in structures based on acoustic spectrum analysis as described in any one of claims 1-6, characterized in that, include: The data acquisition module is used to collect multi-channel time-domain response signals and surface temperature field data of the structure under test, and to determine the coating thickness of the structure under test based on the multi-channel time-domain response signals. The spectral feature extraction and correction module is used to perform time-frequency decomposition on the multi-channel time-domain response signal, extract the spectral feature parameters of each channel, and perform acoustic compensation correction on the spectral feature parameters according to the coating thickness to obtain the corrected spectral feature parameters. An anomaly feature extraction and correction module is used to extract anomaly features from the surface temperature field data to obtain temperature anomaly feature parameters, and to perform thermal compensation correction on the temperature anomaly feature parameters according to the coating thickness to obtain corrected temperature anomaly feature parameters. The void defect prediction module is used to input the corrected spectral feature parameters into a preset nonlinear inversion model based on the physical laws of bridge structures to predict the geometric parameters of void defects. At the same time, the corrected temperature anomaly feature parameters and the corrected spectral feature parameters are orthogonally combined to generate the confidence level of the true defects. The test result output module is used to obtain the non-destructive testing results of voids based on the geometric parameters of the void defect and the confidence level of the true defect.

8. The non-destructive testing device for internal voids in structures based on acoustic spectrum analysis according to claim 7, characterized in that, The data acquisition module specifically includes: The data acquisition unit is used to acquire multi-channel time-domain response signals and surface temperature field data of the structure under test; The time-scale plot construction unit is used to perform continuous wavelet transform on the multi-channel time-domain response signal using the wavelet transform modulus maxima method to construct a time-scale plot. The time coordinate extraction unit is used to extract the time coordinates of the first mode maximum point representing the reflected wave on the upper surface of the coating and the second mode maximum point representing the reflected wave at the coating-steel substrate interface in the time-scale diagram. The coating thickness calculation unit is used to calculate the time delay difference based on the time coordinates of the first modulus maximum point and the second modulus maximum point, and to calculate the coating thickness of the structure under test by combining the preset sound velocity value in the coating.

9. The non-destructive testing device for internal voids in structures based on acoustic spectrum analysis according to claim 7, characterized in that, The spectral feature extraction and correction module specifically includes: The time-frequency decomposition unit is used to perform time-frequency decomposition on the multi-channel time-domain response signal to obtain multiple frequency bands. The main frequency amplitude determination unit is used to calculate the energy proportion of each frequency band, take the frequency band with the highest energy proportion as the main frequency band, and determine the main frequency amplitude based on the main frequency band. The peak-to-width ratio calculation unit is used to determine the half-power point on the power spectrum of the main frequency band, calculate the half-power bandwidth, and determine the peak-to-width ratio based on the half-power bandwidth. The energy concentration calculation unit is used to calculate the sum of the energy proportions of several frequency bands with the highest energy proportions, which is then used as the energy concentration. The main frequency amplitude correction unit is used to correct the main frequency amplitude based on the coating thickness according to the pre-calibrated acoustic compensation coefficient function, so as to obtain the corrected main frequency amplitude, and together with the peak width ratio and energy concentration, constitute the corrected spectral characteristic parameters.

10. The non-destructive testing device for internal voids in structures based on acoustic spectrum analysis according to claim 7, characterized in that, The abnormal feature extraction and correction module specifically includes: The differential processing unit is used to perform differential processing on the temperature field images before and after acoustic excitation to obtain differential thermograms. An abnormal hot spot determination unit is used to identify connected regions in the differential thermal map whose temperature amplitude exceeds a preset background threshold, and define the connected regions as abnormal hot spots. The thermal diffusion gradient calculation unit is used to extract the local temperature rise value as the peak temperature from the abnormal hot spot and calculate the thermal diffusion gradient centered on the peak temperature point, that is, the rate of change of temperature with radial distance. The morphological feature extraction unit is used to extract the morphological feature parameters of the abnormal hot spot using an edge detection algorithm; the morphological feature parameters include hot spot area, eccentricity, and compactness. The local temperature rise amplitude correction unit is used to correct the local temperature rise amplitude based on the coating thickness according to the pre-calibrated thermal compensation coefficient function, so as to obtain the corrected local temperature rise amplitude, and together with the rate of change of temperature with radial distance and morphological characteristic parameters, constitute the corrected temperature anomaly characteristic parameters.

Images