Full-focusing ultrasonic imaging method for multi-modal damage of composite material

Abstract

The invention provides a full-focusing ultrasonic imaging method for multi-modal damage of a composite material, and the method comprises the steps: calculating the correlation of the reflection intensity of a sound wave signal at different moments and spatial positions through employing an established quantitative corresponding relation, recognizing a close contact region and a loose separation region between a bonding layer and a substrate, and carrying out the recognition of the close contact region and the loose separation region. Judging the continuity distribution condition of the interface contact state; according to the continuous distribution condition of the interface contact state, key feature points of dynamic damage evolution are extracted, a time sequence analysis method is adopted to track the feature points to adapt to damage features of dynamic changes, and prediction data of the damage expansion trend are obtained; and according to the dynamic representation result of the interface state, fusing the obtained multi-period signal characteristic data, carrying out quantitative calculation on the area change of the debonding region, judging the residual distribution state of the bonding strength, and obtaining a comprehensive evaluation report of the damage degree of the bonding interface.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to a full-focus ultrasound imaging method for multimodal damage of composite materials. Background Technology

[0002] Composite materials play an irreplaceable role in aerospace, automotive manufacturing, and other fields, with their lightweight and high-strength properties providing crucial support for modern engineering technology. However, the quality of the adhesive interface in composite materials directly affects the overall performance of the structure. Debonding damage can lead to catastrophic consequences, making the accurate detection and assessment of debonding damage a research hotspot. Currently, although various detection methods are widely used, these methods often exhibit limitations when dealing with complex damage morphologies, especially when the damage boundaries are irregular and the damage degree is uneven. Existing technologies struggle to accurately capture subtle changes in the interface state and cannot effectively adapt to the dynamic characteristics of damage evolution, leading to questions about the reliability of the detection results. Against this backdrop, the research field faces significant technical challenges. A core characteristic of debonding damage at adhesive interfaces is the irregularity of the damaged area shape. This irregularity mainly stems from the uneven distribution of adhesive during the bonding process and the random distribution of stress concentration points within the material. The random distribution of stress concentration points is formed by the combined effects of differences in fiber orientation in the composite material, interlayer interface defects, residual stress during manufacturing, and thermal cycling and mechanical loads under service conditions. This irregularity makes the characteristics of sound wave propagation paths and reflected signals exceptionally complex. When this complexity combines with the dynamic evolution of the debonding region, it further exacerbates the difficulty of interface state identification, thus posing a direct threat to imaging quality. A deeper problem lies in the fact that when the debonding region exhibits partial debonding or gradual expansion, the continuity of the interface contact state is disrupted, and the acoustic characteristics at the boundary change drastically. This change further increases the difficulty of sound wave signal analysis, making traditional fixed sampling methods inadequate for adapting to dynamically changing damage characteristics. Therefore, how to dynamically adjust the spatial sampling strategy of the imaging system to accurately characterize changes in the interface state, and on this basis, achieve precise assessment of the debonding area and reliable judgment of residual adhesion strength, has become a critical problem that urgently needs to be solved. Summary of the Invention

[0003] This invention provides a full-focusing ultrasound imaging method for multimodal damage of composite materials, mainly comprising:

[0004] Initial acoustic signal data of the adhesive interface is acquired, and propagation path deviation information related to the shape of the debonding area is extracted to generate a preliminary damage distribution feature map. Based on the preliminary damage distribution feature map, adaptive mesh generation technology is used to locally refine the suspected debonding area, determine the spatial sampling density distribution scheme, re-acquire acoustic signal data, and perform filtering and enhancement processing to generate an optimized damage area boundary description. Based on the optimized damage area boundary description, feature data of adhesive layer thickness change and contact area reduction at the adhesive interface are extracted, and the change in acoustic wave reflection intensity is calculated to establish a quantitative correspondence between thickness, contact area, and reflection intensity. Based on the quantitative correspondence, the correlation of acoustic wave signal reflection intensity at different times and spatial locations is calculated to generate contact state distribution data between the adhesive layer and the substrate. Based on the contact state distribution data, key feature points of dynamic damage evolution are extracted, and time series analysis is used to track the key feature points to generate predicted data of damage expansion trend. Based on the predicted data and local feature changes of the acoustic signal, the spatial sampling density distribution scheme is optimized to generate dynamic characterization data of the interface state.

[0005] Furthermore, the process of acquiring initial acoustic signal data of the adhesive interface, extracting propagation path deviation information related to the shape of the debonding area, and generating a preliminary damage distribution feature map includes:

[0006] The initial acoustic signal data is decomposed in the time domain to extract amplitude and phase sequences. The amplitude attenuation rate of adjacent sampling points is calculated, and potential damage points with attenuation rates exceeding a threshold are marked. The spatial coordinates of the potential damage points are summarized to generate initial positioning data. Based on the initial positioning data, wavelet transform is used to decompose the acoustic signal, and the energy proportion of wavelet coefficients at each scale is calculated to determine damage boundary points and generate damage contour data. Based on the damage contour data, the difference between the measured signal arrival time and the standard propagation time is calculated to determine the path deviation and generate shape feature data of irregular damage. Based on the shape feature data and the path deviation, spatial interpolation is used to integrate the damage location, contour, and depth information to generate the preliminary damage distribution feature map.

[0007] Furthermore, the step of using wavelet transform to decompose the acoustic signal based on the initial positioning data, calculating the energy proportion of wavelet coefficients at each scale, determining damage boundary points, and generating damage contour data includes:

[0008] Based on the initial positioning data, the acoustic signal is decomposed to generate wavelet coefficients of different scales; the energy proportion of the wavelet coefficients is calculated, the energy proportion difference between adjacent positions is determined, and points with differences exceeding a threshold are marked as damage boundary points; damage boundary points are connected to generate damage contour data.

[0009] Furthermore, based on the optimized description of the damaged area boundary, the feature data of the thickness change and contact area reduction of the adhesive layer at the bonding interface are extracted, the change in acoustic wave reflection intensity is calculated, and a quantitative correspondence between thickness, contact area, and reflection intensity is established, including:

[0010] Based on the optimized damage region boundary description, the coordinates and damage depth data of the detection points are extracted, the difference between the detected thickness and the designed thickness is calculated, and thickness change rate data is generated. Based on the thickness change rate data, the contact area ratio is calculated, the crack propagation direction is determined, and a feature dataset is generated. Based on the feature dataset, the acoustic impedance and reflection coefficient of each detection point are calculated, converted into reflection intensity values, and theoretical reflection intensity data is generated. Based on the theoretical reflection intensity data and the measured signal intensity, the relationship between thickness change rate, contact area ratio, and reflection intensity is fitted to generate the quantitative correspondence.

[0011] Furthermore, the step of calculating the acoustic impedance and reflection coefficient of each detection point based on the feature dataset, converting them into reflection intensity values, and generating theoretical reflection intensity data includes:

[0012] Based on the thickness change rate and contact area ratio in the feature dataset, the acoustic impedance value is calculated; based on the acoustic impedance value, the initial value of the reflection coefficient is calculated, multiplied by the contact area ratio, and a corrected reflection coefficient is generated; based on the corrected reflection coefficient, it is converted into a reflection intensity value to generate the theoretical reflection intensity data.

[0013] Furthermore, the step of extracting key feature points of dynamic damage evolution based on the contact state distribution data, tracking the key feature points using time series analysis methods, and generating predictive data of damage propagation trends includes:

[0014] Based on the contact state distribution data, the boundary line coordinate sequence is extracted, the curvature value is calculated, the curvature maxima and contact state abrupt change points are marked, and a set of key monitoring points is generated; based on the set of key monitoring points, the reflection intensity time series of each monitoring point is extracted, a time evolution model is fitted, and the predicted location and expansion velocity data are generated; based on the predicted location and expansion velocity data, the future damage boundary is calculated, and the predicted data is generated.

[0015] Furthermore, the step of calculating the future damage boundary and generating the prediction data based on the predicted location and propagation velocity data includes:

[0016] Based on the predicted location, the position difference between adjacent time points is calculated to generate expansion velocity data; based on the expansion velocity data, the expansion direction is determined and active expansion areas are marked; based on the active expansion areas, future location points are recursively calculated and connected to form the future damage boundary in the predicted data.

[0017] Furthermore, the step of optimizing the spatial sampling density distribution scheme based on the predicted data and the local feature changes of the acoustic signal to generate dynamic characterization data of the interface state includes:

[0018] Based on the predicted data, the boundary position and expansion velocity are extracted, and the sampling priority value is calculated by combining the main frequency of the acoustic signal and the amplitude attenuation coefficient. Based on the sampling priority value, a sampling density distribution function is generated, and the sampling interval and acquisition frequency are output. Based on the sampling density distribution function, new acoustic signals are acquired, feature parameters are extracted, compared with the predicted feature values, the sampling scheme is adjusted, and dynamic characterization data of the interface state is generated.

[0019] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0020] This invention discloses a full-focusing ultrasonic imaging method for multimodal damage of composite materials. By decomposing the time and frequency domain characteristics of the acoustic signal, propagation path deviation information related to the shape of irregular regions is extracted to obtain a preliminary damage distribution feature map. Then, an adaptive mesh generation technique is used to locally refine the suspected debonding area, determining a dynamic adjustment scheme for the spatial sampling density to obtain more detailed information on acoustic propagation path changes. By establishing a quantitative correspondence between adhesive layer thickness, contact area, and acoustic wave reflection signal intensity, the continuous distribution of the interface contact state is identified. Time series analysis is used to track feature points, obtaining predicted data on damage propagation trends, and optimizing the sampling distribution model accordingly. Finally, multi-time period signal feature data are fused to quantitatively calculate the area change of the debonding region, resulting in a comprehensive assessment report of the degree of adhesive interface damage. This invention enables dynamic detection and assessment of debonding damage at adhesive interfaces, improving detection accuracy and efficiency. Attached Figure Description

[0021] Figure 1 This is a flowchart of a full-focus ultrasound imaging method for multimodal damage of composite materials according to the present invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0023] like Figure 1 This embodiment of a full-focus ultrasound imaging method for multimodal damage of composite materials may specifically include:

[0024] Step S101: For the debonding damage characteristics in the adhesive interface, initial acoustic signal data is obtained. By decomposing the time and frequency domain characteristics of the signal, propagation path deviation information related to the shape of the irregular area is extracted to obtain a preliminary damage distribution feature map.

[0025] Initial acoustic signal data of the adhesive interface is acquired, and the signal is decomposed in the time domain. The amplitude and phase sequences of the signal are extracted using short-time Fourier transform. The amplitude attenuation rate between adjacent sampling points is calculated. When the attenuation rate exceeds a preset threshold, it is marked as a potential damage point. The spatial coordinates of all potential damage points are summarized to obtain the initial location data of the debonding area. Based on the damage area range determined by the initial location data, wavelet transform is performed on the acoustic signal within this range to decompose it into wavelet coefficients at different scales. The energy proportion of the wavelet coefficients at each scale is calculated, and the damage boundary position is determined based on the change in energy proportion. When the energy proportion difference between adjacent positions exceeds a preset threshold, it is determined as a damage boundary point, obtaining damage contour data. For the area delineated by the damage contour data, a ray tracing model of acoustic wave propagation is established. The standard propagation time under no-damage conditions is calculated based on the sound velocity parameters of the adhesive material. The arrival time of the measured signal is compared with the standard propagation time, and the time difference is calculated as the path deviation. The shape characteristics of irregular damage are determined based on the spatial distribution of the path deviation. By combining the shape characteristics of irregular damage and the distribution of path deviation, the Kriging interpolation method is used to spatially interpolate the discrete deviation data. The damage depth at the corresponding location is estimated based on the magnitude of the deviation in the interpolation result. The damage location, shape contour and depth information are integrated to generate a complete damage distribution feature map.

[0026] Specifically, in the detection of debonding damage at adhesive interfaces, time-domain decomposition is fundamental to understanding the characteristics of sound wave propagation.

[0027] Specifically, when sound waves propagate through an adhesive structure, they encounter debonding areas, resulting in reflection and scattering, which alters the amplitude of the received signal. Short-time Fourier transform (SFT) can extract frequency information while maintaining time resolution. By setting an appropriate window function length, such as a Hanning window with 256 sampling points, millisecond-level transient changes can be captured. The amplitude attenuation rate is calculated based on the amplitude ratio of adjacent sampling points. When this ratio is below 0.7, it indicates a significant loss of sound wave energy, which often corresponds to discontinuities within the material.

[0028] It should be noted that wavelet transform has unique advantages in processing non-stationary signals. Wavelet coefficients at different scales reflect the energy distribution characteristics of the signal in different frequency bands. Due to changes in interface properties, decoupling regions can lead to energy concentration or dispersion in specific frequency bands.

[0029] In one possible implementation, a 5-level decomposition using the db4 wavelet basis is employed. The energy proportion changes of the wavelet coefficients in the 3rd and 4th levels are most sensitive; when the energy proportion difference between adjacent positions exceeds 15%, the damage boundary can be accurately located. This method is more stable than traditional threshold judgment because it considers the multi-scale characteristics of the signal. The establishment of the ray tracing model requires consideration of the propagation laws of sound waves in multi-layered media.

[0030] Preferably, Fermat's principle is used to determine the shortest propagation path of the sound wave. In a uniform adhesive layer, the sound velocity is typically 2800-3200 m / s, while in the debonding region, due to the presence of an air layer, the sound velocity drops to around 340 m / s. The detection area is divided into several units through meshing, and each unit is assigned a corresponding sound velocity value. The propagation time from the emission point to the receiving point is calculated using a ray tracing algorithm. The path deviation is quantified by the difference between the measured time and the theoretical time; a larger deviation indicates a more severe debonding region traversed by the path. The Kriging interpolation method performs well in spatial data reconstruction. Based on the theory of variograms, it can predict the values ​​of unknown points based on the spatial correlation of known points.

[0031] In one embodiment, a spherical variogram model is selected, and spatial correlation parameters are determined by calculating the semivariogram between each measurement point. The spatial distribution characteristics of the path deviation are considered during interpolation; there is an approximately linear relationship between the deviation and the damage depth, with each 1 microsecond increase in time deviation corresponding to approximately 0.5 millimeters of damage depth. This quantitative relationship makes it possible to directly deduce the damage depth from the path deviation. The resulting feature map not only shows the planar distribution of the damage but also includes depth information, providing comprehensive data support for subsequent repair decisions.

[0032] Step S102: Based on the preliminary damage distribution feature map, analyze the influence of the irregularity of the adhesive interface on the sound wave propagation speed and attenuation. Adaptive mesh generation technology is used to locally densify the suspected debonding area, determine the dynamic adjustment scheme of the spatial sampling density, and re-acquire sound wave signal data through this scheme. The signal in the high-density area is filtered and enhanced to obtain more refined sound wave propagation path change information and obtain an optimized description of the damage area boundary.

[0033] Based on the preliminary damage distribution feature map, the damage depth value and location coordinates of each pixel are extracted. The damage gradient is calculated by the ratio of the depth difference between adjacent pixels to the distance. The damage gradient is multiplied by the material's sound velocity variation coefficient to obtain the rate of change of sound wave propagation speed. The corresponding signal attenuation coefficient is calculated based on the rate of change. If the attenuation coefficient exceeds a preset threshold, the area is marked as a suspected debonding area, resulting in distribution data containing the location and attenuation coefficient of the suspected area. Based on the distribution data of the suspected areas, the attenuation coefficient of each suspected area is read as the basis for judging mesh refinement. A quadtree meshing method is used to recursively subdivide the detection area. The larger the attenuation coefficient, the deeper the subdivision level, until the mesh size reaches a preset minimum value, generating a spatial sampling scheme containing the coordinates of grid nodes of different densities. Based on the grid node coordinates in the spatial sampling scheme, more sound wave transmitting and receiving sensors are deployed in the dense grid area. Sound wave signals are collected according to the positions determined by the node coordinates, and the signal waveform and arrival time of each sampling point are recorded to form a high-resolution sound wave signal dataset that matches the mesh density. Each signal in the high-resolution acoustic signal dataset is subjected to Butterworth filtering. The arrival time of the first wave of the filtered signal is extracted. The local deflection angle of the acoustic propagation path is calculated by the arrival time difference between adjacent sampling points. The boundary position of the damaged area is identified according to the spatial rate of change of the deflection angle. When the rate of change exceeds a preset threshold, it is determined as a boundary point. Connecting all boundary points yields an optimized description of the boundary of the damaged area.

[0034] Specifically, calculating the damage gradient is a key step in understanding the irregularities of the adhesive interface.

[0035] In one possible implementation, the depth values ​​of adjacent pixels are extracted from the damage distribution feature map. For example, if a point has a depth of 2.5 mm, an adjacent point has a depth of 3.2 mm, and the distance between the two points is 0.8 mm, then the damage gradient is 0.875. This gradient value reflects the drastic change in damage in space. The material sound velocity variation coefficient is usually determined based on the physical properties of the adhesive material. The sound velocity of an epoxy resin layer is approximately 2800 m / s. When debonding occurs, the area is actually an air layer, and the sound velocity drops to 340 m / s, resulting in a sound velocity variation coefficient of 8.2 times. Multiplying the damage gradient by the sound velocity variation coefficient yields the actual rate of change of the sound wave propagation speed, which directly affects the propagation behavior of the sound wave in that area. The core of the quadtree mesh generation method lies in adaptive subdivision based on regional characteristics.

[0036] Specifically, the algorithm starts with the entire detection area, dividing it into four equal sub-regions, and calculating the average attenuation coefficient within each sub-region. When the attenuation coefficient of a sub-region exceeds 0.6, the sub-region is further divided into four equal parts, and this process is repeated until the grid size reaches a minimum limit, such as 0.5 mm. This recursive subdivision method allows severely damaged areas to obtain denser sampling points, while normal areas maintain sparser sampling, ensuring both detection accuracy and efficiency. The subdivision level is typically controlled at 5-7 layers; excessively deep subdivision leads to a sharp increase in computational load with limited improvement in accuracy.

[0037] It should be noted that high-resolution signal acquisition relies on the proper arrangement of sensors. In densely gridded areas, the sensor spacing can be reduced to 1 mm, while in sparse areas, a spacing of 5-10 mm can be maintained. The signal at each sampling point contains multiple parameters: waveform amplitude sequence, first arrival time, and dominant frequency component. Accurate measurement of the first arrival time employs a threshold method combined with correlation analysis; the initial arrival time is recorded when the signal amplitude first exceeds three times the noise level, and then cross-correlation calculations are used to correct the time accuracy to the microsecond level. The Butterworth filter exhibits a flat passband characteristic in signal processing.

[0038] In one embodiment, a fourth-order low-pass filter with a cutoff frequency of 500kHz is designed to effectively remove high-frequency noise while retaining damage characteristic information. The first wave arrival time extracted after filtering is more accurate, and the time difference between adjacent sampling points can reflect local changes in the sound wave path.

[0039] For example, during normal propagation, the time difference between adjacent points is 2.8 microseconds, while after passing through the damaged area, the time difference increases to 4.2 microseconds. Through geometric calculations, the deflection angle of the sound wave propagation path can be obtained as approximately 15 degrees. When the rate of change of the deflection angle between adjacent areas exceeds 10 degrees per millimeter, it indicates that the location is a damage boundary. Connecting these boundary points forms a complete damage profile, and compared to the initial detection results, the boundary positioning accuracy is improved to the sub-millimeter level.

[0040] Step S103: Based on the optimized description of the damage area boundary, analyze the characteristics of the adhesive layer thickness change and contact area reduction in the partially debonded state, as well as the crack propagation direction and velocity characteristics in the propagation state. By calculating the influence of different thickness areas and contact areas on the acoustic wave reflection intensity, establish a quantitative correspondence between adhesive layer thickness, contact area and acoustic wave reflection signal intensity.

[0041] For the optimized description of the damaged area boundary, the coordinates and damage depth data of each detection point within the boundary are extracted. The design thickness in the adhesive material specification is used as a reference benchmark. The difference between the current detected thickness and the design thickness is calculated. The difference is divided by the design thickness to obtain the thickness change rate. The debonding coefficient of each point is determined based on the thickness change rate. The thickness change rates and corresponding coordinates of all detection points are summarized to generate thickness change distribution data of the damaged area. Based on the thickness change distribution data, the actual contact area is calculated according to the inverse relationship between the debonding coefficient and the contact area. When the debonding coefficient is zero, the contact area is a unit area; when the debonding coefficient is one, the contact area is zero. An intermediate value is determined by linear interpolation. The spatial derivative of the contact area difference between adjacent detection points is calculated. The direction with the largest difference is determined as the main direction of crack propagation, resulting in a feature dataset containing the contact area and propagation direction. Based on the thickness change rate and contact area data in the feature dataset, the reflection coefficient of each detection point is calculated using the acoustic impedance formula. Acoustic impedance equals the product of material density and sound velocity. The impedance difference between the adhesive layer and the substrate material is divided by the sum of the impedances to obtain the initial reflection coefficient. This initial value is then multiplied by the contact area ratio to obtain the corrected reflection coefficient. The reflection coefficient is converted into a reflection intensity value to obtain the theoretical reflection intensity of each detection point. By comparing the theoretical reflection intensity with the measured signal intensity, a data matrix containing the thickness change rate, contact area ratio, and measured reflection intensity is constructed. The least squares method is used to fit the relationship among these three factors. After determining the fitting coefficients, a quantitative correspondence between the adhesive layer thickness, contact area, and acoustic wave reflection signal intensity is obtained. R measured The measured reflection intensity is represented by dh / dt, and the thickness change rate is represented by A. contact The formula represents the contact area ratio, and α, β, and γ represent the fitting coefficients. This formula describes the linear relationship between the measured reflection intensity and the thickness change rate and the contact area ratio.

[0042] Specifically, the design thickness in the adhesive material specification sheet is an important reference benchmark for quality control.

[0043] Specifically, the designed thickness of aerospace-grade epoxy resin adhesive layers is typically 0.1-0.3 mm, a value determined through rigorous testing at the factory. When actual testing reveals a thickness of 0.08 mm at a certain point, compared to the designed thickness of 0.2 mm, the thickness variation rate is 0.6%, indicating a loss of 60% of the adhesive material at that location. This thickness variation directly reflects the severity of debonding; a larger thickness variation rate indicates more severe damage to the integrity of the adhesive layer.

[0044] It should be noted that there is a clear physical relationship between the debonding degree coefficient and the contact area. In the fully bonded state, the contact area is the nominal area of ​​the detection unit, and the debonding degree coefficient is zero; in the fully debonded state, the actual contact area drops to zero, and the debonding degree coefficient is one. In the partially debonded state, the contact area is calculated using a linear interpolation method. For example, when the debonding degree coefficient is 0.4, the actual contact area is 60% of the nominal area. Although this linear relationship is a simplified model, it has proven to have sufficient accuracy in engineering applications. Spatial derivative calculation plays a crucial role in identifying crack propagation directions.

[0045] In one possible implementation, the contact area gradient in eight directions is calculated for each detection point, and the direction with the largest gradient value is selected as the main direction of crack propagation. For example, if the contact area of ​​the adjacent point to the east of a certain point is 0.8 square millimeters, to the west it is 0.3 square millimeters, and to the north and south it is 0.5 square millimeters, then the gradient in the east-west direction is the largest, indicating that the crack mainly propagates in the east-west direction. This method can accurately capture the crack propagation trend and provide a basis for subsequent structural safety assessment. The calculation of acoustic impedance involves the basic physical properties of the material.

[0046] Preferably, the acoustic impedance of the aluminum alloy substrate is approximately 17 × 10^6 Rayles, and the acoustic impedance of the epoxy resin adhesive layer is approximately 3 × 10^6 Rayles, with an impedance difference of 14 × 10^6 Rayles and a sum of impedances of 20 × 10^6 Rayles. The initial reflection coefficient is 0.7. When considering the influence of the contact area, if the actual contact area is only 40% of the nominal area, the effective reflection coefficient drops to 0.28. This correction method takes into account the actual situation where some areas within the debonding region still maintain contact, making the prediction of reflection intensity more accurate. The least squares fitting process requires the construction of a reasonable data matrix.

[0047] In one embodiment, data from 100 detection points are collected, with each data point containing three parameters: thickness change rate, contact area ratio, and measured reflection intensity. Fitting coefficients are calculated using matrix operations. Typical fitting results show that reflection intensity has a quadratic relationship with the thickness change rate and a linear relationship with the contact area ratio. Establishing this quantitative relationship allows for the subsequent inference of the actual state of the adhesive layer by measuring the reflection intensity, achieving a shift from qualitative detection to quantitative evaluation and significantly improving the practical value of nondestructive testing.

[0048] Step S104: Using the established quantitative correspondence, calculate the correlation of the acoustic signal reflection intensity at different times and spatial locations, identify the tight contact area and loose separation area between the adhesive layer and the substrate, and determine the continuity distribution of the interface contact state.

[0049] By utilizing the established quantitative correspondence between thickness, contact area, and reflection intensity, the acoustic wave reflection intensity data collected at different times are substituted into the formula to deduce the contact area value at each time. The difference in contact area at the same detection point at adjacent times is calculated, and the difference is divided by the sampling time interval to obtain the contact area change rate. The contact stability is judged based on the change rate; the smaller the change rate, the more stable the contact. This yields the contact stability index for each detection point. Based on the distribution of the contact stability index, the reflection intensity values ​​of adjacent detection points at the same time are extracted, and the absolute value of the difference in reflection intensity between adjacent points is calculated. When the difference is less than a preset threshold and the reflection intensity is greater than the benchmark value, it is determined to be a tight contact area. When the difference is less than the preset threshold and the reflection intensity is less than the benchmark value, it is determined to be a loose separation area, resulting in a spatial classification map of the contact state. Based on the spatial classification map of the contact state, the connectivity between tight contact areas and loose separation areas is statistically analyzed. If there is a continuous path between any two points in the same type of area, the area is determined to be continuously distributed. The ratio of the area of ​​the continuous area to the total detection area is calculated, and the continuity distribution of the interface contact state is judged based on the magnitude of the ratio.

[0050] Specifically, the application of quantitative correspondence is the foundation for achieving accurate assessment of contact status.

[0051] Specifically, the established relationship shows that reflection intensity is positively correlated with contact area and negatively correlated with changes in adhesive layer thickness. When the reflection intensity at a certain test point is measured to be 0.65, substituting it into the relationship allows us to deduce that the actual contact area at that point is approximately 85% of the nominal area. This reverse calculation makes it possible to directly obtain the contact area, which is difficult to measure directly, from the easily measurable reflection intensity, greatly improving detection efficiency.

[0052] It is important to note that the determination of the sampling time interval directly affects the accuracy of contact stability assessment. In fatigue testing of aerospace structures, a typical sampling interval is 0.1 seconds. This timescale captures the slow propagation of fatigue cracks without generating excessive redundant data. For example, if the contact area at a test point is 0.8 square millimeters in the first sampling and 0.75 square millimeters in the second sampling 0.1 seconds later, the rate of change of contact area is 0.5 square millimeters per second. A larger rate of change indicates a more unstable bonding state in that area, potentially indicating fatigue damage or stress concentration. The difference in reflection intensity plays a crucial role in distinguishing different contact states.

[0053] In one possible implementation, adjacent detection points with a reflection intensity difference of less than 0.1 are considered to belong to the same contact state. The baseline value is set based on the typical reflection intensity of a well-bonded state, usually between 0.5 and 0.7. When the reflection intensity of two adjacent points is both greater than 0.6 and the difference is less than 0.1, both points belong to the tight contact area; when the reflection intensity is both less than 0.3 and the difference is less than 0.1, they belong to the loose separation area. This classification method avoids the errors that may be caused by single-point judgment. Connectivity analysis is an important means of evaluating the spatial distribution characteristics of contact states.

[0054] Preferably, the eight-connectivity criterion is used to determine the continuity of a region. That is, if a detection point and any of its eight adjacent points in the surrounding directions belong to the same type of region, a connected path is considered to exist. For example, in a 10×10 detection grid, 65 points were found to belong to closely connected regions. Among these, 60 points form a large connected region, and the other 5 points form two isolated small regions. The main connected region accounts for 92% of the total number of closely connected points, indicating that the closely connected state has good continuity. The calculation of the area ratio provides a quantitative indicator for assessing continuity.

[0055] In one embodiment, the total detection area is 100 square centimeters, of which the largest continuous tight contact area is 72 square centimeters, the largest continuous loose separation area is 15 square centimeters, and the remainder is a transition area. The continuity index of the tight contact area is 0.72, indicating that the adhesive structure maintains a good contact condition overall. When the continuity index is below 0.5, it indicates that the contact condition is fragmented, the structural integrity is seriously threatened, and timely maintenance measures are required. This quantitative assessment method provides a clear basis for judging structural health monitoring.

[0056] Step S105: Based on the continuous distribution of the interface contact state, extract key feature points of dynamic damage evolution, use time series analysis to track and process the feature points, adapt to the dynamically changing damage characteristics, and obtain predicted data of damage expansion trend.

[0057] Based on the continuous distribution of the interface contact state, the coordinate sequence of the boundary line between the tightly contacted area and the loosened separation area is extracted. The curvature of each point on the boundary line is calculated using the three-point circular arc method, that is, the reciprocal of the radius of the arc determined by the current point and the two adjacent points is used as the curvature value. The point with the maximum curvature is marked as the boundary shape abrupt change point. At the same time, the difference of the continuity index of each detection point at adjacent time moments is calculated, and the point with the largest difference is marked as the contact state abrupt change point. The two types of abrupt change points are merged to obtain the set of key monitoring points for damage evolution. For each monitoring point in the set of key monitoring points, its spatial coordinates and reflection intensity values ​​at each sampling time are extracted. The reflection intensity is used as a quantitative index to characterize the contact state. A time-indexed data sequence is constructed, and a second-order autoregressive moving average model is used to fit the data sequence. The autoregressive coefficients and moving average coefficients are determined by the least squares method to obtain the model parameter set describing the time evolution law of each monitoring point. The predicted values ​​of each monitoring point are calculated based on the model parameter set. The difference between the predicted positions at adjacent times is divided by the time interval to obtain the expansion rate. The expansion direction is determined based on the position changes at multiple consecutive times. If the expansion rate exceeds a preset threshold, it is determined to be an active expansion area. The future positions of each monitoring point are recursively calculated using the model parameter set. All predicted position points are connected to form the future damage boundary. The area enclosed by the predicted boundary is calculated to obtain the predicted data of the damage expansion trend.

[0058] Specifically, the three-point circular arc method is simple and efficient in curvature calculation.

[0059] Specifically, for any point on the boundary line, selecting that point and its two adjacent points before and after it uniquely defines an arc. By calculating the radius of this arc and taking its reciprocal, the curvature value at that point is obtained.

[0060] For example, when the coordinates of three consecutive boundary points are (10,20), (12,22), and (15,21), geometric calculations show that the radius of the arc is approximately 5.4 mm, corresponding to a curvature value of 0.185 / mm. A larger curvature value indicates a more severe bending of the boundary at that point, often corresponding to a region of rapid change in damage morphology.

[0061] It should be noted that the identification of abrupt changes in the contact state depends on the temporal variation of the continuity index. The continuity index, defined in the preceding steps, is the ratio of the area of ​​the continuous region to the total area, and its rate of change reflects the severity of damage propagation.

[0062] In one possible implementation, if a detection point has a continuity index of 0.85 during the first sampling and drops to 0.75 during the second sampling, with a time interval of 1 second, then the rate of change of the continuity index at that point is 0.1 per second. When the rate of change exceeds a preset threshold, such as 0.05 per second, it indicates that the adhesion at that location is rapidly deteriorating and should be a key monitoring target. The second-order autoregressive moving average model can effectively capture the trend and periodic characteristics of time series. This model represents the current value as a linear combination of the values ​​from the previous two times, plus the influence of current and previous random disturbances. The model parameters include two autoregressive coefficients and two moving average coefficients; these four parameters together determine the dynamic characteristics of the time series.

[0063] For example, after fitting the reflection intensity sequence of a certain monitoring point, the autoregressive coefficients are 0.7 and 0.2, indicating that the current reflection intensity is mainly affected by the previous moment, and the influence of the previous two moments gradually decreases.

[0064] Preferably, the calculation of the expansion rate needs to take into account the accuracy of the location prediction. Using the determined set of model parameters, the future location of each monitoring point can be recursively predicted.

[0065] For example, if a boundary point is currently located at (20, 30), and the model predicts its next location to be (20.5, 30.8), with a time interval of 1 second, then the expansion rate of this point is approximately 0.94 mm / s. The expansion direction is determined by calculating the angle of the displacement vector; in this example, the expansion direction is approximately 58 degrees north of northeast. When the expansion directions of multiple monitoring points tend to be consistent, it indicates that the damage is preferentially developing along a specific direction.

[0066] In one embodiment, the formation of the predicted boundary requires integrating the predicted positions of all monitoring points. A smooth predicted boundary curve can be obtained by connecting the discrete predicted points using spline interpolation. The region between the predicted boundary and the current boundary is the expected damage propagation area.

[0067] For example, if the current damaged area is 15 square centimeters, and the predicted damaged area in 10 seconds is 18 square centimeters, then the damage propagation rate is 0.3 square centimeters per second. This quantitative prediction of the propagation trend provides a scientific basis for developing maintenance plans, enabling maintenance personnel to take preventative measures before the damage reaches a critical state.

[0068] Step S106: Based on the predicted data of damage propagation trend, the dynamic adjustment scheme of spatial sampling density is adjusted. The predicted data of damage propagation trend and the local feature changes of acoustic signal are used as inputs, and the optimized spatial distribution of sampling points and sampling frequency configuration are output. The acquisition and processing flow of acoustic signal is updated, and the signal feature matching is performed on the real-time changes of damage propagation trend of adhesive interface to obtain the dynamic characterization result of interface state.

[0069] Based on the predicted data of damage propagation trend, the coordinates of the predicted boundary position and the propagation velocity are extracted. Simultaneously, the dominant frequency and amplitude attenuation coefficient of the current acoustic signal at each detection point are obtained as local features. The propagation velocity is normalized by dividing by the maximum velocity value, and the local feature parameters are normalized by dividing by their respective maximum values. A weighted sum is calculated with a weight of 0.6 for the propagation velocity and 0.4 for the local features to obtain the sampling priority value for each detection point. Based on the sampling priority values ​​of each detection point, spatial interpolation is performed using a Gaussian radial basis function. The detection point coordinates are used as the center of the radial basis function, and the sampling priority value is used as the function amplitude. The interpolation coefficients are obtained by solving the interpolation equations, constructing a continuous sampling density distribution function. This function takes arbitrary spatial coordinates as input and outputs the sampling interval time and signal acquisition frequency value at that location. Using the sampling interval and acquisition frequency calculated by the sampling density distribution function, sampling points are added in areas where the sampling interval is less than a preset threshold. New acoustic signal data is acquired according to the calculated acquisition frequency. The feature parameters of the new signal are extracted and compared with the feature values ​​estimated based on the predicted data. If the deviation exceeds the preset threshold, the calculation weight of the sampling priority is adjusted, and the sampling density distribution function is regenerated. By continuously acquiring acoustic signals through the adjusted sampling scheme, the reflection intensity, dominant frequency, and first wave arrival time of each sampling point at each moment are extracted. These three parameters are combined into a feature vector, and the Euclidean distance between the feature vectors of adjacent moments is calculated as the rate of change. The active damage area is determined based on the spatial distribution of the rate of change. Combined with the damage area, boundary shape, and expansion velocity information, the dynamic characterization result of the interface state, including the damage location, morphology, and evolution trend, is obtained.

[0070] Specifically, normalization plays a crucial balancing role in multi-parameter fusion.

[0071] Specifically, the typical range of the expansion speed is 0 to 5 mm / s, while the range of the amplitude attenuation coefficient is 0 to 1, with significant differences in their dimensions and numerical ranges. A normalization method, dividing by their respective maximum values, maps all parameters to a unified range of 0 to 1.

[0072] For example, the expansion velocity at a certain detection point is 3 mm / s, the maximum velocity is 5 mm / s, and the normalized value is 0.6; the amplitude attenuation coefficient is 0.4, which is within the range of 0 to 1. The weights set to 0.6 and 0.4 reflect the importance attached to the dynamic expansion characteristics, because the expansion velocity is directly related to the development trend of the damage.

[0073] It should be noted that the Gaussian radial basis function exhibits good smoothness and locality characteristics in spatial interpolation. The function's shape resembles a bell curve, reaching its maximum value at the center point and smoothly decaying with increasing distance.

[0074] In one possible implementation, the radius of influence of the radial basis function is set to 10 mm. When a point is 5 mm away from the center of the function, the function value decays to about 60%. By arranging radial basis functions at all detection points and adjusting the function amplitude according to the sampling priority, a continuous sampling density field is formed after superposition.

[0075] For example, near the damage boundary, the radial basis functions of multiple high-priority detection points superimpose to form a peak region of sampling density. The practical application of the sampling density distribution function is reflected in dynamic resource allocation.

[0076] Preferably, the sampling interval of the function output ranges from 0.05 to 1 second, and the acquisition frequency ranges from 1 to 20 kHz. When the input coordinates are located at the leading edge of damage propagation, the function may output a sampling interval of 0.1 seconds and an acquisition frequency of 15 kHz, indicating the need for intensive monitoring; while in the stable region, the output may be a 0.8-second interval and a 5 kHz frequency, achieving reasonable resource allocation. Compared with uniform sampling, this adaptive sampling strategy can reduce the amount of data by approximately 40% while ensuring detection accuracy.

[0077] In one embodiment, the deviation of the feature parameters is determined using a relative error criterion. The predicted dominant frequency is 8 kHz, while the measured frequency is 7.2 kHz, resulting in a relative deviation of 10%. When the deviation exceeds a threshold of 15%, it indicates that the current sampling priority calculation may not be accurate enough. Weight adjustment employs a gradient descent approach; if the predicted expansion velocity is too high, its weight is reduced from 0.6 to 0.55, while the weight of local features is correspondingly increased to 0.45. This dynamic adjustment mechanism allows the sampling scheme to adapt to the uncertainties of damage evolution. Euclidean distance, as a measure of feature vector change, has an intuitive geometric meaning. The three-dimensional feature vector includes reflection intensity, dominant frequency, and first wave arrival time; for example, at one moment it is (0.6, 8 kHz, 2.5 μs), and at the next moment it is (0.5, 7.5 kHz, 2.8 μs). The calculated Euclidean distance is approximately 0.58, reflecting the overall degree of change in the state at that point. When the rate of change of multiple points within a certain region exceeds 0.5, it can be identified as a region of active damage. The dynamic characterization results integrate spatial distribution, temporal evolution, and multi-parameter features to form a dataset that comprehensively reflects the interface state, providing a quantitative basis for structural health assessment.

[0078] Step S107: By combining the dynamic characterization results of the interface state with the acquired multi-time period signal feature data, the area change of the debonding area is quantitatively calculated to determine the residual distribution state of the bonding strength and obtain a comprehensive assessment report on the degree of damage to the adhesive interface.

[0079] By dynamically characterizing the interface state, the boundary coordinates, damage location information, and propagation velocity data of the damaged area at each time point are extracted. Simultaneously, various stored characteristic parameter data, including time series of reflection intensity, contact area distribution, and thickness variation information, are read. The boundary coordinates at different times are imported into the area calculation program, and the debonding area at each time point is calculated using the polygon area formula. The area increment is obtained by subtracting the area values ​​from adjacent time points, forming a dataset of debonding area changes over time. Based on the area values ​​and growth rate in the debonding area change dataset, the initial bond strength parameter from the material specification is used as a baseline value. Combined with the contact area ratio data at each monitoring point, the residual bond strength at each point is calculated by multiplying the contact area ratio by the initial strength value. When the residual strength is lower than the threshold ratio of the initial strength, it is marked as a failure point. The number of failure points per unit area is counted to obtain a failure density distribution map. Based on the failure density distribution map and the debonding area change dataset, the debonding area is divided by the total adhesive area to obtain the damage area ratio, and the total number of failure points is divided by the total number of monitoring points to obtain the failure rate. Combined with the average propagation rate, a weighted sum is obtained with a weight of 0.4 for the area ratio, 0.4 for the failure rate, and 0.2 for the propagation rate to obtain a comprehensive index of damage severity. Based on the index value, three damage levels are divided into mild, moderate, and severe. By integrating the damage location coordinates, area ratio, failure rate, propagation rate, and damage level information, a comprehensive assessment report on the degree of damage to the adhesive interface is generated.

[0080] Specifically, polygon area calculation plays a fundamental role in the quantification of damaged areas.

[0081] Specifically, the damaged boundary is usually irregular in shape, consisting of dozens to hundreds of boundary points. Using Green's theorem in a discrete form, the polygon is decomposed into multiple triangles, and the total area is obtained by summing the areas of each triangle.

[0082] For example, if the damage boundary at a certain moment consists of 50 points, and the areas at two adjacent moments are 120 square millimeters and 135 square millimeters respectively, with an area increase of 15 square millimeters, it indicates that the damage is expanding. This precise area calculation lays the data foundation for subsequent quantitative assessment.

[0083] It is important to note that the initial bond strength parameter is a crucial benchmark for evaluating residual strength. In aerospace structural bonding, the initial shear strength of epoxy resin layers is typically 25-35 MPa, a value clearly stated in the material specifications. Residual strength is calculated based on a simple proportional relationship: when the contact area at a point accounts for 70%, its residual strength is approximately 70% of the initial strength, i.e., 17.5-24.5 MPa. When the residual strength is below 40% of the initial strength, that point is considered a failure point because it can no longer withstand the design load. The generation of the failure density distribution map reflects an analytical approach from the local to the global perspective.

[0084] In one possible implementation, the detection area is divided into 10×10 mm grid cells, and the number of failure points within each cell is counted. If a cell has 8 monitoring points, 3 of which are failure points, then the failure density of that cell is 37.5%. Different density levels are visualized using color coding, with dark red representing high failure density areas and light yellow representing low density areas. This spatial distribution information intuitively reflects the degree of damage aggregation.

[0085] Preferably, the weighting settings reflect the contribution of different indicators to the overall damage assessment. The area ratio weight of 0.4 reflects the importance of the macroscopic damage range, the failure rate weight of 0.4 emphasizes the impact of the microscopic failure degree, and the propagation rate weight of 0.2 considers the development trend of the damage.

[0086] For example, a certain adhesive structure has a damaged area ratio of 0.15, a failure rate of 0.25, and a normalized average propagation rate of 0.3. The comprehensive index is calculated as 0.15 × 0.4 + 0.25 × 0.4 + 0.3 × 0.2 = 0.22. According to a pre-defined grading standard, an index less than 0.2 indicates minor damage, 0.2-0.5 indicates moderate damage, and greater than 0.5 indicates severe damage. This structure is therefore classified as moderately damaged.

[0087] In one embodiment, the comprehensive assessment report contains multiple levels of information. The first page presents key indicators: total bonded area 500 square centimeters, current debonded area 75 square centimeters, damaged area ratio 15%, average failure rate 25%, maximum propagation rate 3 mm / day, and overall damage level moderate. Detailed pages include a coordinate table of damage locations, a time evolution curve, and a failure density distribution map. This multi-dimensional assessment provides comprehensive technical support for maintenance decisions, enabling maintenance personnel to accurately grasp the structural health status and formulate appropriate maintenance plans.

[0088] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the concept of this application. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this application.

Claims

1. A method for full-focused ultrasound imaging of multimodal damage in composite materials, characterized in that, The method includes: Initial acoustic signal data of the adhesive interface is acquired, and propagation path deviation information related to the shape of the debonding area is extracted to generate a preliminary damage distribution feature map. Based on the preliminary damage distribution feature map, adaptive mesh generation technology is used to locally refine the suspected debonding area, determine the spatial sampling density distribution scheme, re-acquire acoustic signal data, and perform filtering and enhancement processing to generate an optimized damage area boundary description. Based on the optimized damage area boundary description, feature data of adhesive layer thickness change and contact area reduction at the adhesive interface are extracted, and the change in acoustic wave reflection intensity is calculated to establish a quantitative correspondence between thickness, contact area, and reflection intensity. Based on the quantitative correspondence, the correlation of acoustic wave signal reflection intensity at different times and spatial locations is calculated to generate contact state distribution data between the adhesive layer and the substrate. Based on the contact state distribution data, key feature points of dynamic damage evolution are extracted, and time series analysis is used to track the key feature points to generate predicted data of damage expansion trend. Based on the predicted data and local feature changes of the acoustic signal, the spatial sampling density distribution scheme is optimized to generate dynamic characterization data of the interface state.

2. The method for full-focus ultrasound imaging of multimodal damage in composite materials according to claim 1, characterized in that, The process of acquiring initial acoustic signal data of the adhesive interface, extracting propagation path deviation information related to the shape of the debonding area, and generating a preliminary damage distribution feature map includes: The initial acoustic signal data is decomposed in the time domain to extract amplitude and phase sequences. The amplitude attenuation rate of adjacent sampling points is calculated, and potential damage points with attenuation rates exceeding a threshold are marked. The spatial coordinates of the potential damage points are summarized to generate initial positioning data. Based on the initial positioning data, wavelet transform is used to decompose the acoustic signal, and the energy proportion of wavelet coefficients at each scale is calculated to determine damage boundary points and generate damage contour data. Based on the damage contour data, the difference between the measured signal arrival time and the standard propagation time is calculated to determine the path deviation and generate shape feature data of irregular damage. Based on the shape feature data and the path deviation, spatial interpolation is used to integrate the damage location, contour, and depth information to generate the preliminary damage distribution feature map.

3. The method for full-focus ultrasound imaging of multimodal damage in composite materials according to claim 2, characterized in that, The step of using wavelet transform to decompose the acoustic signal based on the initial positioning data, calculating the energy proportion of wavelet coefficients at each scale, determining damage boundary points, and generating damage contour data includes: Based on the initial positioning data, the acoustic signal is decomposed to generate wavelet coefficients of different scales; the energy proportion of the wavelet coefficients is calculated, the energy proportion difference between adjacent positions is determined, and points with differences exceeding a threshold are marked as damage boundary points; damage boundary points are connected to generate damage contour data.

4. The method for full-focus ultrasound imaging of multimodal damage in composite materials according to claim 1, characterized in that, Based on the optimized description of the damaged area boundary, the feature data of the thickness change and contact area reduction of the adhesive layer at the bonding interface are extracted, the change in acoustic wave reflection intensity is calculated, and a quantitative correspondence between thickness, contact area, and reflection intensity is established, including: Based on the optimized damage region boundary description, the coordinates and damage depth data of the detection points are extracted, the difference between the detected thickness and the designed thickness is calculated, and thickness change rate data is generated. Based on the thickness change rate data, the contact area ratio is calculated, the crack propagation direction is determined, and a feature dataset is generated. Based on the feature dataset, the acoustic impedance and reflection coefficient of each detection point are calculated, converted into reflection intensity values, and theoretical reflection intensity data is generated. Based on the theoretical reflection intensity data and the measured signal intensity, the relationship between thickness change rate, contact area ratio, and reflection intensity is fitted to generate the quantitative correspondence.

5. The method for full-focus ultrasound imaging of multimodal damage in composite materials according to claim 4, characterized in that, The step of calculating the acoustic impedance and reflection coefficient of each detection point based on the feature dataset, converting them into reflection intensity values, and generating theoretical reflection intensity data includes: Based on the thickness change rate and contact area ratio in the feature dataset, the acoustic impedance value is calculated; based on the acoustic impedance value, the initial value of the reflection coefficient is calculated, multiplied by the contact area ratio, and a corrected reflection coefficient is generated; based on the corrected reflection coefficient, it is converted into a reflection intensity value to generate the theoretical reflection intensity data.

6. The method for full-focus ultrasound imaging of multimodal damage in composite materials according to claim 1, characterized in that, The step of extracting key feature points of dynamic damage evolution based on the contact state distribution data, tracking the key feature points using time series analysis methods, and generating predictive data of damage propagation trends includes: Based on the contact state distribution data, the boundary line coordinate sequence is extracted, the curvature value is calculated, the curvature maxima and contact state abrupt change points are marked, and a set of key monitoring points is generated; based on the set of key monitoring points, the reflection intensity time series of each monitoring point is extracted, a time evolution model is fitted, and the predicted location and expansion velocity data are generated; based on the predicted location and expansion velocity data, the future damage boundary is calculated, and the predicted data is generated.

7. The method for full-focus ultrasound imaging of multimodal damage in composite materials according to claim 6, characterized in that, The step of calculating the future damage boundary and generating the predicted data based on the predicted location and propagation velocity data includes: Based on the predicted location, the position difference between adjacent time points is calculated to generate expansion velocity data; based on the expansion velocity data, the expansion direction is determined and active expansion areas are marked; based on the active expansion areas, future location points are recursively calculated and connected to form the future damage boundary in the predicted data.

8. The method for full-focus ultrasound imaging of multimodal damage in composite materials according to claim 1, characterized in that, The step of optimizing the spatial sampling density distribution scheme based on the predicted data and local feature changes of the acoustic signal to generate dynamic characterization data of the interface state includes: Based on the predicted data, the boundary position and expansion velocity are extracted, and the sampling priority value is calculated by combining the main frequency of the acoustic signal and the amplitude attenuation coefficient. Based on the sampling priority value, a sampling density distribution function is generated, and the sampling interval and acquisition frequency are output. Based on the sampling density distribution function, new acoustic signals are acquired, feature parameters are extracted, compared with the predicted feature values, the sampling scheme is adjusted, and dynamic characterization data of the interface state is generated.

Images