A non-destructive testing method and system based on polyester fiber material

Abstract

The present application relates to the technical field of polyester fiber ultrasonic detection, in particular to a nondestructive testing method and system based on polyester fiber material, comprising: collecting initial ultrasonic echo time domain signals of different detection points of polyester fiber material, processing through an improved ultrasonic signal reconstruction algorithm of fusion adaptive noise suppression model and waveform matching criterion, and generating characteristic enhanced ultrasonic reconstruction signals. Relying on the structure feature library of polyester fiber material, the reconstruction signals are analyzed layer by layer to generate a multi-layer structure feature mapping containing fiber orientation distribution, interface adhesion state and internal defect mask. Through the material performance degradation model integrating the mapping relationship of defect size, density and stress concentration coefficient, the defect related performance degradation parameters are calculated, the material structure integrity is evaluated in combination with the preset structure safety threshold, and the corresponding evaluation report is generated. The present application improves the signal feature recognition degree, realizes the synchronous analysis of multi-dimensional structure features, and accurately reflects the internal structure state of the material.

Description

Technical Field

[0001] This invention relates to the field of ultrasonic testing technology for polyester fibers, and in particular to a non-destructive testing method and system based on polyester fiber materials. Background Technology

[0002] Conventional non-destructive testing of polyester fiber materials employs ultrasonic testing. This involves acquiring ultrasonic echo time-domain waveform data from different testing points, performing basic signal noise reduction using a general filtering algorithm, and then identifying internal defects through a simplified analytical method. However, this testing method lacks dedicated signal processing logic tailored to the characteristics of polyester fibers. It only utilizes general ultrasonic signal processing techniques for preliminary signal processing, and the signal analysis stage lacks a structural feature database specific to polyester fibers. Defect identification and material structural feature analysis are therefore performed separately.

[0003] Existing ultrasonic testing technologies for polyester fibers suffer from insufficient adaptability in signal processing. General noise suppression methods cannot match the waveform characteristics of ultrasonic echoes from polyester fiber materials, easily leading to weakening of effective structural features in the signal. Furthermore, there is no unified standard for waveform matching, resulting in low structural feature recognition after signal reconstruction. Signal analysis can only extract single information about internal defects, failing to simultaneously acquire data on fiber orientation distribution and interfacial bonding status. The correlation analysis between defects and material properties relies solely on single defect parameters, neglecting to integrate the mapping relationship between defect size, defect density, and stress concentration factor, resulting in poor accuracy in structural integrity evaluation.

[0004] In the process of ultrasonic testing of polyester fiber materials, it is necessary to use an appropriate ultrasonic signal reconstruction method to process the original time-domain waveform data, rely on a dedicated material structure feature library to complete multi-dimensional structural feature analysis, and generate a feature mapping map containing multiple types of structural information to make up for the shortcomings of existing detection technologies in signal processing and feature analysis. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the existing technology and to propose a non-destructive testing method and system based on polyester fiber materials.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a non-destructive testing method based on polyester fiber materials, comprising: Acquire an initial set of ultrasonic echo signals of polyester fiber material, wherein the initial set of ultrasonic echo signals contains original time-domain waveform data at different detection points; An improved ultrasonic signal reconstruction algorithm is applied to the initial ultrasonic echo signal set to generate a feature-enhanced ultrasonic reconstructed signal. The improved ultrasonic signal reconstruction algorithm processes the original time-domain waveform data based on an adaptive noise suppression model and a waveform matching criterion. The ultrasonic reconstruction signal enhanced by the feature enhancement is subjected to hierarchical analysis using a polyester fiber material structural feature library to generate a multi-layer structural feature map, which includes fiber orientation distribution, interface adhesion state and internal defect mask. Based on the multi-layer structure feature mapping map, defect-related performance degradation parameters are calculated through a material performance degradation model, which integrates the mapping relationship between defect size, defect density and stress concentration factor. By combining the defect-related performance degradation parameters with the predefined structural safety threshold, the structural integrity of polyester fiber materials is graded and evaluated, generating a structural integrity evaluation report that includes defect level, location coordinates, and risk level.

[0007] As a further aspect of the present invention, an improved ultrasound signal reconstruction algorithm is applied to the initial ultrasound echo signal set to generate a feature-enhanced ultrasound reconstruction signal, including: The original time-domain waveform data is decomposed into multiple sub-frequency band signal components. For each sub-band signal component, the noise energy distribution of the sub-band in the time domain is calculated according to the adaptive noise suppression model, and the standard waveform with the highest similarity to the sub-band signal component is selected from the preset reference waveform library according to the waveform matching criterion. Based on the calculated noise energy distribution, adaptive threshold noise reduction processing is performed on the sub-frequency band signal components to obtain a preliminarily denoised sub-frequency band signal. The preliminarily denoised sub-band signal is time-domain aligned and amplitude-normalized with the selected standard waveform; Based on the similarity metric defined by the waveform matching criterion, calculate the phase difference and waveform distortion coefficient between the initially denoised sub-band signal and the standard waveform; The phase difference is used to perform phase correction on the sub-band signal of the initial denoising, and the waveform distortion coefficient is used to perform morphological compensation on the waveform of the corrected signal. The signals from all sub-bands, after phase correction and morphological compensation, are combined to generate the enhanced ultrasonic reconstructed signal.

[0008] As a further aspect of the present invention, calculating the noise energy distribution of the sub-frequency band in the time domain based on the adaptive noise suppression model includes: In the time-domain sequence of the sub-frequency band signal components, a sliding time window is used to extract multiple local signal segments; Power spectral density estimation is performed on each local signal segment. Frequency components in the power spectrum below a preset energy threshold are identified as noise components, and frequency components above the preset energy threshold are identified as signal components. The energy values ​​of all frequency components identified as noise components in each local signal segment are accumulated and calculated as the local noise energy of the local signal segment. Using the local noise energy as the vertical axis and the center time of the sliding time window as the horizontal axis, a noise energy distribution curve of the sub-frequency band in the time domain is constructed. The noise energy distribution curve is smoothed, and the smoothed curve is used as the output of the adaptive noise suppression model to guide the dynamic setting of the threshold in the adaptive threshold noise reduction process.

[0009] As a further aspect of the present invention, the step of calculating the phase difference and waveform distortion coefficient between the preliminarily denoised sub-band signal and the standard waveform based on the similarity metric defined by the waveform matching criterion is achieved through the following steps: The Hilbert transform is performed on the preliminarily denoised sub-band signal and the selected standard waveform respectively to obtain the analytical signal form of the two signals; The instantaneous phase sequences of the two signals are extracted from the analytical signals of the sub-band signals after preliminary denoising and the analytical signals of the standard waveform, respectively. Calculate the difference between the instantaneous phase sequence of the preliminarily denoised sub-band signal and the instantaneous phase sequence of the standard waveform at each sampling time point; The arithmetic mean of the phase difference values ​​at all sampling time points is calculated, and the arithmetic mean is used as the phase difference between the sub-frequency band signal of the initial denoising and the standard waveform; The sub-band signal and the standard waveform after preliminary denoising are respectively normalized for amplitude processing to ensure that the maximum amplitude of the two signals is consistent. In the time domain, using a sliding window of fixed length, the cross-correlation coefficient between the preliminarily denoised sub-band signal and the standard waveform within the corresponding window signal segment is calculated. Find the relative time shift between windows that makes the cross-correlation coefficient reach its maximum value, and perform time shift alignment operation on the standard waveform based on this relative time shift; On the aligned signal, the root mean square value of the amplitude difference between the corresponding sampling points of the preliminarily denoised sub-band signal and the standard waveform is calculated, and the root mean square value is divided by the root mean square value of the amplitude of the standard waveform. The resulting ratio is used as the waveform distortion coefficient.

[0010] As a further aspect of the present invention, the step of using a polyester fiber material structural feature library to perform layered analytical processing on the feature-enhanced ultrasonic reconstructed signal to generate a multi-layer structural feature mapping map includes: From the polyester fiber material structural feature library, a standard ultrasonic response template corresponding to the model of the polyester fiber material to be tested is retrieved. The standard ultrasonic response template contains ideal reflection waveform features of different material interfaces from the surface layer to the deep layer. The feature-enhanced ultrasonic reconstructed signal is subjected to point-by-point correlation analysis with the standard ultrasonic response template in the time domain to identify the position of the reflection peak in the feature-enhanced ultrasonic reconstructed signal corresponding to each material interface. Based on the identified positions of each reflected wave peak, the enhanced ultrasonic reconstructed signal is divided into multiple signal segments in the time domain, with each signal segment corresponding to a structural layer of the polyester fiber material. For each signal segment, its envelope, center frequency offset, and signal attenuation rate are extracted to form the primary feature vector of the structural layer. The primary feature vector of each structural layer is input into a pre-trained structural state classification model, which outputs the fiber orientation distribution state score, interface adhesion state score, and probability of internal defect existence of the structural layer. All detection points, fiber orientation distribution status scores, interface adhesion status scores, and internal defect existence probabilities of all structural layers are interpolated and rendered in a three-dimensional spatial grid to generate independent fiber orientation distribution maps, interface adhesion status maps, and internal defect masks, which together constitute the feature mapping map of the multi-layer structure.

[0011] As a further aspect of the present invention, the step of dividing the feature-enhanced ultrasonic reconstructed signal into multiple signal segments in the time domain based on the identified positions of each reflection peak includes: The midpoint between two adjacent reflected wave peak positions is used as the dividing point; The portion of the signal preceding the first reflected wave peak is taken as the signal segment of the corresponding material surface layer; Between two adjacent reflection peaks, the signal segment in the middle is taken as the signal segment of the corresponding internal structural layer, with the dividing point closer to the previous reflection peak as the starting point and the dividing point closer to the next reflection peak as the ending point. The signal portion following the last reflected wave peak is taken as the signal segment of the corresponding deep or bottom layer of the material; Each segmented signal is labeled with its corresponding structural layer number and time domain start and end positions.

[0012] As a further aspect of the present invention, based on the multilayer structure feature mapping map, defect-related performance degradation parameters are calculated using a material performance degradation model, including: From the internal defect mask, all independent defect regions are identified, and the equivalent diameter, area, and projected length in the thickness direction of each independent defect region are calculated. From the fiber orientation distribution map, read the average fiber orientation angle and its distribution dispersion at the location of each independent defect region; From the interface adhesion state diagram, read the adhesion strength score of the material interface adjacent to each individual defect area; The equivalent diameter, area, projected length, average fiber orientation angle, distribution dispersion, and bond strength score are used as input features for the material performance degradation model. The material property degradation model calculates a local stress concentration factor and a strength reduction factor for each independent defect region based on the pre-learned mapping relationship. Summarize the local stress concentration factors of all independent defect areas, and calculate the overall maximum stress concentration factor and average stress concentration factor; Summarize the strength reduction factors of all independent defect areas, and calculate the overall average strength reduction factor and the strength reduction factor of the weakest area; The overall maximum stress concentration factor, average stress concentration factor, overall average strength reduction factor, and strength reduction factor of the weakest region are collectively used as the defect-related performance degradation parameter.

[0013] As a further aspect of the present invention, the calculation of the equivalent diameter, area, and projected length in the thickness direction of each independent defect region includes: Perform connected component analysis on the internal defect mask and mark each independent connected region as an independent defect region; Calculate the total pixel area of ​​each independent defect region, and convert the pixel area into the actual physical area according to the preset spatial resolution; The diameter of a circle with the same actual physical area is taken as the equivalent diameter of the independent defect region. Project each individual defect region along a direction perpendicular to the material surface, i.e., the thickness direction, and calculate the maximum length of the projected region in the thickness direction as the projection length.

[0014] As a further aspect of the present invention, the structural integrity of polyester fiber materials is graded and evaluated by combining the defect-related performance degradation parameters with a predefined structural safety threshold, including: From the predefined structural safety threshold table, read the allowable stress concentration factor threshold and allowable strength reduction factor threshold that match the current polyester fiber material grade, heat treatment state and usage environment; The overall maximum stress concentration factor in the defect-related performance degradation parameter is compared with the allowable stress concentration factor threshold. Compare the strength reduction factor of the weakest region in the defect-related performance degradation parameter with the allowable strength reduction factor threshold. If the overall maximum stress concentration factor does not exceed the allowable stress concentration factor threshold, and the strength reduction factor of the weakest region is not lower than the allowable strength reduction factor threshold, then the structural integrity is determined to be at a safe level. If the overall maximum stress concentration factor exceeds the allowable stress concentration factor threshold, or the strength reduction factor of the weakest region is lower than the allowable strength reduction factor threshold, but the average stress concentration factor and the overall average strength reduction factor in the defect-related performance degradation parameters are still within acceptable ranges, then the structural integrity is determined to be at the level of concern. If the overall maximum stress concentration factor exceeds the allowable stress concentration factor threshold, or the strength reduction factor of the weakest region is lower than the allowable strength reduction factor threshold, then the structural integrity is determined to be at risk level. Based on the determined safety level, attention level, or risk level, and combined with the location coordinates of the defects in the internal defect mask, a structural integrity evaluation report containing the defect level, location coordinates, and risk level is generated.

[0015] As a further aspect of the present invention, the present invention also includes a non-destructive testing system based on polyester fiber materials. The system includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of the non-destructive testing method based on polyester fiber materials as described above.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows: An improved ultrasonic signal reconstruction algorithm is applied to the original time-domain waveform data based on an adaptive noise suppression model and waveform matching criteria. This algorithm effectively suppresses noise by conforming to the characteristic patterns of ultrasonic echo signals from polyester fiber materials, eliminating irrelevant interference noise components. Simultaneously, it corrects the distortion of the original time-domain waveform according to the waveform matching criteria, preserving subtle feature information related to the fiber structure and interface state. The integrity of the reconstructed ultrasonic signal is maintained, and the clarity of various structural features within the signal is improved. This avoids the obscuring of effective features or waveform distortion that occurs in conventional signal processing, resulting in a higher degree of correspondence between the signal features and the actual structural state of the polyester fiber material.

[0017] By utilizing a structural feature library of polyester fibers, layered analytical processing is performed on the enhanced ultrasonic reconstructed signal. This allows for one-to-one matching of the reconstructed signal features with standard data within the feature library, simultaneously extracting fiber orientation distribution, interfacial adhesion state, and internal defect masking information. Multiple structural features are then integrated to form a multi-layered structural feature map. This analytical method overcomes the limitations of conventional single-feature extraction, presenting multi-dimensional structural features simultaneously in the same map. Defect information is correlated with fiber arrangement and interfacial adhesion information, and the feature map comprehensively reflects the overall internal structural state of the material, expanding the analytical dimensions and completeness of structural features. Attached Figure Description

[0018] Figure 1 This is a flowchart of a non-destructive testing method based on polyester fiber materials according to the present invention; Figure 2 A flowchart illustrating the generation of feature-enhanced reconstructed ultrasound signals using an improved ultrasound signal reconstruction algorithm; Figure 3 This is a flowchart for calculating the noise energy distribution of a sub-band in the time domain based on an adaptive noise suppression model. Detailed Implementation

[0019] 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.

[0020] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0021] See Figure 1 This invention provides a non-destructive testing method based on polyester fiber materials, the specific method including: The process begins by acquiring an initial set of ultrasonic echo signals from polyester fiber materials. This set encompasses raw time-domain waveform data from different detection points. An improved ultrasonic signal reconstruction algorithm is then applied to this initial set of ultrasonic echo signals. This algorithm relies on an adaptive noise suppression model and waveform matching criteria to process the raw data, thereby generating a feature-enhanced ultrasonic reconstruction signal. A pre-built structural feature library of polyester fiber materials is used to perform layered analytical processing on the ultrasonic reconstruction signal, obtaining a multi-layered structural feature map that includes fiber orientation distribution, interfacial adhesion state, and internal defect masks. This map is then input into a material performance degradation model, which integrates the mapping relationship between defect size, defect density, and stress concentration factor, thereby calculating defect-related performance degradation parameters. Combining these parameters with a predefined structural safety threshold, a graded evaluation of the structural integrity of the polyester fiber material is completed, outputting a structural integrity evaluation report covering defect level, location coordinates, and risk level.

[0022] In one embodiment of the present invention, see [reference] Figure 2 The original time-domain waveform data is decomposed into multiple sub-band signal components. For each sub-band signal component, its noise energy distribution in the time domain is calculated based on an adaptive noise suppression model. Simultaneously, a standard waveform with the highest similarity to the sub-band signal component is selected from a preset reference waveform library according to a waveform matching criterion. Based on the calculated noise energy distribution, adaptive threshold denoising is applied to the sub-band signal component to obtain a preliminary denoised sub-band signal. The preliminary denoised signal is then time-domain aligned and amplitude normalized with the selected standard waveform. According to the similarity metric defined by the waveform matching criterion, the phase difference and waveform distortion coefficient between the preliminary denoised sub-band signal and the standard waveform are calculated. The obtained phase difference is used to perform phase correction on the preliminary denoised signal, and the waveform distortion coefficient is used to perform morphological compensation on the corrected signal. The signals of all sub-bands after phase correction and morphological compensation are then band-synthesized to form a feature-enhanced ultrasonic reconstructed signal.

[0023] In practice, the process of applying the improved ultrasonic signal reconstruction algorithm to the initial ultrasonic echo signal set is implemented through a signal processing system. This system is equipped with a signal acquisition module, a frequency band decomposition filter bank, and an adaptive noise suppression module. The initial ultrasonic echo signal set originates from ultrasonic scanning of a polyester fiber material sample. The scan covers a pre-defined grid array of detection points on the sample surface, and each detection point records a segment of original time-domain waveform data containing complete reflection information. The frequency band decomposition process uses a set of orthogonal mirror filters to decompose the original time-domain waveform data, uniformly dividing the effective bandwidth of the original signal into several interconnected and non-overlapping sub-bands. Each sub-band signal component carries the ultrasonic response characteristics within a specific frequency range.

[0024] In some embodiments, the reference waveform library is stored in non-volatile memory and contains typical ultrasonic reflection waveforms of defect-free standard polyester fiber material samples prepared under different process conditions. Each standard waveform is labeled with the corresponding material type, detection frequency, and incident angle information. The waveform matching criterion uses the normalized cross-correlation coefficient as a similarity index, calculates the cross-correlation coefficient between the sub-band signal component and each candidate standard waveform in the reference waveform library, and selects the candidate standard waveform that maximizes the cross-correlation coefficient as the matching result.

[0025] In practical implementation, the adaptive noise suppression model relies on local spectral analysis within a sliding time window. For each sub-band signal component, a sliding window with a fixed step size is used along the time axis, with the window length set to cover at least two complete ultrasound cycles. The power spectral density is calculated using a Fast Fourier Transform (FFT) on the local signal segment within the window. A preset energy threshold is set based on the average power level of the background noise in the defect-free region. The sum of the energy of frequency components below the threshold in the local power spectrum is statistically analyzed and taken as the local noise energy corresponding to the center time of that window. The center times of each window are fitted with the corresponding local noise energy values ​​to generate a continuous curve characterizing the noise energy change over time. This curve is then filtered using a moving average to remove random fluctuations, resulting in a smooth noise energy distribution curve.

[0026] Optionally, adaptive threshold denoising dynamically adjusts the threshold value based on the noise energy distribution curve. A higher threshold is used during periods of high noise energy to suppress strong interference, while a lower threshold is used during periods of low noise energy to preserve weak signal characteristics. The threshold setting function is defined as:

[0027] in: Indicates time The noise reduction threshold, The smoothed noise energy distribution curve at time [time value missing] The value, This is the proportionality coefficient. This is a constant bias term. For sub-band signal components with amplitudes lower than... The sample points are either zeroed out or attenuated, and the amplitude is higher than The sample points are retained to obtain the initial denoised sub-band signal.

[0028] In practice, time-domain alignment is achieved by calculating the cross-correlation function between the preliminarily denoised sub-band signal and the selected standard waveform. The time delay corresponding to the peak value of the cross-correlation function is then identified, and the standard waveform is shifted along the time axis by this time delay to achieve precise alignment. Amplitude normalization linearly scales the amplitudes of the two signals, unifying their maximum absolute amplitude to the same reference value. The similarity metrics defined by the waveform matching criterion include two indicators: phase difference and waveform distortion coefficient. The phase difference reflects the overall phase shift, while the waveform distortion coefficient reflects shape differences.

[0029] The Hilbert transform is used to convert a real signal into an analytic signal. The real part of the analytic signal is the original signal, and the imaginary part is the result of the Hilbert transform. Instantaneous phase sequences are extracted from the analytic signals of the initially denoised sub-band signals and from the analytic signals of the standard waveform. The difference between the two sequences at the same sampling time point is calculated, and the arithmetic mean of the phase differences at all sampling points is taken. This average value is the phase difference between the initially denoised sub-band signals and the standard waveform. Phase correction subtracts this phase difference from the phase of the initially denoised sub-band signals to achieve phase synchronization with the standard waveform.

[0030] In practice, the waveform distortion coefficient is calculated based on amplitude normalization and time-shift alignment. A fixed-length sliding window slides synchronously along the aligned two signals, and the cross-correlation coefficient between the two signal segments within the window is calculated. The relative time shift between windows that maximizes the cross-correlation coefficient is found, and the time shift position of the standard waveform is further fine-tuned to ensure optimal local alignment. On the aligned signal, the amplitude difference between the initially denoised sub-band signal and the standard waveform is calculated point by point. The root mean square (RMS) values ​​of these differences are obtained, and then the RMS value is divided by the RMS amplitude value of the standard waveform. The resulting dimensionless ratio is the waveform distortion coefficient. Morphological compensation applies a correction weight proportional to the morphological difference of the standard waveform to the phase-corrected signal based on the magnitude of the waveform distortion coefficient to reduce waveform distortion.

[0031] In some embodiments, bandgap synthesis employs a synthesis filter bank that is the inverse of bandgap decomposition. All sub-band signals, after phase correction and morphological compensation, are used as input to reconstruct a complete broadband time-domain signal. The channel gain of the synthesis filter bank is adjusted according to the signal-to-noise ratio (SNR) of each sub-band; sub-bands with higher SNR are assigned higher gain, and sub-bands with lower SNR are assigned lower gain. The synthesized signal is the feature-enhanced ultrasonic reconstructed signal, which retains the key structural features of the original signal while significantly reducing the influence of random noise and mode noise.

[0032] In one embodiment of the present invention, see [reference] Figure 3In the time-domain sequence of the sub-frequency band signal components, a sliding time window is used to extract several local signal segments. Power spectral density estimation is performed on each local signal segment, and frequency components with power spectral density below a preset energy threshold are classified as noise components, while those above the threshold are considered as signal components. The energy values ​​of all frequency components judged as noise components in each local signal segment are accumulated and calculated as the local noise energy of that segment. The noise energy distribution curve of the sub-frequency band in the time domain is plotted with the local noise energy as the vertical axis and the center time of the sliding window as the horizontal axis. After smoothing the curve, it is used as the output of the adaptive noise suppression model to guide the dynamic setting of the threshold in the adaptive threshold noise reduction process. Furthermore, the steps for calculating the phase difference and waveform distortion coefficient between the initial denoised sub-band signal and the standard waveform include: performing Hilbert transforms on both signals to obtain analytic signal forms, and extracting their respective instantaneous phase sequences; calculating the phase difference between the two sequences at each sampling point, and taking their arithmetic mean as the phase difference between them; normalizing the amplitude of the two signals to make their maximum amplitudes consistent, and calculating the cross-correlation coefficient of the signal segments within the corresponding window using a fixed-length sliding window in the time domain; determining the relative time shift between windows when the cross-correlation coefficient is maximum, and aligning the standard waveform accordingly; calculating the root mean square value of the amplitude difference between the corresponding sampling points of the two signals on the aligned signal, and then dividing it by the root mean square value of the amplitude of the standard waveform; the resulting ratio is the waveform distortion coefficient.

[0033] In practical implementation, the digital signal processor (DSP) is responsible for calculating the noise energy distribution of sub-bands in the time domain based on the adaptive noise suppression model. The DSP receives sub-band signal components from the frequency band decomposition module; these components are discrete time series in complex form. The sliding time window is set to cover at least ten signal cycles corresponding to the center frequency of the sub-band, and the window sliding step is set to one-quarter of the window length to ensure sufficient overlap between local signal segments. For each truncated local signal segment, the Welch method is used to estimate the power spectral density, dividing the frequency interval into several narrowband units. The preset energy threshold is three times the average power spectral density of the defect-free reference area. The power value within each narrowband unit is statistically analyzed, and units with power values ​​below the preset energy threshold are designated as noise-dominant units. The sum of the power values ​​of these units is accumulated as the local noise energy. Discrete noise energy distribution data points are generated using the center time of the sliding time window as the independent variable and the corresponding local noise energy as the dependent variable. A continuous noise energy distribution curve is then fitted using cubic spline interpolation. The fitted curve is smoothed using a low-pass filter with a cutoff frequency of one-fifth of the sub-band bandwidth to eliminate spike fluctuations caused by sudden interference. The smoothed noise energy distribution curve is the final output of the adaptive noise suppression model, which is used to guide the dynamic adjustment of the threshold in the subsequent adaptive threshold noise reduction process.

[0034] In some embodiments, the operation of calculating the phase difference between the initially denoised sub-band signal and the standard waveform is performed by an analytical signal processing unit. Hilbert transforms are applied to both the initially denoised sub-band signal and the standard waveform to obtain analytical signal forms for both signals. Instantaneous phase sequences are extracted from the analytical signals. The instantaneous phase is obtained by performing an arctangent operation on the ratio of the imaginary to the real part of the analytical signal, and the calculation result is mapped to the range of 0 to 2π to avoid phase jumps. The difference between the instantaneous phase sequence of the initially denoised sub-band signal and the instantaneous phase sequence of the standard waveform at each sampling time point is calculated. If the phase difference exceeds π, it is adjusted to the range of -π to π by adding or subtracting 2π. The arithmetic mean of the phase differences at all sampling time points is calculated; this arithmetic mean is the constant phase difference between the initially denoised sub-band signal and the standard waveform. The phase correction operation subtracts this phase difference from the instantaneous phase sequence of the initially denoised sub-band signal, achieving overall phase alignment between the two signals.

[0035] In practice, the waveform distortion coefficient calculation is based on amplitude normalization and fine time-shift alignment. The amplitude of the initially denoised sub-band signal and the standard waveform is normalized so that the maximum absolute amplitude of both signals is in units of 1. A fixed-length sliding window is set to cover twice the width of the main lobe of the standard waveform. The window slides along the time axis, and the cross-correlation coefficient between the initially denoised sub-band signal segment and the standard waveform segment within the window is calculated. The formula for calculating the cross-correlation coefficient is:

[0036] in: Represents relative time shift Cross-correlation coefficients at time This indicates the first sub-band signal in the window after initial denoising. Each sample value, This represents the mean value of the sub-band signal samples after initial denoising within this window. This indicates the relative time shift of the standard waveform within the window. The next Each sample value, This represents the mean of the standard waveform samples within the window. This represents the number of sampling points within the window. Iterate through all possible integer time shifts. , find Time shift to obtain the maximum value Based on this, time-shift alignment operations are performed on the standard waveform.

[0037] It is understandable that, in the aligned signal, the waveform distortion coefficient reflects the degree of deviation in shape between the initially denoised sub-band signal and the standard waveform. The square of the amplitude difference between corresponding sampling points of the two signals is calculated, the average of these squares is taken, and then the square root is used to obtain the root mean square (RMS) value of the amplitude difference. The RMS value of the amplitude of the standard waveform over its entire length is calculated separately, and the RMS value of the amplitude difference is divided by the RMS value of the standard waveform's amplitude; the resulting dimensionless ratio is the waveform distortion coefficient. The morphology compensation module performs weighted correction on the phase-corrected signal based on the magnitude of the waveform distortion coefficient. The larger the waveform distortion coefficient, the higher the correction weight, thereby reducing waveform morphology distortion caused by differences in propagation paths. In some embodiments, the cross-correlation coefficient calculation within the sliding window uses a fast convolution algorithm to improve computational efficiency. The time-shift alignment operation includes not only integer multiples of the sampling interval for translation but also fine adjustments based on fractional multiples of the sampling interval using polynomial interpolation to further improve alignment accuracy. The calculation results of the phase difference and waveform distortion coefficient are stored in a feature register for subsequent signal reconstruction and quality assessment modules to use.

[0038] In one embodiment of the present invention, a standard ultrasonic response template matching the material under test is retrieved from a structural feature library. This template covers the ideal reflection waveform characteristics of each material interface from the surface to the deep layers. The ultrasonic reconstructed signal is subjected to point-by-point correlation analysis with the standard template in the time domain to identify the reflection peak positions corresponding to each material interface in the signal. Based on the identified peak positions, the signal is divided into multiple signal segments in the time domain, each segment corresponding to a structural layer. The envelope, center frequency offset, and signal attenuation rate of each signal segment are extracted and combined to form a primary feature vector of the structural layer. This vector is input into a pre-trained structural state classification model, and the model outputs the fiber orientation distribution state score, interface adhesion state score, and internal defect existence probability of the layer. All detection points and the above scores and probabilities of all structural layers are interpolated and rendered in a three-dimensional spatial grid to generate independent fiber orientation distribution maps, interface adhesion state maps, and internal defect masks, which together constitute a multi-layer structural feature mapping map. The specific operation of segmenting the signal based on the position of the reflected wave peak is as follows: the midpoint between two adjacent wave peaks is taken as the segmentation point; the signal portion before the first wave peak is classified as the signal segment of the material surface layer; between two adjacent wave peaks, the middle signal is extracted as the signal segment of the internal structural layer, with the segmentation point of the preceding wave peak as the starting point and the segmentation point of the following wave peak as the ending point; the signal portion after the last wave peak is classified as the signal segment of the deep or bottom layer of the material; and each segmented signal segment is labeled with the corresponding structural layer number and the start and end positions of the time domain.

[0039] In practice, the process of performing hierarchical analysis on the enhanced ultrasonic reconstructed signal using a polyester fiber material structural feature library is executed by a hierarchical analysis server. The server's memory is loaded with a standardized database adapted to different types of polyester fiber materials. A standard ultrasonic response template that perfectly matches the batch number of the sample under test is retrieved from the polyester fiber material structural feature library. The standard ultrasonic response template is a multi-channel time-domain waveform matrix, with each row corresponding to the ideal interface reflection waveforms at different depths from the surface coating, fiber layer to the substrate. A point-by-point sliding correlation operation is performed between the enhanced ultrasonic reconstructed signal and the waveforms of each channel of the standard ultrasonic response template. Peak positions with a correlation coefficient exceeding 0.85 are identified as valid material interface reflection peaks, and the absolute time coordinates and interface types of these peaks are recorded.

[0040] In some embodiments, the temporal segmentation of the ultrasonic reconstructed signal based on the identified reflection peak positions follows the principle of interface reflection symmetry. The midpoint between two adjacent reflection peak positions is defined as the theoretical segmentation point of the interface between the two material layers. All signal points before the first reflection peak position belong to the material surface protective layer, and all signal points after the last reflection peak position belong to the material backing layer or matrix layer. For the signal segment between two adjacent reflection peaks, the segmentation point corresponding to the previous peak is used as the starting index, and the segmentation point corresponding to the next peak is used as the ending index. The intermediate continuous sampling points are extracted to form an independent signal segment of the internal fiber structure layer. Each segmented signal segment is assigned a globally unique hierarchical identifier, and its starting sampling point number and ending sampling point number in the original ultrasonic reconstructed signal are recorded.

[0041] In practical implementation, the primary feature vector extraction operation for a single structural layer signal segment is performed in the feature extraction engine. The envelope is obtained by calculating the modulus of the analytic signal through Hilbert transform. The center frequency offset is determined by calculating the difference between the centroid of the power spectrum of the signal segment and the reference frequency of the corresponding layer of the standard template using short-time Fourier transform. The signal attenuation rate is calculated by linearly fitting the slope of the logarithmic magnitude. The structural state classification model adopts a deep convolutional neural network architecture. The number of nodes in the network input layer is consistent with the dimension of the primary feature vector. The hidden layer contains three fully connected layers and a ReLU activation function. The output layer contains three independent neurons, corresponding to the fiber orientation distribution state score, the interface adhesion state score, and the probability of internal defect existence, respectively.

[0042] The fiber orientation distribution score quantifies the degree of orderliness of the fiber arrangement, with a score range of 0 to 1. A higher value indicates that the orientation is closer to the ideal process standard. The interface adhesion score assesses the tightness of the bond between the fiber and the matrix, with a score range of 0 to 1. A higher value indicates better adhesion quality. The probability of internal defects characterizes the probability of the presence of pores, cracks, or inclusions in the structural layer, with a value range of 0 to 1. A value exceeding 0.5 is considered to indicate a risk of defects. All detection points are located in a three-dimensional spatial grid based on their physical coordinates. The grid node spacing is set to half the diameter of the ultrasonic probe, and the eigenvalues ​​of missing points are filled using radial basis function interpolation.

[0043] In some embodiments, the interpolation rendering process is executed in parallel on the graphics processor. The interpolation result of the fiber orientation distribution status score generates a grayscale image, with the grayscale value proportional to the score; the interpolation result of the interface adhesion status score generates a pseudo-color heatmap, using a Jet color chart; the interpolation result of the probability of internal defects generates a binary mask image, with pixels having a probability greater than or equal to 0.5 assigned a value of 1 and pixels having a probability less than 0.5 assigned a value of 0. These three images are superimposed on the same coordinate system to form a multi-layer structural feature map containing spatial location, structural attributes, and defect markers. Referring to Table 1, the primary feature vectors and model output results of three different structural layers in a typical polyester fiber composite material sample are shown: Table 1: Primary Feature Vectors of Structural Layers and Output of Classification Model

[0044] In practice, the training data for the structural state classification model comes from a large dataset of ultrasonic testing of polyester fiber material samples with known structural states. The loss function uses a combination of cross-entropy and mean squared error, and the optimization objective is to minimize the difference between the predicted score and the true label. After the model is trained, the parameters are fixed and deployed to the online detection system to achieve real-time hierarchical analysis of the ultrasonic reconstructed signal and generation of feature maps.

[0045] In one embodiment of the present invention, all independent defect regions are identified from the internal defect mask, and the equivalent diameter, area, and thickness projection length of each region are calculated; the average fiber orientation angle and distribution dispersion at each defect region are read from the fiber orientation distribution map; the adhesion strength score of the interface adjacent to each defect region is obtained from the interface adhesion state map; the equivalent diameter, area, projection length, average fiber orientation angle, distribution dispersion, and adhesion strength score are used as model input features; the model calculates the local stress concentration coefficient and strength reduction coefficient for each defect region based on the pre-learned mapping relationship; the local stress concentration coefficients of all defect regions are summarized, and the overall maximum and average stress concentration coefficients are calculated; the strength reduction coefficients of all defect regions are summarized, and the overall average and the strength reduction coefficient of the weakest region are calculated; the overall maximum stress concentration coefficient, average stress concentration coefficient, overall average strength reduction coefficient, and the strength reduction coefficient of the weakest region are used together as defect-related performance degradation parameters. The steps for calculating the geometric features of each independent defect region are as follows: perform connected component analysis on the internal defect mask to mark each independent region; calculate the total pixel area of ​​each region and convert it into the actual physical area according to the preset spatial resolution; take the diameter of a circle with the same area as the equivalent diameter of the region; project each independent region along the direction perpendicular to the material surface and calculate the maximum length of the projected area in the thickness direction as the projection length.

[0046] In practice, the process of calculating defect-related performance degradation parameters based on a multi-layer structural feature map and a material performance degradation model is performed on a defect assessment workstation. The workstation is loaded with a fully trained material performance degradation model and 3D spatial mapping data. From the internal defect mask layer of the multi-layer structural feature map, all independent defect regions are identified using an eight-neighbor connected component labeling algorithm. Each independent defect region consists of one or more connected pixels and is assigned a unique identifier. Based on preset spatial resolution parameters, the number of pixels in the independent defect regions is converted into actual physical areas. Assuming a spatial resolution of 0.1 mm per pixel, a defect region with an area of ​​50 pixels corresponds to an actual physical area of ​​0.5 square millimeters. The diameter of a circle with the same actual physical area is used as the equivalent diameter of the defect region; a defect region with an area of ​​0.5 square millimeters has an equivalent diameter of approximately 0.798 mm. The pixel distribution of each independent defect region is projected along the thickness direction perpendicular to the material surface. The maximum pixel span occupied by the projected region in the thickness direction is calculated, and then multiplied by the spatial resolution to obtain the projection length.

[0047] In some embodiments, the fiber orientation distribution map and the interface adhesion state map are raster data of the same scale registered with the internal defect mask. For each independent defect region, the fiber orientation angle data of all pixels within the region are extracted, and the arithmetic mean of these angles is calculated as the average fiber orientation angle of the defect region. The standard deviation of these angles is calculated as the distribution dispersion. Simultaneously, all interface adhesion state score pixels within a range extending outward from the edge of the independent defect region are extracted, and the average of these scores is calculated as the adhesion strength score of the adjacent interface. These geometric parameters and structural parameters are combined into an input feature vector for each independent defect region. The input feature vector contains six dimensions: equivalent diameter, area, projected length, average fiber orientation angle, distribution dispersion, and adhesion strength score.

[0048] In practice, the material property degradation model is constructed using a gradient boosting tree regression algorithm. The model input is a six-dimensional feature vector, and the output consists of two target parameters: the local stress concentration factor and the strength reduction factor. The local stress concentration factor represents the factor by which the local stress is amplified by the defect, while the strength reduction factor represents the proportion by which the material strength decreases due to the defect. The model training data comes from a combination of finite element simulation and destructive mechanics experiments, covering various defect sizes, distributions, and material conditions. For each independent defect region, the model calculates the corresponding local stress concentration factor and strength reduction factor based on its feature vector, and the calculation results are stored in the defect feature list.

[0049] It is understandable that the defect-related performance degradation parameter aims to reflect the overall structural performance degradation. It summarizes the local stress concentration factors of all independent defect regions, selects the maximum value as the overall maximum stress concentration factor, and calculates the arithmetic mean of all local stress concentration factors as the average stress concentration factor. It also summarizes the strength reduction factors of all independent defect regions, calculates the arithmetic mean of all strength reduction factors as the overall average strength reduction factor, and selects the minimum value as the strength reduction factor for the weakest region. These four statistics together constitute the defect-related performance degradation parameter, used for subsequent structural integrity grading evaluation. Refer to Table 2, which shows the input features and model output calculations for a sample containing three independent defect regions: Table 2: Input of Independent Defect Region Features and Calculation of Performance Degradation Parameters

[0050] In the example above, the overall maximum stress concentration factor is 2.41, the average stress concentration factor is 1.94, the overall average strength reduction factor is 0.72, and the strength reduction factor of the weakest region is 0.63. The calculation process of the material property degradation model can be summarized mathematically, and the local stress concentration factor is generated by the characteristic mapping function:

[0051] in: Indicates the local stress concentration factor. This represents the nonlinear mapping relationship of the material property degradation model. Indicates the equivalent diameter. Indicates area, Indicates the projection length. Indicates the average orientation angle of the fibers. Indicates the dispersion of the distribution. This represents the bond strength score. The strength reduction factor is calculated following a similar mapping logic, and the final output set of defect-related performance degradation parameters is used to quantitatively assess the safety margin of the structure.

[0052] In one embodiment of the present invention, the allowable stress concentration factor threshold and allowable strength reduction factor threshold, which match the current material grade, heat treatment state, and usage environment, are read from a predefined structural safety threshold table; the overall maximum stress concentration factor in the defect-related performance degradation parameters is compared with the allowable stress concentration factor threshold; the strength reduction factor of the weakest region is compared with the allowable strength reduction factor threshold; if the overall maximum stress concentration factor does not exceed the threshold and the strength reduction factor of the weakest region is not lower than the threshold, the structural integrity is assessed as safe; if the overall maximum stress concentration factor exceeds the standard or the strength reduction factor of the weakest region is lower than the threshold, but the average stress concentration factor and the overall average strength reduction factor are still within an acceptable range, it is assessed as a caution level; if the overall maximum stress concentration factor exceeds the standard or the strength reduction factor of the weakest region is lower than the threshold, it is assessed as a risk level; based on the assessment results, combined with the defect location coordinates in the internal defect mask, a structural integrity evaluation report containing the defect level, location coordinates, and risk level is generated.

[0053] In practice, the process of grading and evaluating the structural integrity of polyester fiber materials by combining defect-related performance degradation parameters with predefined structural safety thresholds is completed on a structural evaluation terminal. This terminal has a built-in structural safety threshold database and an automatic rating engine. The predefined structural safety threshold table uses material grade, heat treatment status code, and usage environment code as a joint primary key. Records that perfectly match the properties of the current test sample are retrieved, and the allowable stress concentration factor threshold and allowable strength reduction factor threshold are read from these records. The allowable stress concentration factor threshold represents the upper limit of the maximum allowable stress amplification factor for the material under safe service conditions, and the allowable strength reduction factor threshold represents the lower limit of the minimum allowable strength retention ratio for the material under safe service conditions.

[0054] In some embodiments, the defect-related performance degradation parameters include four indicators: the overall maximum stress concentration factor, the average stress concentration factor, the overall average strength reduction factor, and the strength reduction factor of the weakest region. The rating engine compares the overall maximum stress concentration factor with the allowable stress concentration factor threshold, and compares the strength reduction factor of the weakest region with the allowable strength reduction factor threshold. If the overall maximum stress concentration factor is less than or equal to the allowable stress concentration factor threshold, and the strength reduction factor of the weakest region is greater than or equal to the allowable strength reduction factor threshold, the evaluation engine marks the structural integrity status as a safe level.

[0055] In practice, if the overall maximum stress concentration factor exceeds the allowable stress concentration factor threshold, or the strength reduction factor of the weakest area is less than the allowable strength reduction factor threshold, the evaluation engine will further check whether the average stress concentration factor and the overall average strength reduction factor are within acceptable ranges. The acceptable range is determined based on the secondary threshold field in the predefined structural safety threshold table. When the average stress concentration factor does not exceed 1.1 times the allowable stress concentration factor threshold, and the overall average strength reduction factor is not less than 0.9 times the allowable strength reduction factor threshold, the structural integrity is classified as a concern level. If the above conditions are not met—that is, the overall maximum stress concentration factor exceeds the limit or the strength reduction factor of the weakest area is insufficient, and the average level also deviates from the safe range—then the structural integrity is directly classified as a risk level.

[0056] Understandably, the structural integrity assessment report is output in a structured data format. The report header includes the sample number, inspection time, and assessment conclusion level. The main body of the report integrates the defect location coordinates from the internal defect mask, marking the X, Y, and Z coordinates of each defect area in three-dimensional space, and classifying the defect level based on the magnitude of the local stress concentration factor and strength reduction factor. The defect levels are divided into three categories: minor, moderate, and severe. The risk level is consistent with the structural integrity level, divided into three categories: safe, cautious, and risky. The final report document includes a complete defect list, location map, and risk description to guide material maintenance or replacement decisions.

[0057] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A non-destructive testing method based on polyester fiber materials, characterized in that, The method includes: Acquire an initial set of ultrasonic echo signals of polyester fiber material, wherein the initial set of ultrasonic echo signals contains original time-domain waveform data at different detection points; An improved ultrasonic signal reconstruction algorithm is applied to the initial ultrasonic echo signal set to generate a feature-enhanced ultrasonic reconstructed signal. The improved ultrasonic signal reconstruction algorithm processes the original time-domain waveform data based on an adaptive noise suppression model and a waveform matching criterion. The ultrasonic reconstruction signal enhanced by the feature enhancement is subjected to hierarchical analysis using a polyester fiber material structural feature library to generate a multi-layer structural feature map, which includes fiber orientation distribution, interface adhesion state and internal defect mask. Based on the multi-layer structure feature mapping map, defect-related performance degradation parameters are calculated through a material performance degradation model, which integrates the mapping relationship between defect size, defect density and stress concentration factor. By combining the defect-related performance degradation parameters with the predefined structural safety threshold, the structural integrity of polyester fiber materials is graded and evaluated, generating a structural integrity evaluation report that includes defect level, location coordinates, and risk level.

2. The non-destructive testing method based on polyester fiber material according to claim 1, characterized in that, An improved ultrasound signal reconstruction algorithm is applied to the initial set of ultrasound echo signals to generate a feature-enhanced ultrasound reconstructed signal, including: The original time-domain waveform data is decomposed into multiple sub-frequency band signal components. For each sub-band signal component, the noise energy distribution of the sub-band in the time domain is calculated according to the adaptive noise suppression model, and the standard waveform with the highest similarity to the sub-band signal component is selected from the preset reference waveform library according to the waveform matching criterion. Based on the calculated noise energy distribution, adaptive threshold noise reduction processing is performed on the sub-frequency band signal components to obtain a preliminarily denoised sub-frequency band signal. The preliminarily denoised sub-band signal is time-domain aligned and amplitude-normalized with the selected standard waveform; Based on the similarity metric defined by the waveform matching criterion, calculate the phase difference and waveform distortion coefficient between the initially denoised sub-band signal and the standard waveform; The phase difference is used to perform phase correction on the sub-band signal of the initial denoising, and the waveform distortion coefficient is used to perform morphological compensation on the waveform of the corrected signal. The signals from all sub-bands, after phase correction and morphological compensation, are combined to generate the enhanced ultrasonic reconstructed signal.

3. The non-destructive testing method based on polyester fiber materials according to claim 2, characterized in that, The noise energy distribution of the sub-band in the time domain is calculated based on the adaptive noise suppression model, including: In the time-domain sequence of the sub-frequency band signal components, a sliding time window is used to extract multiple local signal segments; Power spectral density estimation is performed on each local signal segment. Frequency components in the power spectrum below a preset energy threshold are identified as noise components, and frequency components above the preset energy threshold are identified as signal components. The energy values ​​of all frequency components identified as noise components in each local signal segment are accumulated and calculated as the local noise energy of the local signal segment. Using the local noise energy as the vertical axis and the center time of the sliding time window as the horizontal axis, a noise energy distribution curve of the sub-frequency band in the time domain is constructed. The noise energy distribution curve is smoothed, and the smoothed curve is used as the output of the adaptive noise suppression model to guide the dynamic setting of the threshold in the adaptive threshold noise reduction process.

4. The non-destructive testing method based on polyester fiber material according to claim 2, characterized in that, The step of calculating the phase difference and waveform distortion coefficient between the preliminarily denoised sub-band signal and the standard waveform using a similarity metric defined by the waveform matching criterion is achieved through the following steps: The Hilbert transform is performed on the preliminarily denoised sub-band signal and the selected standard waveform respectively to obtain the analytical signal form of the two signals; The instantaneous phase sequences of the two signals are extracted from the analytical signals of the sub-band signals after preliminary denoising and the analytical signals of the standard waveform, respectively. Calculate the difference between the instantaneous phase sequence of the preliminarily denoised sub-band signal and the instantaneous phase sequence of the standard waveform at each sampling time point; The arithmetic mean of the phase difference values ​​at all sampling time points is calculated, and the arithmetic mean is used as the phase difference between the sub-frequency band signal of the initial denoising and the standard waveform; The sub-band signal and the standard waveform after preliminary denoising are respectively normalized for amplitude processing to ensure that the maximum amplitude of the two signals is consistent. In the time domain, using a sliding window of fixed length, the cross-correlation coefficient between the preliminarily denoised sub-band signal and the standard waveform within the corresponding window signal segment is calculated. Find the relative time shift between windows that makes the cross-correlation coefficient reach its maximum value, and perform time shift alignment operation on the standard waveform based on this relative time shift; On the aligned signal, the root mean square value of the amplitude difference between the corresponding sampling points of the preliminarily denoised sub-band signal and the standard waveform is calculated, and the root mean square value is divided by the root mean square value of the amplitude of the standard waveform. The resulting ratio is used as the waveform distortion coefficient.

5. The non-destructive testing method based on polyester fiber material according to claim 1, characterized in that, The step of using a polyester fiber material structural feature library to perform layered analytical processing on the enhanced ultrasonic reconstructed signal to generate a multi-layer structural feature map includes: From the polyester fiber material structural feature library, a standard ultrasonic response template corresponding to the model of the polyester fiber material to be tested is retrieved. The standard ultrasonic response template contains ideal reflection waveform features of different material interfaces from the surface layer to the deep layer. The feature-enhanced ultrasonic reconstructed signal is subjected to point-by-point correlation analysis with the standard ultrasonic response template in the time domain to identify the position of the reflection peak in the feature-enhanced ultrasonic reconstructed signal corresponding to each material interface. Based on the identified positions of each reflected wave peak, the enhanced ultrasonic reconstructed signal is divided into multiple signal segments in the time domain, with each signal segment corresponding to a structural layer of the polyester fiber material. For each signal segment, its envelope, center frequency offset, and signal attenuation rate are extracted to form the primary feature vector of the structural layer. The primary feature vector of each structural layer is input into a pre-trained structural state classification model, which outputs the fiber orientation distribution state score, interface adhesion state score, and probability of internal defect existence of the structural layer. All detection points, fiber orientation distribution status scores, interface adhesion status scores, and internal defect existence probabilities of all structural layers are interpolated and rendered in a three-dimensional spatial grid to generate independent fiber orientation distribution maps, interface adhesion status maps, and internal defect masks, which together constitute the feature mapping map of the multi-layer structure.

6. The non-destructive testing method based on polyester fiber material according to claim 5, characterized in that, The process involves dividing the enhanced ultrasonic reconstructed signal into multiple signal segments in the time domain based on the identified positions of each reflected wave peak, including: The midpoint between two adjacent reflected wave peak positions is used as the dividing point; The portion of the signal preceding the first reflected wave peak is taken as the signal segment of the corresponding material surface layer; Between two adjacent reflection peaks, the signal segment in the middle is taken as the signal segment of the corresponding internal structural layer, with the dividing point closer to the previous reflection peak as the starting point and the dividing point closer to the next reflection peak as the ending point. The signal portion following the last reflected wave peak is taken as the signal segment of the corresponding deep or bottom layer of the material; Each segmented signal is labeled with its corresponding structural layer number and time domain start and end positions.

7. The non-destructive testing method based on polyester fiber material according to claim 5, characterized in that, Based on the aforementioned multilayer structure feature map, defect-related performance degradation parameters are calculated using a material performance degradation model, including: From the internal defect mask, all independent defect regions are identified, and the equivalent diameter, area, and projected length in the thickness direction of each independent defect region are calculated. From the fiber orientation distribution map, read the average fiber orientation angle and its distribution dispersion at the location of each independent defect region; From the interface adhesion state diagram, read the adhesion strength score of the material interface adjacent to each individual defect area; The equivalent diameter, area, projected length, average fiber orientation angle, distribution dispersion, and bond strength score are used as input features for the material performance degradation model. The material property degradation model calculates a local stress concentration factor and a strength reduction factor for each independent defect region based on the pre-learned mapping relationship. Summarize the local stress concentration factors of all independent defect areas, and calculate the overall maximum stress concentration factor and average stress concentration factor; Summarize the strength reduction factors of all independent defect areas, and calculate the overall average strength reduction factor and the strength reduction factor of the weakest area; The overall maximum stress concentration factor, average stress concentration factor, overall average strength reduction factor, and strength reduction factor of the weakest region are collectively used as the defect-related performance degradation parameter.

8. The non-destructive testing method based on polyester fiber material according to claim 7, characterized in that, The calculation of the equivalent diameter, area, and projected length in the thickness direction of each independent defect region includes: Perform connected component analysis on the internal defect mask and mark each independent connected region as an independent defect region; Calculate the total pixel area of ​​each independent defect region, and convert the pixel area into the actual physical area according to the preset spatial resolution; The diameter of a circle with the same actual physical area is taken as the equivalent diameter of the independent defect region. Project each individual defect region along a direction perpendicular to the material surface, i.e., the thickness direction, and calculate the maximum length of the projected region in the thickness direction as the projection length.

9. The non-destructive testing method based on polyester fiber material according to claim 1, characterized in that, By combining the aforementioned defect-related performance degradation parameters with predefined structural safety thresholds, the structural integrity of polyester fiber materials is graded and evaluated, including: From the predefined structural safety threshold table, read the allowable stress concentration factor threshold and allowable strength reduction factor threshold that match the current polyester fiber material grade, heat treatment state and usage environment; The overall maximum stress concentration factor in the defect-related performance degradation parameter is compared with the allowable stress concentration factor threshold. Compare the strength reduction factor of the weakest region in the defect-related performance degradation parameter with the allowable strength reduction factor threshold. If the overall maximum stress concentration factor does not exceed the allowable stress concentration factor threshold, and the strength reduction factor of the weakest region is not lower than the allowable strength reduction factor threshold, then the structural integrity is determined to be at a safe level. If the overall maximum stress concentration factor exceeds the allowable stress concentration factor threshold, or the strength reduction factor of the weakest region is lower than the allowable strength reduction factor threshold, but the average stress concentration factor and the overall average strength reduction factor in the defect-related performance degradation parameters are still within acceptable ranges, then the structural integrity is determined to be at the level of concern. If the overall maximum stress concentration factor exceeds the allowable stress concentration factor threshold, or the strength reduction factor of the weakest region is lower than the allowable strength reduction factor threshold, then the structural integrity is determined to be at risk level. Based on the determined safety level, attention level, or risk level, and combined with the location coordinates of the defects in the internal defect mask, a structural integrity evaluation report containing the defect level, location coordinates, and risk level is generated.

10. A non-destructive testing system based on polyester fiber materials, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the non-destructive testing method based on polyester fiber material as described in any one of claims 1 to 9.

Images