Method and system for detecting internal defects in a composite material part

Abstract

The present application relates to the technical field of composite material detection, and specifically discloses a kind of composite material workpiece internal defect detection method and system, the method comprises: by obtaining defect detection feature and calculating feature overlap, judging modal feature degradation and constructing hierarchical decomposition model to separate thermal stress coupling signal, difference analysis area is required to be applied gradient humid heat load to generate dynamic response matrix, extract mutation direction angle and filter anisotropic interference by time-frequency coupling analysis, determine layered hidden boundary and boundary hidden area based on mutation direction angle, then image extraction geometric parameter is optimized by error correction matrix after model;System includes: overlap extraction module, confusion analysis module, difference analysis module, mutation identification module and feedback correction module, the present application is favorable to reduce the boundary fuzzy and small defect missed detection problem in traditional detection, realize the positioning and closed loop optimization of defect under multi-field coupling, improve composite material quality control efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of composite material detection, and in particular to a composite part internal defect detection method and system. BACKGROUND

[0002] Carbon fiber-epoxy laminated plate composite materials are widely used in high-end fields such as aerospace and rail transportation due to their high specific strength, fatigue resistance and other excellent properties. However, during the preparation and service of the composite material, the existence of internal defects will significantly reduce the mechanical properties of the component. Therefore, detecting internal defects is the key to ensuring the reliability of the composite component.

[0003] The prior art is prone to misjudgment of defects in the feature overlap area when capturing features of different types of defects in a single mode. The anisotropic intrinsic properties of the composite material interfere with the detection signal, causing the defect boundary to be fuzzy. Traditional static detection cannot simulate the influence of the coupling load of heat and stress in the service environment on the evolution of the defect, and it is difficult to identify potential defect expansion under dynamic load.

[0004] Therefore, the present application provides a composite part internal defect detection method and system. SUMMARY

[0005] The present application aims to provide a composite part internal defect detection method and system to solve the above background problems.

[0006] The object of the present application can be achieved by the following technical solutions:

[0007] A composite part internal defect detection method, comprising the following steps:

[0008] In the historical level micro-defect inspection of carbon fiber and epoxy resin layer composite materials, defect detection features are obtained and overlap detection is performed to obtain the feature overlap of the historical level micro-defects;

[0009] Based on the feature overlap, it is determined whether different defects have modal feature degradation, and if so, a hierarchical decomposition model is constructed to separate the signal confusion caused by thermal stress coupling, the defect detection features are evaluated for confusion, the confusion improvement rate is obtained, and it is determined whether difference analysis is needed;

[0010] If difference analysis is needed, a diagnostic detection interface is divided and a gradient humidity load is applied, and it is analyzed whether the defect detection feature difference of the diagnostic detection interface is significant, and if so, a dynamic response matrix under dynamic load is generated;

[0011] Based on the dynamic response matrix, time-frequency coupling analysis is performed to identify the strain gradient of the layered fiber direction, the mutation direction angle is extracted, and the anisotropy of the composite material is identified to interfere with the mutation direction angle, and if there is interference, the defect area is filtered.

[0012] As a further technical solution of the present application: the way of carrying out the overlap detection is:

[0013] Obtaining an overlap detection vector, mapping the overlap detection vector in a three-dimensional space to form a feature distribution area of each of the three types of defect detection features;

[0014] Calculate the intersection and union of the feature distribution areas of different types of defect detection features, and calculate the overlap degree to obtain the feature overlap degree.

[0015] As a further technical solution of the present application: the way of obtaining the overlap detection vector is:

[0016] Obtaining the hierarchical defect detection feature, the porosity defect detection feature, and the debonding defect detection feature of the defect area, and taking the three defect detection features of the defect area as the defect detection features;

[0017] Simplify the defect area into sampling points, and construct an overlap detection vector containing the defect detection features of each sampling point.

[0018] As a further technical solution of the present application: the way of carrying out confusion evaluation on the defect detection features is:

[0019] If the modal feature degradation occurs, obtain the three types of defect detection features of the overlap detection vector of all historical sampling points, and perform relevant coupling analysis on the historical P-level thermal stress load to determine whether the three types of defect detection features and the P-level thermal stress load exist relevant coupling;

[0020] If there is relevant coupling, construct a three-order tensor and perform low-rank constraint decomposition on the three-order tensor, and construct a defect classification confusion matrix before and after the low-rank constraint decomposition;

[0021] Calculate the confusion improvement rate based on the defect classification confusion matrix, and if the confusion improvement rate is lower than a preset confusion improvement threshold, difference analysis is needed.

[0022] As a further technical solution of the present application: the judgment way of the modal feature degradation is:

[0023] Constructing a defect pair of any two groups of defect detection features of the hierarchical defect detection feature, the porosity defect detection feature, and the debonding defect detection feature;

[0024] Obtaining the coordinates of the corresponding sampling points of the two groups of defects, calculating the Euclidean distance of the coordinates of the two groups of defect sampling points, and distinguishing the defect pairs into spatial adjacent pairs and spatial irrelevant pairs;

[0025] The probability of appearing in the defect detection area is calculated, the spatial proximity entropy of the spatial proximity pair is calculated by an information entropy algorithm, and the feature overlap degree is divided into fixed overlap degree and random overlap degree;

[0026] The feature overlap degree of the current hierarchical micro-defect inspection is obtained, and a state feature degradation equation is established based on the feature overlap degree of the current carbon fiber and epoxy resin layer composite material, the average fixed overlap degree of all hierarchical micro-defects, and the average random overlap degree of all hierarchical micro-defects.

[0027] If the modal feature degradation equation is satisfied, i.e., modal feature degradation occurs.

[0028] As a further technical solution of the application: the way to analyze whether the defect detection feature difference of the diagnostic detection interface is significant is:

[0029] The defect detection area where the modal feature degradation occurs is divided into a diagnostic detection interface.

[0030] Gradient humidity and heat load is applied to the defect detection area, the feature sensitivity of different types of defect detection features in the defect detection area is obtained, and if the defect detection feature difference of the diagnostic detection interface is significant, the feature sensitivity of different types of defects is sorted and processed.

[0031] If the feature sensitivity of the layered defect is the highest, the diagnostic detection interface is re-divided according to the feature sensitivity, the number of divided interface partitions is obtained, and a dynamic response matrix is constructed.

[0032] As a further technical solution of the application: the way to filter the defect area is:

[0033] Obtain the strain gradient and the feature frequency band, map the strain gradient value of the feature frequency band to the spatial position, and extract the mutation point of the strain gradient;

[0034] Obtain the angle between the principal direction of the strain gradient point and the fiber direction to obtain the mutation direction angle;

[0035] Decompose the strain gradient into components in different directions, calculate the frequency domain energy proportion of each direction component of the strain gradient at the feature frequency, and calculate the anisotropic correlation factor;

[0036] Filter out the mutation direction angle with interference and the layered defect area to which the layered defect with interference belongs based on the anisotropic correlation factor.

[0037] As a further technical solution of the application: it further includes the following steps: the way to obtain the strain gradient is:

[0038] Based on the dynamic response matrix, time-frequency domain conversion is carried out, a time-frequency matrix is constructed, a spatial gradient of the time-frequency matrix is calculated, and the spatial derivative in the fiber direction is obtained as the strain gradient.

[0039] As a further technical solution of the application, the method further comprises the following steps:

[0040] Based on the mutation direction angle, the hidden boundary of the layered area is determined, the boundary hidden area is determined, the boundary hidden area is re-imaged, the geometric parameters of the hidden detection features are extracted, the error correction matrix is generated after comparison with the detection model output, the error correction matrix is fed back to the hierarchical decomposition model, and the corrected defect recognition result is output and embedded into the composite material quality control system.

[0041] A composite material part internal defect detection system comprises the following modules:

[0042] The overlap extraction module is used for obtaining defect detection features and performing overlap detection in the historical hierarchical micro-defect inspection of carbon fiber and epoxy resin layer composite materials, and obtaining the feature overlap degree of the historical hierarchical micro-defects.

[0043] The confusion analysis module is used for judging whether different defects appear modal feature degradation based on the feature overlap degree, and if the modal feature degradation appears, a hierarchical decomposition model is constructed to separate the signal confusion caused by thermal stress coupling, the confusion of the defect detection features is evaluated, the confusion improvement rate is obtained, and it is judged whether difference analysis is needed.

[0044] The difference analysis module is used for dividing a diagnostic detection interface and applying a gradient hygrothermal load if difference analysis is needed, and judging whether the difference of the defect detection features of the diagnostic detection interface is significant, and if the difference is significant, a dynamic response matrix under dynamic load is generated.

[0045] The mutation recognition module is used for performing time-frequency coupling analysis to recognize the strain gradient of the layered fiber direction based on the dynamic response matrix, extracting the mutation direction angle, and recognizing the interference of the anisotropy of the composite material on the mutation direction angle, and if the interference exists, the defect area is filtered.

[0046] The application has the following beneficial effects:

[0047] By fusing the ultrasonic echo amplitude decay rate, thermal diffusion time constant and gray threshold porosity three types of defect detection features, the standardization is carried out, the overlapping detection vector is constructed and three-dimensional space mapping is carried out, which is beneficial to realize the cross verification of multi-modal features of delamination, debonding and porosity defects, reduce the one-sidedness of single detection means, improve the quantitative fusion accuracy of complex defect detection features, and provide visual spatial distribution basis for subsequent modal degradation judgment; based on the spatial proximity entropy, the fixed overlap degree and the random overlap degree are distinguished, the modal degradation is judged through the threshold value, and the thermal stress coupling interference is separated by using the third-order tensor low-rank constraint decomposition model, which is beneficial to diagnose the feature confusion phenomenon of different defects, reduce the misjudgment caused by signal aliasing, and effectively alleviate the interference of thermal stress on defect detection features, and improve the classification recognition reliability of multiple defect types.

[0048] By calculating and analyzing the dynamic response difference of defect detection features through feature sensitivity, the evolution characteristics of defects under the action of multi-field coupling are captured, the problem that the traditional static detection cannot reflect the dynamic expansion of defects is solved, and data support is provided for defect evolution path analysis and multi-field response matrix construction; through wavelet transform of the dynamic response matrix by time-frequency coupling analysis, the strain gradient spatial derivative is calculated along the fiber direction, and the anisotropic correlation factor is used to filter the intrinsic property interference of composite materials, which is beneficial to locate the strain gradient mutation point of the edge of delamination defects, suppress the pseudo-signal interference caused by fiber lay-up, and improve the authenticity and accuracy of defect hidden boundary positioning.

[0049] The ultrasonic phased array and infrared thermal imaging high-resolution reimaging are carried out on the boundary hidden area, the defect geometric parameters are extracted and compared with the model predicted value to generate an error correction matrix, which is beneficial to capture the micro-defects and boundary fuzzy area missed by traditional detection, realize the iterative optimization of hierarchical decomposition model through error feedback, and improve the adaptability and recognition reliability of the detection system to different materials and defect types. BRIEF DESCRIPTION OF DRAWINGS

[0050] The application will be further described below with reference to the drawings.

[0051] Figure 1 is a flowchart of a composite material part internal defect detection method of the application;

[0052] Figure 2 is a flowchart of whether difference analysis is needed according to the application;

[0053] Figure 3 is a module diagram of a composite material part internal defect detection system according to the application. DETAILED DESCRIPTION

[0054] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of the present application.

[0055] Embodiment 1:

[0056] Please refer to Figure 1 The present application is a kind of composite material part internal defect detection method, comprising the following steps:

[0057] S1, in the historical level micro defect inspection of carbon fiber and epoxy resin layer composite material, the defect detection characteristics are obtained and the degree of overlap is detected, and the characteristic overlap of the historical level micro defect is obtained.

[0058] The way of obtaining the defect detection characteristics of delamination, debonding and porosity is:

[0059] In the historical level micro defect inspection of carbon fiber and epoxy resin layer composite material, the micro defect inspection is divided into different defect detection areas, and the defect area and the defect-free area of the carbon fiber and the epoxy resin layer composite material are obtained.

[0060] It can be understood that, in the historical level micro defect inspection, the ultrasonic echo amplitude of the defect area is obtained by ultrasonic detection, and compared with the amplitude of the defect-free area, the difference percentage of the baseline value is calculated, and the ultrasonic echo amplitude attenuation rate is obtained as the delamination defect detection characteristic.

[0061] Using infrared thermal imaging technology, the time required for the surface temperature of the defect area to decay from the peak value to the preset decay value 1-1 / e is recorded, and the thermal diffusion time constant is obtained by exponential fitting of the thermal decay curve as the debonding defect detection characteristic.

[0062] With the help of X-ray CT imaging, the defect area is extracted from the CT image by using the gray threshold segmentation method, the percentage of the pore volume to the detection volume is calculated, and the CT gray threshold porosity is obtained as the pore defect detection characteristic.

[0063] Based on the three types of defect detection characteristics, if the detection characteristics are higher than the preset corresponding detection characteristic threshold, the defect boundary of the detection characteristics higher than the preset corresponding detection characteristic threshold is determined, and the defect area and the corresponding defect-free area are determined.

[0064] The defect contained in the defect area can be delamination defect, porosity defect and debonding defect.

[0065] It can be understood that one defect detection area can contain multiple hierarchical defect areas, pore defect areas, and debonding defect areas;

[0066] The hierarchical defect detection feature, the pore defect detection feature, and the debonding defect detection feature are taken as the defect detection features;

[0067] The feature overlap degree of the historical hierarchical micro-defects is obtained by performing overlap detection on the defect detection features.

[0068] The defect area is simplified into sampling points, and the hierarchical defect detection feature, the debonding defect detection feature, and the pore defect detection feature of each sampling point are normalized respectively to construct an overlap detection vector containing the hierarchical defect detection feature, the debonding defect detection feature, and the pore defect detection feature of each sampling point.

[0069] The overlap detection vector is mapped in a three-dimensional space to form a feature distribution area V of each of the three types of defect detection features.

[0070] The feature overlap degree of the historical hierarchical micro-defects is obtained by performing overlap detection on the defect detection features.

[0071] Preferably, the feature overlap degree O of the defect detection area is obtained by the formula

[0072] Wherein, represents the intersection of the feature distribution areas of the i-th type of defect and the j-th type of defect, represents the union of the feature distribution areas of the i-th type of defect and the j-th type of defect.

[0073] Vo represents the volume of the intersection or union of the feature distribution areas.

[0074] It can be understood that the feature overlap degree quantifies the degree of overlap of different types of defect detection features in the spatial distribution, reflects the confusion risk caused by the coupling of the material intrinsic properties or external load of the defect detection features, and provides a basis for the optimization of the multi-defect classification and detection model.

[0075] The role of calculating the feature overlap degree is to:

[0076] Role one, improve the defect detection accuracy and reliability, by fusing the ultrasonic echo amplitude decay rate, thermal diffusion time constant, and CT gray threshold porosity three types of features, normalizing them to construct an overlap detection vector and mapping it to a three-dimensional space, realizing multi-modal feature cross verification of hierarchical, debonding, and pore defects, reducing the one-sidedness and signal overlap of single modal detection, and improving the quantitative fusion accuracy of complex defect detection features.

[0077] ​Function 2: Dynamic load analysis captures defect evolution characteristics: Gradient damp heat load is applied to the feature confusion region. Through feature sensitivity calculation and dynamic response matrix construction, it is beneficial to identify the differences in the dynamic response of defects under multi-field coupling and reduce the problem that traditional static detection cannot reflect defect expansion.

[0078] S2. Based on feature overlap, determine whether different defects exhibit modal feature degradation. If so, construct a hierarchical decomposition model to separate signal confusion caused by thermal stress coupling, evaluate the confusion of defect detection features, obtain the confusion improvement rate, and determine whether difference analysis is needed.

[0079] The method for determining whether different defects exhibit modal feature degradation is as follows:

[0080] Defect spatial correlation analysis was performed on carbon fiber and epoxy resin composite materials to distinguish fixed overlap. and random overlap ;

[0081] Preferably, the method for performing defect spatial correlation analysis on carbon fiber and epoxy resin composite materials is as follows:

[0082] In the historical hierarchical micro-defect inspection, the defect region is constructed into a defect pair by any two sets of defect detection features, including layered defect detection features, porosity defect detection features, and debonding defect detection features;

[0083] Obtain the coordinates of the sampling points corresponding to the two sets of defects, and calculate the Euclidean distance between the coordinates of the two sets of defect sampling points;

[0084] The Euclidean distance between the coordinates of the two sets of defect sampling points is compared with the preset critical distance for each defect pair to classify the defect pairs into spatially adjacent pairs and spatially independent pairs.

[0085] For example, in the testing of carbon fiber and epoxy resin composite materials, delamination and debonding defect pairs are constructed, and the coordinates of two sets of defect sampling points are obtained (e.g., the coordinates of delamination defect point A are (2,3) and the coordinates of debonding defect point B are (6,6)). The Euclidean distance between points A and B is calculated to be 5 mm. If the preset critical distance is 5 mm, this distance is equal to the critical value and is determined to be a spatially adjacent pair. Then, delamination defect point C (1,1) and pore defect point D (8,9) are taken, and the distance is calculated to be 10.63 mm. Since it exceeds the critical value, it is determined to be a spatially unrelated pair. The spatial correlation of defect pairs can be quickly distinguished by Euclidean distance, providing a basis for subsequent modal degradation analysis.

[0086] The probability of spatial proximity pairs appearing in the defect detection area is calculated, and the spatial proximity entropy of the spatial proximity pairs is calculated using the information entropy algorithm.

[0087] Based on spatial proximity entropy, the overlap of inspection features of micro-defects at all historical levels is divided into fixed overlap. and random overlap degree ;

[0088] It needs to be explained that the spatial proximity entropy measures the uncertainty of the spatial distribution of defects; based on the spatial proximity entropy, the historical feature overlap degree is divided into fixed overlap degree and random overlap degree ;

[0089] Set the high entropy value (H=0.97) of the spatial proximity entropy, the high entropy value indicates that the distribution of spatial proximity pairs is random, and the feature overlap may be caused by external interference, and the random overlap degree is dominant; if the spatial proximity entropy is lower than 0.2: indicates that the defects are concentratedly distributed, and the overlap is mainly caused by the intrinsic properties of the material (such as the fiber layer direction, the resin curing shrinkage), and the fixed overlap degree is dominant;

[0090] Different from the historical hierarchical micro-defect inspection, the feature overlap degree of the current hierarchical micro-defect inspection of the carbon fiber and epoxy resin layer composite material is obtained ;

[0091] If the modal feature degradation equation is satisfied: , that is, it is judged whether the modal feature degradation of different defects occurs;

[0092] It needs to be explained that the role of judging whether the modal feature degradation of different defects occurs is:

[0093] Role one, identify the essence of feature confusion: by calculating the spatial proximity pair occurrence probability and the spatial proximity entropy, the feature overlap degree is divided into fixed overlap degree (stable confusion caused by material intrinsic properties) and random overlap degree (abnormal confusion caused by external interference), reducing the misjudgment of material inherent characteristics as defects or ignoring real abnormal confusion;

[0094] Role two, trigger hierarchical decomposition model optimization classification, when it is judged that the modal feature degradation occurs, the three-order tensor low-rank constraint decomposition model is forcibly started, the thermal stress coupling signal is separated, the confusion matrix before and after decomposition is calculated to improve the rate, the reliability of multi-defect type classification is improved, and the signal aliasing misjudgment is reduced;

[0095] Role three, determine the starting condition of dynamic load analysis: if the modal feature degradation and the confusion improvement rate are lower than the threshold value, it indicates that static detection cannot effectively distinguish defects, and dynamic response analysis under gradient hygrothermal load needs to be triggered to capture the dynamic differences of defect detection features under multi-field coupling, and to reduce the problem that traditional static detection cannot reflect defect evolution;

[0096] wherein, is the feature overlap degree of the current carbon fiber and epoxy resin layer composite material, is the average value of the fixed overlap degree of all hierarchical micro-defects, The mean value of the random overlap degree of all levels of micro-defects, k is a preset random overlap coefficient, used to adjust the threshold sensitivity of the modal degradation judgment, and is set according to the detection accuracy requirement;

[0097] In the method, the hierarchical decomposition model is constructed to separate the signal confusion caused by thermal stress coupling, and the confusion evaluation of the defect detection features is performed in the following manner:

[0098] If the modal feature degradation occurs, the three types of defect detection features of the overlap detection vectors of all historical sampling points and the historical P-level thermal stress load are acquired to perform correlation coupling analysis, and it is judged whether the three types of defect detection features and the P-level thermal stress load have correlation coupling;

[0099] As understood by those skilled in the art, the Pearson correlation coefficient r of the three types of defect detection features and the P-level thermal stress load is calculated, and if , it is determined that there is thermal coupling, and it is judged that the three types of defect detection features and the P-level thermal stress load have correlation coupling;

[0100] If there is correlation coupling, based on the three types of defect detection features of the overlap detection vectors of all current sampling points and the P-level thermal stress load, a third-order tensor X is constructed, and is subjected to standardization processing;

[0101] In the formula, M is the total number of sampling points, N is the number of categories of defect detection features, P is the number of levels of thermal stress load, and R is a real number space.

[0102] Preferably, a low-rank constraint is established:

[0103] The third-order tensor is subjected to low-rank constraint decomposition, and a defect classification confusion matrix before and after the low-rank constraint decomposition is constructed.

[0104] In the formula, G, A, B, and C are factor matrices after the third-order tensor decomposition, latent feature components of sampling points, defect detection features, and thermal stress load, respectively, and are used to extract defect detection features independent of thermal stress interference, , is a regularization parameter, used to balance the low-rank constraint of tensor decomposition and fitting error, and reduce overfitting;

[0105] F is a norm, which quantifies the error between the original data and the decomposition result in the tensor decomposition process, realizes the separation accuracy of the thermal stress coupling signal, and simultaneously cooperates with the low-rank constraint term to optimize the reliability of defect detection feature extraction;

[0106] The confusion improvement rate is calculated based on the defect classification confusion matrix, the confusion improvement rate is compared with a preset confusion improvement threshold, if the confusion improvement rate is lower than the preset confusion improvement threshold, difference analysis is required, otherwise, the change of the confusion improvement rate is continuously monitored.

[0107] The skilled in the art can understand that the accuracy rate is obtained by ratio processing of the number of correctly classified samples and the total number of samples, the classification accuracy rates of defects before and after decomposition are calculated respectively, the deviation proportion of the accuracy rate after decomposition and the accuracy rate before decomposition is obtained, and the confusion improvement rate is obtained;

[0108] The purpose of the confusion evaluation is:

[0109] Purpose I: evaluate the effectiveness of the hierarchical decomposition model in separating the thermal stress coupled signals, calculate the confusion improvement rate by constructing the confusion matrix of the defects before and after the low-rank constraint decomposition, judge the separation effect of the model on the defect detection features and the thermal stress coupled signals, and improve the classification and recognition reliability of multiple defect types;

[0110] Purpose II: determine whether to enter the dynamic load difference analysis process, if the confusion improvement rate is lower than the preset threshold, it indicates that the static detection cannot effectively distinguish the defect detection feature confusion, and the dynamic response analysis under the gradient wet heat load is triggered to capture the defect detection feature difference under the multi-field coupling.

[0111] The technical scheme of the embodiment is: in the historical hierarchical micro-defect inspection of carbon fiber and epoxy resin layer composite material, the defect detection features are obtained and the overlap degree detection is performed to obtain the feature overlap degree of the historical hierarchical micro-defect; based on the feature overlap degree, it is used to judge whether the modal feature degradation of different defects occurs, if it occurs, a hierarchical decomposition model is constructed to separate the signal confusion caused by thermal stress coupling, the confusion of the defect detection features is evaluated, the confusion improvement rate is obtained, and it is judged whether difference analysis is needed; it is beneficial to diagnose the feature confusion phenomenon of different defects, reduce the misjudgment caused by signal aliasing, effectively alleviate the interference of thermal stress on the defect detection features, and improve the classification and recognition reliability of multiple defect types.

[0112] Embodiment 2:

[0113] Please refer to Figure 1 The present application is a kind of composite parts internal defect detection method, comprising the following steps:

[0114] S3, if difference analysis is needed, then divide the diagnostic detection interface and apply gradient wet heat load, analyze whether the defect detection feature difference of the diagnostic detection interface is significant, if it is significant, generate the dynamic response matrix under dynamic load;

[0115] The way of dividing the diagnostic detection interface and applying gradient wet heat load is:

[0116] The defect detection area where the modal feature degradation occurs is divided into a diagnostic detection interface;

[0117] According to the load parameter design principle, gradient wet heat load is applied to the defect detection area.

[0118] It can be understood by those skilled in the art that, on the basis of historical thermal stress analysis, the temperature range is set to cover the actual service temperature of the material and extend by 20% (such as -10°C to 150°C), and is gradually increased by 10°C steps to capture the characteristic mutation near the resin glass transition point;

[0119] The humidity gradient is set to 30%RH to 95%RH, 10%RH steps, and each load is moisturized for 4 hours to make the resin moisture balance. At the same time, a static compressive stress of 0.1 times the interlaminar shear strength is applied to the defect detection area with high overlap to simulate the service load, and the multi-field parameters are synchronously controlled by the temperature and humidity environment box and the built-in servo loading system. After each load is stabilized, the defect detection characteristics are dynamically detected by ultrasonic phased array, infrared thermal imaging and CT, so as to obtain the evolution data of delamination, debonding and porosity characteristics under the action of coupled humidity-thermal-stress, and provide a basis for dynamic response matrix construction;

[0120] Wherein, the way to analyze and diagnose whether the difference of the defect detection characteristics of the interface is significant is:

[0121] On the basis of applying gradient humidity-thermal load, the formula is: The characteristic sensitivity S of the ith defect detection characteristic of the defect detection area is obtained i ;

[0122] Wherein, Lmax and Lmin represent the maximum and minimum values of the ith defect detection characteristic under the applied gradient humidity-thermal load, respectively, L is the applied gradient humidity-thermal load, L max and L min represent the maximum and minimum values in the applied gradient humidity-thermal load range, respectively, is the partial derivative symbol;

[0123] For example, Z1 represents a delamination defect detection characteristic; Z2 represents a debonding defect detection characteristic, and Z3 represents a porosity defect detection characteristic;

[0124] It can be understood that the characteristic sensitivity is used to measure the dynamic response capability of the defect detection characteristic under the gradient humidity-thermal load, and can quantify the variation amplitude of the ith defect detection characteristic (such as delamination, debonding, and porosity characteristics) in the gradient humidity-thermal load range, the response slope of the defect detection characteristic to the multi-field coupled load (temperature, humidity, stress), and reflect the evolution sensitivity of the defect under dynamic load;

[0125] Wherein, the role of obtaining the characteristic sensitivity is:

[0126] Role one, quantifying dynamic response, by calculating the variation amplitude of the defect characteristic under the gradient humidity-thermal load, quantifying its dynamic response capability to the multi-field coupled load, reflecting the evolution sensitivity of the defect detection characteristic;

[0127] Action II, verify the validity of the load, determine whether the gradient of the applied wet heat load effectively stimulates the defect feature difference, realizes the dynamic response matrix reflecting the defect behavior, and provides data basis for subsequent analysis;

[0128] Action III, the sensitivity data is used as a feedback parameter to correct the detection model error and improve the adaptability of the system to complex defects;

[0129] Significance test is performed on the feature sensitivity of all defect detection features. If the defect detection feature difference of the diagnostic detection interface is significant, the feature sensitivity of different types of defects is sorted and processed;

[0130] If the feature sensitivity of the layered defect is the highest, the diagnostic detection interface is re-divided according to the feature sensitivity, and the number of divided interface partitions is obtained;

[0131] The dynamic response matrix based on the number of regional partitions, feature dimensions, and load combinations.

[0132] S4, based on the dynamic response matrix, time-frequency coupling analysis is performed to identify the strain gradient of the fiber direction of the layered defect, the mutation direction angle is extracted, and the interference of the anisotropy of the composite material on the mutation direction angle is identified. If there is interference, the defect area is filtered;

[0133] The way to perform time-frequency coupling analysis to identify the strain gradient of the fiber direction of the layered defect is:

[0134] Based on the dynamic response matrix, time-frequency domain conversion is performed to construct the time-frequency matrix W(a,b);

[0135] Preferably, the dynamic response matrix is preprocessed, the dynamic response matrix after data preprocessing is wavelet transformed, the axial strain signal is time-frequency decomposed, and the time-frequency matrix W(a,b) of the strain signal under different loads is obtained;

[0136] Wherein, a is the scale factor of frequency, and b is the translation factor;

[0137] In the time-frequency domain, the spatial gradient of the time-frequency matrix is calculated along the fiber direction of the carbon fiber and epoxy resin layer composite material, and the spatial derivative of the fiber direction is obtained as the strain gradient;

[0138] Wherein, is the coordinate of the fiber direction;

[0139] Based on the strain gradient, the feature frequency band is obtained by extraction and analysis, the strain gradient value of the feature frequency band is mapped to the spatial position, and the mutation point of the strain gradient is extracted;

[0140] The skilled in the art can understand that the characteristic frequency band of energy concentration in the strain gradient is determined by the frequency spectrum analysis (such as screening by frequency energy proportion), the strain gradient value in the frequency band is mapped to the corresponding spatial position coordinates to form a position-strain gradient distribution map; finally, the edge detection algorithm is used to extract the points in the distribution map whose strain gradient mutation amplitude exceeds the preset threshold, which are the strain gradient mutation points of the defect edge, used for locating the hidden boundary of the delamination defect;

[0141] An angle between the main direction of the strain gradient point and the fiber direction is obtained as a mutation direction angle;

[0142] The way of identifying the interference of the anisotropy of the composite material on the mutation direction angle is:

[0143] The strain gradient is decomposed into components in different directions, and the frequency energy proportions of the components in different directions at the characteristic frequency are calculated;

[0144] The anisotropy correlation factor η is obtained by the formula: ;

[0145] wherein, is the frequency energy in the fiber direction, is the angle between the mutation direction angle and the fiber direction, is the sum of the frequency energy proportions of the components in different directions at the characteristic frequency;

[0146] Based on the anisotropy correlation factor η, the mutation direction angles with interference are filtered out, and the defect regions to which the delamination defects corresponding to the mutation direction angles with interference belong are filtered out; It needs to be explained that the purpose of filtering out the mutation direction angles with interference is to exclude the interference of the intrinsic properties (such as the fiber layup direction) of the anisotropic composite material on the detection signal, and to avoid the misjudgment of the defect boundary caused by the strain gradient pseudo-signal in the fiber direction. The interference degree is quantified by calculating the anisotropy correlation factor (η), and the mutation direction angles interfered by the inherent properties of the material are filtered out to ensure that the extracted mutation direction angles truly reflect the strain gradient mutation of the delamination defect edge, thereby locating the hidden boundary of the delamination, improving the accuracy of the spatial position recognition of the defect region, and reducing the problem of missing small defects caused by signal confusion;

[0147] S5, determining the hidden boundary of the delamination region based on the mutation direction angle, determining the boundary hidden region, re-imaging the boundary hidden region, extracting the geometric parameters of the hidden detection features, comparing with the output of the detection model to generate an error correction matrix, feeding back to the hierarchical decomposition model, outputting the corrected defect recognition result and embedding into the composite material quality control system;

[0148]

[0149] ​​​The hidden boundary of the delamination region is determined based on the mutation direction angle, and the boundary hidden region is determined based on the hidden boundary in a manner as follows:

[0150] Preferably, based on the defect region corresponding to the filtered delamination defect, the coordinates of the mutation point of the strain gradient and the mutation direction angle corresponding to the mutation point are obtained, the coordinates of the mutation point and the mutation direction angle are spatially mapped, and a mutation direction angle-position range map is established.

[0151] Based on the mutation direction angle-position range map, an edge detection algorithm is used to identify the region in the distribution map where the direction angle change rate exceeds a preset threshold, which is the hidden boundary of the defect region corresponding to the delamination defect.

[0152] Taking the hidden boundary as a reference, a distance is expanded to both sides to form a strip-shaped region as a boundary hidden region.

[0153] The geometric parameters of the defect detection features are extracted from the re-imaging of the edge defect suspected region in a manner as follows:

[0154] The boundary hidden region is subjected to high-resolution re-imaging scanning by using ultrasonic phased array and infrared thermal imaging detection means, the geometric parameters of the defect, including the length, width, depth, shape profile, and boundary position of the defect, are extracted from the scanning image by using an edge extraction algorithm, and the distribution position of the defect in the composite material level is recorded.

[0155] The extracted geometric parameters are compared with the predicted parameters output by the hierarchical decomposition model, the deviation values of the same parameters are calculated, and an error correction matrix containing the position and size dimension of the delamination defect is generated.

[0156] The error correction matrix is input into the hierarchical decomposition model, and a corrected defect recognition result is output and embedded into the composite material quality control system.

[0157] As can be understood by those skilled in the art, the error correction matrix is input into the hierarchical decomposition model as a feedback signal, the factor matrix (such as G, A, B, and C) of the third-order tensor decomposition and the low-rank constraint parameter are adjusted, the fitting accuracy of the model for complex boundaries and small defects is iteratively optimized, and a corrected defect recognition result is output; finally, the correction result is integrated into the composite material quality control system database, the dynamic update of the detection model is realized, and closed-loop management of the quality control process is realized.

[0158] The technical scheme of the embodiment is as follows: a difference analysis module: if difference analysis is needed, a diagnostic detection interface is divided and a gradient wet heat load is applied, whether the defect detection feature difference of the diagnostic detection interface is significant is analyzed, and if the difference is significant, a dynamic response matrix under dynamic load is generated; based on the dynamic response matrix, a strain gradient of a layered fiber direction is analyzed and identified for time-frequency coupling, a mutation direction angle is extracted, and interference of anisotropy of the composite material on the mutation direction angle is identified, and if the interference exists, a filtering process is performed on a defect area; a hidden boundary of a delamination area is determined based on the mutation direction angle, and a boundary hidden area is determined, the boundary hidden area is re-imaged, geometric parameters of a hidden detection feature are extracted, an error correction matrix is generated after comparison with an output of a detection model, and the error correction matrix is fed back to a hierarchical decomposition model, a corrected defect identification result is output, and the corrected defect identification result is embedded into a composite material quality control system, error feedback is realized to iteratively optimize the hierarchical decomposition model, and self-adaptability and identification reliability of the detection system to different materials and defect types are improved.

[0159] Embodiment 3

[0160] As shown in Figure 3 The present application is a composite part internal defect detection system, comprising the following modules

[0161] Overlap extraction module: for obtaining defect detection features and performing overlap detection in historical hierarchical micro-defect inspection of carbon fiber and epoxy resin layer composite materials, to obtain feature overlap of historical hierarchical micro-defects;

[0162] Confusion analysis module: based on the feature overlap, whether modal feature degradation occurs for different defects is judged, if the degradation occurs, a hierarchical decomposition model is constructed to separate signal confusion caused by thermal stress coupling, confusion evaluation is performed on the defect detection features, confusion improvement rate is obtained, and whether difference analysis is needed is judged;

[0163] Difference analysis module: if difference analysis is needed, a diagnostic detection interface is divided and a gradient wet heat load is applied, whether the defect detection feature difference of the diagnostic detection interface is significant is analyzed, and if the difference is significant, a dynamic response matrix under dynamic load is generated;

[0164] Mutation identification module: based on the dynamic response matrix, a strain gradient of a layered fiber direction is analyzed and identified for time-frequency coupling, a mutation direction angle is extracted, and interference of anisotropy of the composite material on the mutation direction angle is identified, and if the interference exists, a filtering process is performed on a defect area;

[0165] Feedback correction module: based on the mutation direction angle, a hidden boundary of a delamination area is determined, and a boundary hidden area is determined, the boundary hidden area is re-imaged, geometric parameters of a hidden detection feature are extracted, an error correction matrix is generated after comparison with an output of a detection model, the error correction matrix is fed back to a hierarchical decomposition model, a corrected defect identification result is output, and the corrected defect identification result is embedded into a composite material quality control system.

[0166] The above detailed description has shown, by way of example, an embodiment of the application. It is specifically contemplated that the application is not limited to the embodiments described herein, but rather the scope of the application is defined by the claims.

Claims

1. A composite material part internal defect detection method, characterized in that: in the historical hierarchical micro-defect inspection of carbon fiber and epoxy resin layer composite material, defect detection features are obtained and overlap detection is performed to obtain feature overlap of historical hierarchical micro-defects; the overlap detection is performed in the following manner: an overlap detection vector is obtained, the overlap detection vector is mapped in a three-dimensional space to form feature distribution regions of three types of defect detection features respectively; the intersection and union of the feature distribution regions of different types of defect detection features are calculated, and overlap detection is performed to obtain the feature overlap; based on the feature overlap, it is judged whether modal feature degradation occurs for different defects, and if it occurs, a hierarchical decomposition model is constructed to separate the signal confusion caused by thermal stress coupling, the defect detection features are evaluated for confusion, an improvement rate of confusion is obtained, and it is judged whether difference analysis is needed; the defect detection features are evaluated for confusion in the following manner: if modal feature degradation occurs, the three types of defect detection features of the overlap detection vector of all historical sampling points are obtained, and historical P-level thermal stress loads are analyzed for correlation coupling to judge whether the three types of defect detection features and the P-level thermal stress loads have correlation coupling; wherein P is the number of thermal stress loads; if there is correlation coupling, a three-order tensor is constructed and low-rank constraint decomposition is performed, and a defect classification confusion matrix before and after the low-rank constraint decomposition is constructed; the improvement rate of confusion is calculated based on the defect classification confusion matrix, and if the improvement rate of confusion is lower than a preset confusion improvement threshold, difference analysis is needed; if difference analysis is needed, a diagnostic detection interface is divided and a gradient humid heat load is applied, and it is analyzed whether the defect detection feature difference of the diagnostic detection interface is significant, and if it is significant, a dynamic response matrix under dynamic load is generated; based on the dynamic response matrix, time-frequency coupling analysis is performed to identify the strain gradient of the layered fiber direction, a mutation direction angle is extracted, and the interference of the anisotropy of the composite material on the mutation direction angle is identified, and if there is interference, the defect area is filtered. The overlap detection vector is obtained in the following manner: layered defect detection features, pore defect detection features, and debonding defect detection features of the defect area are obtained, and the three types of defect detection features of the defect area are taken as defect detection features; the defect area is simplified into sampling points, and an overlap detection vector containing the defect detection features of each sampling point is constructed. The judgment manner of the modal feature degradation is as follows: a defect pair of any two groups of defect detection features of layered defect detection features, pore defect detection features, and debonding defect detection features is constructed; the coordinates of the sampling points corresponding to the two groups of defects in the defect pair are obtained, the Euclidean distance of the coordinates of the two groups of defect sampling points is calculated, and the defect pair is divided into a spatially adjacent pair and a spatially unrelated pair; the probability of the spatially adjacent pair appearing in the defect detection area is calculated, the spatial proximity entropy of the spatially adjacent pair is calculated through an information entropy algorithm, and the feature overlap is divided into fixed overlap and random overlap; the feature overlap of the current hierarchical micro-defect inspection is obtained, and a modal feature degradation equation is established based on the feature overlap of the current carbon fiber and epoxy resin layer composite material, the average value of the fixed overlap of all hierarchical micro-defects, and the average value of the random overlap of all hierarchical micro-defects. ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ 2. The method of claim 1, wherein: ​ ​ ​ 3. The method of claim 1, wherein: ​ ​ ​ ​ ​ If the modal characteristic degradation equation is satisfied, i.e., modal characteristic degradation occurs.

4. The method of claim 1, wherein: The manner of analyzing whether the defect detection characteristic difference of the diagnostic detection interface is significant is: The defect detection area where modal characteristic degradation occurs is divided into a diagnostic detection interface; Gradient humidity and heat load is applied to the defect detection area to obtain the characteristic sensitivity of different types of defect detection characteristics in the defect detection area, and if the defect detection characteristic difference of the diagnostic detection interface is significant, the characteristic sensitivity of different types of defects is sorted and processed; If the characteristic sensitivity of the delamination defect is the highest, the diagnostic detection interface is re-divided according to the characteristic sensitivity, the number of divided interface partitions is obtained, and a dynamic response matrix is constructed.

5. The method of claim 1, wherein: The manner of performing the filtering processing on the defect area is: Obtain the strain gradient and the characteristic frequency band, map the strain gradient value of the characteristic frequency band to the spatial position, and extract the mutation point of the strain gradient; Obtain the angle between the main direction of the mutation point of the strain gradient and the fiber direction to obtain the mutation direction angle; The strain gradient is decomposed into components in different directions, the frequency domain energy proportion of each direction component of the strain gradient at the characteristic frequency is calculated, and an anisotropy correlation factor is calculated; Based on the anisotropy correlation factor, the mutation direction angle with interference is filtered out, and the delamination defect belonging to the defect area with interference is filtered out.

6. The method of claim 1, wherein: Further comprising the following steps: Determine the hidden boundary of the delamination area based on the mutation direction angle, determine the boundary hidden area, re-image the boundary hidden area, extract the geometric parameters of the hidden detection characteristics, compare with the detection model output to generate an error correction matrix, feed back to the hierarchical decomposition model, output the corrected defect recognition result and embed into the composite material quality control system.

7. A system for detecting internal defects in a composite article, the system comprising: A composite material part internal defect detection method for realizing any one of claims 1-6, comprising the following modules: Overlap extraction module: used for obtaining defect detection characteristics and performing overlap detection in the historical hierarchical micro-defect inspection of carbon fiber and epoxy resin layer composite material, to obtain the characteristic overlap degree of the historical hierarchical micro-defect; Confusion analysis module: used for judging whether different defects have modal characteristic degradation based on the characteristic overlap degree, if yes, constructing a hierarchical decomposition model to separate the signal confusion caused by thermal stress coupling, evaluating the defect detection characteristics, obtaining the confusion improvement rate and judging whether difference analysis is needed; Difference analysis module: if difference analysis is needed, divide the diagnostic detection interface and apply gradient humidity and heat load, analyze whether the defect detection characteristic difference of the diagnostic detection interface is significant, if yes, generate a dynamic response matrix under dynamic load; Mutation recognition module: used for performing time-frequency coupling analysis to identify the strain gradient of the delamination fiber direction based on the dynamic response matrix, extracting the mutation direction angle and recognizing the anisotropy of the composite material to interfere with the mutation direction angle, and filtering the defect area if there is interference.

Images