CFRP defect and performance prediction method based on ultrasonic detection and finite element simulation

Abstract

The invention relates to the field of carbon fiber reinforced resin matrix composite detection, in particular to a CFRP defect and performance prediction method based on ultrasonic detection and finite element simulation. A set of closed-loop automatic evaluation system from defect identification to performance prediction is established by taking construction of a standard sample library and multi-source data acquisition as a basis and combining a convolutional neural network model and a finite element simulation data driven regression model; the technical problems that a traditional CFRP structure detection and performance evaluation link is separated, manual experience is relied on, efficiency is low, and precision is insufficient are effectively solved. Specifically, according to the scheme, a convolutional neural network model is trained through ultrasonic scanning signals and prefabricated defect parameters, intelligent and accurate recognition of the depth and area of internal defects is achieved, and the defects that traditional ultrasonic detection depends on manual interpretation and is poor in consistency are overcome.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of carbon fiber reinforced resin matrix composite material detection, and in particular to a CFRP defect and performance prediction method based on ultrasonic detection and finite element simulation. BACKGROUND

[0002] Carbon fiber reinforced resin matrix composite material (CFRP) is a high-performance structural material with carbon fiber as the reinforcing phase and resin as the matrix. It has high specific strength, low density and excellent corrosion resistance, and is widely used in aerospace, rail transportation and new energy vehicles. Due to the multi-layered structure of CFRP, internal defects such as interlaminar delamination and delamination are easily produced under complex load or fatigue conditions, which leads to a decrease in load-bearing performance and structural safety hazards. Therefore, how to accurately detect the internal defects of CFRP and evaluate their influence on mechanical properties is a key technical problem in the engineering application of composite materials.

[0003] The existing CFRP internal defect detection method mainly uses ultrasonic detection technology to obtain internal echo signals of the material through ultrasonic C-scan or A-scan mode, and uses the pulse echo method to measure the amplitude and time difference of the signal to determine the existence and position of the defect. In recent years, some research has tried to combine signal processing algorithms to extract defect depth and area parameters, thereby realizing quantitative detection. On the other hand, finite element simulation methods are widely used in the analysis and damage prediction of the mechanical properties of composite materials. By setting different defect sizes and depths in the simulation model, the load response characteristics of the material can be obtained, providing a reference for performance evaluation. The above detection methods and simulation analysis methods have been applied in engineering, providing an important means for the structural evaluation of composite materials.

[0004] The existing technology still has problems such as separation of detection and performance evaluation, data flow interruption and insufficient intelligence. Specifically, the recognition of defect signal characteristics by ultrasonic detection still relies on manual experience, and lacks automated and high-precision recognition capabilities; the defect parameters obtained by detection cannot be directly imported into the finite element simulation model, resulting in the inability of simulation calculation based on real defect morphology; there is a lack of unified modeling mechanism between detection data and performance response data, making it difficult to form a calculable mapping relationship between defect parameters and mechanical properties; at the same time, the detection, modeling and prediction processes are independent of each other, and do not form a closed-loop system from ultrasonic detection to performance prediction. The above problems result in the performance evaluation of existing CFRP structures still relying on step-by-step operation and manual intervention, making it difficult to meet the demand for high-precision, rapid and intelligent evaluation in engineering applications. SUMMARY

[0005] The application provides a CFRP defect and performance prediction method based on ultrasonic detection and finite element simulation, and solves the problem of the existing technology that the CFRP structure detection and performance evaluation links are disconnected, rely on artificial experience, are inefficient, and the traditional method cannot realize integrated closed-loop evaluation from nondestructive testing to accurate prediction of residual strength.

[0006] To achieve the above object, the embodiments of the application disclose the following technical solutions:

[0007] The application discloses a CFRP defect and performance prediction method based on ultrasonic detection and finite element simulation, comprising the following steps:

[0008] S1, a carbon fiber reinforced resin matrix composite defect sample containing delamination defects is prepared, and during the lamination process of the carbon fiber reinforced resin matrix composite, a plurality of polytetrafluoroethylene sheets with different areas are embedded at a set lamination depth, and after curing and forming, a plurality of defect samples containing different delamination defect areas are processed;

[0009] S2, ultrasonic C-scan detection is performed on the defect sample, an ultrasonic probe is excited by using a pulse echo method, ultrasonic A-scan echo signals of the defect sample are obtained, and time information and amplitude information are taken as original ultrasonic detection signals;

[0010] S3, the original ultrasonic detection signals are input into a convolutional neural network model, corresponding delamination defect depth parameters, delamination defect area parameters and delamination defect region length parameters are output, and a defect parameter data set is formed;

[0011] S4, a finite element simulation model of the carbon fiber reinforced resin matrix composite is established according to the defect parameter data set, the delamination defect parameters are mapped into a finite element grid structure, compression simulation calculation is performed, compression load response data are obtained, and a performance response data set is formed;

[0012] S5, feature fusion is performed on the defect parameter data set and the performance response data set, and a mathematical model between the defect parameters and the compression load response is established by using a least square method or a nonlinear fitting algorithm;

[0013] S6, the ultrasonic detection signals of the carbon fiber reinforced resin matrix composite to be tested are input into the convolutional neural network model, delamination defect parameters of the test sample are obtained, predicted compression performance values of the test sample are calculated based on the mathematical model, and a performance evaluation result is output.

[0014] The application creates a closed-loop automatic evaluation system from defect identification to performance prediction based on the construction of a standard sample library and multi-source data acquisition, combined with a convolutional neural network model and a regression model driven by finite element simulation data, effectively solving the technical problems of the traditional CFRP structure detection and performance evaluation link, such as fragmentation, dependence on artificial experience, low efficiency and insufficient precision. Specifically, the scheme uses ultrasonic scanning signals and pre-prepared defect parameters to train a convolutional neural network model, achieving intelligent and accurate identification of internal defect depth and area, overcoming the shortcomings of traditional ultrasonic detection, such as dependence on artificial interpretation and poor consistency. At the same time, by fusing physical test and finite element simulation data to establish a performance prediction mathematical model, the remaining bearing capacity can be quickly calculated after obtaining the defect parameters, avoiding the tedious process of repeated destructive testing or complex simulation modeling. Finally, by integrating the above two models into the ultrasonic signal of the component to be tested, an integrated process of non-destructive testing, parameter extraction, performance prediction and safety decision is realized, significantly improving the evaluation efficiency and reliability, and providing strong technical support for the rapid and safe evaluation of CFRP structures in high-end equipment fields such as aerospace. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 A flowchart of the method of the embodiments of the application is provided.

[0016] Figure 2 A model diagram of the linear regression equation of the method of the embodiments of the application is provided.

[0017] Figure 3 A decision curve diagram of the method of the embodiments of the application is provided. DETAILED DESCRIPTION

[0018] Reference will now be made in detail to the specific implementations of the application. While the application will be described in conjunction with these specific implementations, it will be understood that they are not intended to limit the application to these specific implementations. On the contrary, the application is intended to cover alternatives, modifications, and equivalents, which can be included within the spirit and scope of the application as defined by the claims. In the following description, numerous specific details are set forth in order to provide a thorough understanding of the application. The application can be practiced without some or all of these specific details. In other instances, well known process operations have not been described in detail in order not to unnecessarily obscure the application.

[0019] The singular forms "a," "an," and "the" include plural referents unless the context clearly dictates otherwise. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0020] Example One

[0021] The CFRP defect and performance prediction method based on ultrasonic detection and finite element simulation comprises the following steps:

[0022] S1, a standard specimen library of carbon fiber reinforced resin-based composite materials containing known defect parameters is prepared;

[0023] S2, a data acquisition step: ultrasonic detection signals, real mechanical property data and simulation mechanical property data based on the defect parameters of each specimen in the standard specimen library are acquired;

[0024] S3, a defect identification model construction step: based on the ultrasonic detection signals and the corresponding known defect parameters, a machine learning model is trained to obtain a defect identification model capable of identifying defect parameters from ultrasonic signals;

[0025] S4, a performance prediction model construction step: based on the real mechanical property data and / or simulation mechanical property data and the corresponding defect parameters, a performance prediction model capable of predicting mechanical properties is established through data fitting;

[0026] S5, ultrasonic detection is performed on the to-be-tested component to obtain its ultrasonic detection signal;

[0027] S6, the ultrasonic detection signal of the to-be-tested component is input into the defect identification model to output the identification result of its defect parameters;

[0028] S7, the identification result of the defect parameters is input into the performance prediction model to output the predicted value of its mechanical properties;

[0029] S8, based on the predicted value of the mechanical properties, the safety of the to-be-tested component is evaluated.

[0030] In the specific implementation of the present application, a standard specimen library is first prepared. During the lamination process of carbon fiber reinforced resin-based composite materials, polytetrafluoroethylene sheets of different areas are embedded at a set lamination depth, and then the lamination is continued to the designed thickness. After curing and forming, n defect specimens are processed, and the delamination defect area and the delamination defect depth are known parameters. Subsequently, data acquisition is performed, including ultrasonic C-scan detection of the defect specimens to obtain ultrasonic A-scan echo signals and compression testing of the defect specimens to obtain real compression loads and finite element simulation based on known defect parameters to obtain simulation compression loads Then the defect recognition model is constructed, that is, the convolution neural network model is trained with the ultrasonic A-scan echo signal as the input and the corresponding known defect parameters as the target output. Meanwhile, the performance prediction model is constructed, that is, the mathematical model is established by the least square method based on the obtained compression load data and the corresponding defect parameters. The model is obtained by solving the normal equation set, and the normal equation set is: ;

[0031] The equation set is written in the matrix form:

[0032] ;

[0033] Among them, , , , and the following is obtained:

[0034] ;

[0035] According to the matrix form, the compression load of the defect sample and the slope and the longitudinal intercept of the fitting line of the delamination defect area and the delamination defect depth are obtained, that is, the mathematical model of the delamination defect area , the delamination defect depth , and the compression load of the defect sample is obtained: .

[0036] Then, the prediction application is carried out. The ultrasonic detection is carried out on the to-be-tested component, the signal is input into the defect recognition model, the defect parameter recognition result is output, the result is input into the performance prediction model, the mechanical performance prediction value is output, and the safety is evaluated according to the mechanical performance prediction value.

[0037] Among them, represents the delamination defect area of the defect sample i;

[0038] Un represents the delamination defect area of the defect sample n;

[0039] represents the delamination defect depth of the defect sample i;

[0040] Vn represents the delamination defect depth of the defect sample n;

[0041] represents the compression load of the defect sample i;

[0042] Wn represents the compression load of the defect sample n;

[0043] a vertical intercept of a mathematical model fitting curve representing the delamination defect area Ui of the defect sample i and the compressive load Wi of the defect sample i;

[0044] a slope of a mathematical model fitting curve representing the delamination defect area Ui of the defect sample i and the compressive load Wi of the defect sample i;

[0045] a slope of a mathematical model fitting curve representing the delamination defect depth Vi of the defect sample i and the compressive load Wi of the defect sample i;

[0046] T represents a symbol of matrix transposition.

[0047] As Figure 2 , a model graph of a linear regression equation.

[0048] The scheme further proposes that the data acquisition step in the step S2 specifically comprises: performing ultrasonic C scanning on each standard sample to obtain ultrasonic A scanning echo signals, performing a compression test to obtain a real compressive load, and performing finite element simulation to obtain a simulated compressive load.

[0049] The specific implementation of the data acquisition step is to perform ultrasonic C scanning on each standard sample to obtain ultrasonic A scanning echo signals. The compression test is performed to obtain a real compressive load , and the finite element simulation is performed to obtain a simulated compressive load . The ultrasonic C scanning uses an ultrasonic probe with pulse echo method to scan the sample surface at a fixed step and , and records the echo signals of each point.

[0050] The compression test is performed on a universal testing machine until the sample is destroyed to record the limit load. The finite element simulation is based on known defect parameters and to establish a composite laminate model in software to perform compression calculation and obtain a simulated compressive load .

[0051] The scheme further proposes that the machine learning model in the step S3 is a convolutional neural network model.

[0052] The machine learning model is specifically implemented as a convolutional neural network model. The model structure includes an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer. The input layer receives one-dimensional ultrasonic A scanning echo signal time domain waveform data. The output layer sets neurons to respectively output the time and amplitude of the defect echo position. The training process uses the time of the defect echo position in the ultrasonic A scanning echo signals of all n defect samples.and amplitude The CNN model is trained as a training sample after integration.

[0053] The performance prediction model in step S4 is further proposed to be established by a regression analysis algorithm.

[0054] The regression analysis algorithm used by the performance prediction model is specifically implemented as a least squares method. The algorithm solves model coefficients by constructing a normal equation system. The specific mathematical model is a binary linear regression equation with the layered defect area and the layered defect depth as independent variables and the compression load as the dependent variable, and its form is:

[0055] ;

[0056] Among them, the regression coefficients , , are obtained by least squares estimation.

[0057] The scheme further proposes that before constructing the performance prediction model in step S4, the real mechanical performance data and the simulation mechanical performance data are compared to verify the accuracy of the simulation model.

[0058] The specific implementation of verifying the accuracy of the simulation model is to compare the real compression load with the simulation compression load . In turn, the test compression load of the defect sample 1, the test compression load of the defect sample i, the test compression load of the defect sample n, the finite element simulation compression load of the defect sample 1, the finite element simulation compression load of the defect sample i, and the finite element simulation compression load of the defect sample n are obtained. Comparing the test and finite element simulation compression loads verifies the accuracy of the finite element simulation.

[0059] The scheme further proposes that the safety evaluation in step S8 is achieved by comparing the predicted value with a preset safety threshold.

[0060] The specific implementation of safety evaluation is achieved by comparing the predicted compression load with a preset safety threshold. The safety threshold is set as a certain percentage of the compression load of the defect-free sample. If the compression load of the test sample reaches 80% of the compression load corresponding to the defect sample with an area of 0 of the polytetrafluoroethylene sheet, the performance of the test sample is qualified.

[0061] Example Two

[0062] This embodiment further refines the method described in Example One, particularly the specific construction and application method of the defect recognition model. First, a standard sample library is prepared. In specific implementation, the area of the polytetrafluoroethylene sheet should be less than 100 mm², and the thickness is 0.02 mm to 0.025 mm. The defect sample processed after solidification has a size of 140 mm x 13 mm x 4 mm, to ensure the normativity of the test and the comparability of the results.

[0063] In the ultrasonic C-scan step of data acquisition, the ultrasonic probe scans along the sample surface with a fixed transverse step distance Lx and a longitudinal step distance Ly. The number of step points judged as delamination defects is denoted as m, and the length L of the defect area can be calculated by the formula L = m x Lx. This is a key step for quantifying the geometric size of the defect from the C-scan image.

[0064] In constructing the defect recognition model, the training target of the convolutional neural network model can be set as the time T and amplitude A of the defect echo position. More preferably, the defect recognition can be comprehensively determined in combination with the determination curve, such as Figure 3 . The specific method is to construct a depth-amplitude curve based on the known defect depth H and ultrasonic A-scan echo amplitude A under the condition of determining the length L of the delamination defect area; then comprehensively construct the complete determination curve based on the depth-amplitude data by synthesizing the depth-amplitude curves under different defect area lengths L. In application, the depth Hj and amplitude Aj extracted from the ultrasonic signal of the sample to be tested by the CNN model correspond to the determination curve, and the length Lj of the defect area can be accurately determined. This method fully utilizes the topographic information of ultrasonic C-scan and the signal characteristics of A-scan, and improves the accuracy and reliability of defect parameter recognition.

[0065] Before constructing the performance prediction model, the finite element model needs to be verified. The real compression load Yi of the defect sample is compared with the finite element simulation compression load Zi. If the error between the two is within an acceptable range, it is proved that the simulation model is accurate, and it can be used to construct the performance prediction model by data augmentation. Finally, in the safety evaluation stage, the predicted compression load Wj of the sample to be tested is compared with the compression load of the defect-free sample. If the predicted value reaches more than eighty percent of the bearing capacity of the defect-free sample, the performance of the sample is determined to be qualified. This threshold value can be adjusted according to the safety margin of actual engineering application.

[0066] It should be pointed out finally that the above examples are only used to illustrate the technical solutions of the present application but not to limit it; although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the specific embodiments of the present application can be modified or some technical features can be replaced by equivalent ones without departing from the spirit of the technical solutions of the present application, and all of them should be covered in the technical solution range claimed by the present application.

Claims

1. A method for CFRP defect and performance prediction based on ultrasonic testing and finite element simulation, characterized in that, The method comprises the following steps: S1, preparing a standard sample library of carbon fiber reinforced resin matrix composites containing known defect parameters; S2, a data acquisition step: acquiring the ultrasonic detection signal, real mechanical property data, and simulated mechanical property data based on the defect parameters of each sample in the standard sample library; S3, a defect identification model construction step: training a machine learning model based on the ultrasonic detection signal and its corresponding known defect parameters to obtain a defect identification model capable of identifying defect parameters from ultrasonic signals; S4, a performance prediction model construction step: based on the real mechanical property data or simulated mechanical property data and its corresponding defect parameters, a performance prediction model capable of predicting mechanical properties is established through data fitting; S5, ultrasonic detection of the to-be-tested component to obtain its ultrasonic detection signal; S6, inputting the ultrasonic detection signal of the to-be-tested component into the defect identification model to output the identification result of its defect parameters; S7, inputting the identification result of the defect parameters into the performance prediction model to output the predicted value of its mechanical properties; S8, based on the predicted value of the mechanical properties, the safety of the to-be-tested component is evaluated.

2. The method for CFRP defect and performance prediction based on ultrasonic testing and finite element simulation according to claim 1, characterized in that, The data acquisition step in step S2 specifically comprises: performing ultrasonic C-scan on each standard sample to obtain ultrasonic A-scan echo signals, performing compression test to obtain real compression load, and performing finite element simulation to obtain simulated compression load. 3.The CFRP defect and performance prediction method based on ultrasonic detection and finite element simulation according to claim 1, wherein, The machine learning model in step S3 is a convolutional neural network model.

4. The method for CFRP defect and performance prediction based on ultrasonic testing and finite element simulation according to claim 1, characterized in that, The performance prediction model in step S4 is established by a regression analysis algorithm.

5. The method for CFRP defect and performance prediction based on ultrasonic testing and finite element simulation according to claim 1, characterized in that, In step S4, before constructing the performance prediction model, the real mechanical property data and the simulated mechanical property data are compared to verify the accuracy of the simulation model.

6. The method for CFRP defect and performance prediction based on ultrasonic testing and finite element simulation according to claim 1, characterized in that, The safety evaluation in step S8 is realized by comparing the predicted value with a preset safety threshold.

Images