Method and system for analyzing integrity of CFRP after lightning stroke

Abstract

The invention provides a CFRP post-lightning stroke integrity analysis method and system, and belongs to the technical field of electric digital data processing. According to the invention, the portable ATR-FTIR test and the artificial neural network are combined, the CFRP integrity analysis nondestructive testing method after lightning stroke is provided, and the residual strength of the composite material after lightning stroke can be accurately predicted. According to the method, the CFRP integrity after lightning stroke can be directly, accurately and efficiently evaluated, the destructiveness and high cost of a traditional method can be avoided, the defects that an existing neural network model is complex in structure and accumulates errors can be overcome, and more concise and accurate CFRP integrity evaluation after lightning stroke is achieved.

Description

Technical Field

[0001] The present invention belongs to the technical field of electrical digital data processing, relates to the use of machine learning in computer-aided design, relates to neural networks, and particularly relates to a method and system for analyzing the integrity of CFRP after a lightning strike. Background Art

[0002] Carbon fiber reinforced polymer composites (CFRP) have been widely used in aerospace, transportation and other fields. However, when CFRP is used in the main load-bearing structures of modern engineering structures such as aircraft, due to the insufficient electrical conductivity of CFRP, it is difficult to quickly conduct high-amplitude lightning currents, and is exposed to the threat of lightning strikes. Although modern aircraft have adopted lightning strike protection (LSP) systems such as metal mesh / metal sheets, if the LSP system is not properly designed, the CFRP surface may still suffer thermal damage after being struck by lightning. This thermal damage can sometimes be detected through visual inspection and manifests as blistering or delamination. In the long run, this type of thermal damage can have catastrophic consequences. Therefore, accurately assessing the structural integrity of CFRP after a lightning strike is crucial to improving flight safety, reducing maintenance costs and extending the service life of modern composite aircraft.

[0003] Traditional composite material integrity analysis methods include in-plane compression tests, drop hammer impact tests, or quasi-static indentation tests. Liu Xiaoming's team conducted axial compression tests to evaluate the residual strength of carbon fiber composite laminates after lightning strikes (Liu Xiaoming, Yu Xiaosang, Wang Jiu, et al. Axial compression test of carbon fiber composite laminates with different protection after lightning strikes [J]. Journal of Materials Science and Engineering, 2016(3):4. DOI:10.14136 / j.cnki.issn1673-2812.2016.03.008.). However, this test requires sampling from engineering structures and testing in the laboratory. In aerospace applications, the cost of preparing specimens from aircraft is extremely high, so on-site non-destructive testing (NDI) may be an effective way to solve this problem. However, existing ultrasonic testing can only detect the location and area of ​​damage, and cannot quantitatively evaluate the degree of damage and residual mechanical properties.

[0004] The prior art Chinese invention patent CN118571381A discloses a neural network-based lightning damage assessment method for composite materials. The method predicts damage area, damage depth, and residual strength by cascading three neural network models. This prior art has the following disadvantages: (1) The process is cumbersome, and the cascaded network structure causes errors to accumulate step by step, resulting in reduced prediction accuracy; (2) The macroscopic mechanical properties of the material are actually determined by its microscopic chemical structure (determined by the chemical bond state and molecular structure). Optical images can only reflect the damage morphology and cannot directly reveal the relationship between structure and performance. This prior art relies on visual images as input, which can only capture macroscopic morphological features and cannot reflect changes in the material's microscopic chemical structure.

[0005] The prior art Chinese invention patent CN119164911A discloses a method and system for determining the degree of aging on the surface of silicone rubber. This method uses FTIR data to train a neural network model to achieve aging degree classification. This prior art has the following disadvantages: (1) its application is limited to silicone rubber materials, and the evaluation target is the degree of aging rather than the residual strength. It fails to establish a direct quantitative relationship (mapping relationship) between chemical structure and mechanical properties; (2) this method does not consider the material response characteristics under extreme thermal effects such as lightning strikes, and its model structure and feature extraction method are not suitable for lightning damage assessment of CFRP. Summary of the Invention

[0006] The present invention is made to solve the above problems, and its purpose is to provide a method and system for analyzing the integrity of CFRP after lightning strike.

[0007] The present invention provides a method for analyzing the integrity of CFRP after a lightning strike, which has the following characteristics and includes the following steps: S10, performing high-temperature exposure tests on CFRP at different temperatures, comparing the changes in its infrared spectrum, thereby determining the chemical bond evolution law of CFRP at different temperatures to determine the key chemical structure therein; S20, using an ATR-FTIR spectrometer to scan the lightning-struck area of ​​CFRP to obtain its infrared spectrum historical data, selecting the characteristic peaks of the key chemical structure therein and calculating the intensity as target input data; S30, obtaining the mechanical property historical data of CFRP after a lightning strike through a mechanical property test, and using it as target output data; S40, using an artificial neural network to establish a mapping relationship between the target input data and the target output data, thereby training a prediction model for the integrity of CFRP after a lightning strike; S50, using an ATR-FTIR spectrometer to scan the lightning-struck area of ​​CFRP that needs to be actually analyzed to obtain its infrared spectrum historical data, selecting the characteristic peaks of the key chemical structure therein, calculating the intensity, and inputting them into the prediction model, outputting the corresponding mechanical property data of CFRP, thereby realizing the integrity analysis of CFRP after a lightning strike.

[0008] The CFRP post-lightning strike integrity analysis method provided by the present invention may also have the following characteristics: wherein, in step S10, the CFRP is placed in a muffle furnace and ablated at a constant temperature of 100°C to 300°C in an air atmosphere, and the infrared spectrum changes of the CFRP at different temperatures are compared, thereby obtaining the chemical bond evolution law of the CFRP at different temperatures to determine the key chemical structure therein.

[0009] The CFRP post-lightning strike integrity analysis method provided by the present invention may also have the following characteristics: wherein, in steps S10 to S20, the key chemical structures include the CH stretching vibration of the methyl and methylene groups, the stretching vibration of the carbonyl group, and the skeletal vibration of the benzene ring.

[0010] The CFRP post-lightning strike integrity analysis method provided by the present invention may also have the following feature: wherein, in step S20, baseline adjustment and smoothing are performed on the infrared spectrum historical data.

[0011] The CFRP post-lightning strike integrity analysis method provided by the present invention may also have the following features: wherein, in step S30, the mechanical property test includes a three-point bending test, The historical data of mechanical properties include data on flexural strength.

[0012] The CFRP post-lightning strike integrity analysis method provided by the present invention may also have the following features: wherein, in step S40, constraints are set for the prediction models constructed under different training conditions, and the constraint settings include the maximum number of iterations and the minimum training target convergence error, and the artificial neural network is a BP neural network.

[0013] The CFRP post-lightning strike integrity analysis method provided by the present invention may also have the following features: wherein, in step S40, the prediction models under different training conditions are also evaluated, and the accuracy of their prediction results is reflected by calculating the relative error. If the relative error does not meet the evaluation requirements, the prediction accuracy of the prediction model is improved by expanding the training data set of the target input data and the target output data and optimizing the structure of the artificial neural network until the relative error meets the evaluation requirements.

[0014] The CFRP post-lightning strike integrity analysis method provided by the present invention may also have the following characteristics: wherein, optimizing the structure of the artificial neural network includes adjusting the number of layers, adjusting the number of neurons in the hidden layer, and optimizing the activation function.

[0015] The CFRP post-lightning strike integrity analysis method provided by the present invention may also have the following features: wherein, in step S50, the CFRP post-lightning strike integrity analysis further includes quantitatively determining the concentration of key chemical structures. cThe relationship between the characteristic peak intensity of the key chemical structure and the characteristic peak intensity is determined by the absorbance value in the infrared spectrum historical data. A Calculate the difference from the baseline and the concentration of key chemical structures c : , in the above formula, θ is the angle of incidence, n s and n c are the refractive indices of CFRP and ATR crystal, K is the molar absorption coefficient, K Depends on the material of CFRP and the wavelength of incident light λ .

[0016] The present invention also provides a CFRP post-lightning strike integrity analysis system having the following characteristics: it uses any of the aforementioned CFRP post-lightning strike integrity analysis methods, and includes: a data input module for a user to input target input data and target output data therein; a model construction and training module, based on an artificial neural network and connected to the data input module, for establishing a mapping relationship between target input data and target output data, thereby training a CFRP post-lightning strike integrity prediction model; a prediction analysis module, connected to the model construction and training module, for a user to input the intensity of characteristic peaks of key chemical structures in infrared spectrum historical data of a CFRP lightning strike area that needs to be actually analyzed, and then output mechanical property data of the CFRP corresponding to the prediction model trained by the model construction and training module; and an output module, connected to the prediction analysis module, for outputting and displaying the mechanical property data and generating an integrity evaluation report.

[0017] The CFRP post-lightning strike integrity analysis method and system of the present invention have the following beneficial effects: (1) This paper proposes a nondestructive testing method for analyzing the integrity of CFRP after a lightning strike by combining portable ATR-FTIR testing with an artificial neural network. This method can accurately predict the residual strength of composite materials after a lightning strike. The handheld ATR-FTIR can complete on-site chemical structure acquisition without disassembly or sampling. In conjunction with a trained prediction model (BP-ANN model), it can instantly output the residual bending strength, enabling on-site testing and quantitative evaluation, avoiding the high cost, long cycle time, and difficulty in quantification associated with traditional evaluation methods.

[0018] (2) The present invention relies on the sensitivity of FTIR to the characteristic functional groups (-C=O, -CH2-, benzene ring skeleton, etc.) of composite materials after lightning strikes, and establishes a nonlinear mapping relationship between the chemical structure concentration of composite materials after lightning strikes and the macroscopic residual strength. For the first time, the direct correlation between the chemical degradation of the matrix and the bearing capacity of the structure is realized, providing data support for revealing the mechanism of lightning damage, predicting life span and formulating maintenance standards.

[0019] (3) The present invention dynamically optimizes the number of network layers, the number of nodes, and the activation function, and the prediction error is stably converged to <5%. This method provides a new method for non-destructive on-site evaluation of aircraft composite structures and a general method for evaluating the residual mechanical properties of composite structures under thermal damage in other situations.

[0020] (4) The present invention can directly, accurately and efficiently evaluate the integrity of CFRP after lightning strike, which can avoid the destructiveness and high cost of traditional methods, and overcome the shortcomings of the existing neural network model such as complex structure and error accumulation, thus achieving a more concise and accurate evaluation of the integrity of CFRP after lightning strike. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 4 is a flow chart of a method for analyzing the integrity of CFRP after a lightning strike according to an embodiment of the present invention.

[0022] Figure 2 This is a comparison diagram of infrared spectra of CFRP before and after lightning strike in the test example of the present invention.

[0023] Figure 3 This is a comparison chart of the mechanical properties of CFRP before and after lightning strike in the test example of the present invention.

[0024] Figure 4 This is a diagram of the integrity evaluation of CFRP after lightning strike in the test example of the present invention.

[0025] Figure 5 It is a neural network design diagram of the test example of the present invention. DETAILED DESCRIPTION

[0026] In order to make the technical means, creative features, objectives and effects of the present invention easier to understand, the following embodiments and accompanying drawings specifically illustrate a CFRP post-lightning strike integrity analysis method and system of the present invention.

[0027] Example Figure 1 4 is a flow chart of a method for analyzing the integrity of CFRP after a lightning strike according to an embodiment of the present invention.

[0028] like Figure 1 As shown, this embodiment provides a method for analyzing the integrity of CFRP after a lightning strike, comprising the following steps: S10, determining the key chemical structure in CFRP, includes the following sub-steps S11 to S13: S11, high temperature exposure test: The CFRP was placed in a muffle furnace and ablated at a constant temperature of 100°C to 300°C for 15 minutes in an air atmosphere. The surface and internal temperature distributions of the samples were similar, and the degree of ablation damage was similar.

[0029] S12, Comparison of infrared spectrum changes of CFRP at room temperature and after high temperature exposure test.

[0030] S13, obtain the chemical bond evolution rules of CFRP at different temperatures, and determine that the key chemical structures are methylene, methyl, carbonyl and benzene ring skeleton.

[0031] S20, obtaining target input data, includes the following sub-steps S21-S22: S21, a portable ATR-FTIR spectrometer was used to scan the lightning-struck area of ​​CFRP to obtain its infrared spectrum historical data.

[0032] S22, after baseline adjustment and smoothing of the infrared spectrum historical data, the characteristic peaks of the key chemical structures are selected and the intensities are calculated as the target input data.

[0033] Among them, the characteristic peak intensity is determined by the absorbance value in the infrared spectrum historical data. A Calculate the difference from the baseline.

[0034] S30, obtaining historical data of mechanical properties of CFRP after lightning strike through a three-point bending test as target output data.

[0035] Among them, the historical data of mechanical properties are the data of bending strength.

[0036] S40, using a BP neural network to train a CFRP post-lightning strike integrity prediction model, specifically including the following sub-steps S41 to S43: S41 , using a BP neural network to establish a mapping relationship between the target input data obtained in step S22 and the target output data obtained in step S30 , thereby training a CFRP post-lightning strike integrity prediction model.

[0037] S42, setting constraints for the constructed prediction models under different training conditions, wherein the constraint settings include setting a maximum number of iterations and setting a minimum training target convergence error.

[0038] S43, evaluate the prediction models under different training conditions, and reflect the accuracy of their prediction results by calculating the relative error. If the relative error does not meet the evaluation requirements, return to step S41 and improve the prediction accuracy of the prediction model by expanding the training data set of target input data and target output data and optimizing the structure of the artificial neural network until the relative error meets the evaluation requirements.

[0039] Among them, optimizing the structure of artificial neural networks includes adjusting the number of layers, adjusting the number of hidden layer neurons, and optimizing activation functions.

[0040] S50, integrity analysis, includes the following sub-steps S51 to S53: S51, after using the ATR-FTIR spectrometer to scan the CFRP lightning-struck area that needs to be actually analyzed to obtain its infrared spectrum historical data, the characteristic peaks of the key chemical structure are selected and the intensity is calculated and then input into the prediction model, and the corresponding mechanical property data of the CFRP is output.

[0041] S52, quantification of concentrations of key chemical structures c , specifically including the following sub-steps S521 to S523: S521, determine the thickness of the CFRP that can be penetrated by infrared light / absorption layer thickness using a portable ATR-FTIR spectrometer: (Formula 1) In the above formula, θ is the angle of incidence, n s and n c are the refractive indices of CFRP and ATR crystal, λ represents the wavelength of the incident light. d p and λ ,θ, n s and n c About, when λ When and θ are fixed, if n s is a fixed value, then d p It is also a fixed value.

[0042] S522, absorbance formula is: A = Kd p c (Equation 2) In the above formula, A is the absorbance value, K is the molar absorption coefficient, KDepends on the material of CFRP and the wavelength of incident light λ , c is the concentration of key chemical structures in CFRP after lightning strike.

[0043] S523, combining Equations 1 and 2, to determine the concentration of key chemical structures in the CFRP after a lightning strike c : .

[0044] S53, the mechanical properties data output in step S51 and the concentration of the key chemical structure output in step S52 are combined. c Together they serve as a CFRP integrity analysis report after lightning strike.

[0045] This embodiment also provides a CFRP post-lightning strike integrity analysis system, which uses the CFRP post-lightning strike integrity analysis method of this embodiment, including a data input module, a model building and training module, a prediction analysis module, and an output module.

[0046] The data input module is used for a user to input the target input data obtained in step S22 and the target output data obtained in step S30 .

[0047] The model building and training module is based on an artificial neural network and is connected to the data input module, and is used to establish a mapping relationship between the target input data and the target output data according to the method of step S40, thereby training a CFRP post-lightning strike integrity prediction model.

[0048] The prediction analysis module is connected to the model building and training module, and is used to output the mechanical properties data of CFRP and the concentration of the key chemical structure after the user inputs the intensity of the characteristic peak of the key chemical structure of the CFRP lightning strike area that needs to be actually analyzed according to the method of step S51 and step S52. c .

[0049] The output module is used to output and display the mechanical performance data and concentration of key chemical structures output by the prediction analysis module according to the method of step S53. c And generate a completeness evaluation report.

[0050] Test Case This test example uses the CFRP post-lightning strike integrity analysis system in the embodiment and performs the test according to the CFRP post-lightning strike integrity analysis method in the embodiment.

[0051] The CFRP post-lightning strike integrity analysis method in this test case is similar to that in the embodiment and will not be repeated here. The specific data / methods selected in this test case are as follows: In step S21, the scanning position of the portable ATR-FTIR spectrometer is selected using an array method, expanding outward from the center of the lightning injection point of the CFRP, and collecting infrared spectrum data every 1.5 cm. All collected infrared spectrum data are used as infrared spectrum historical data.

[0052] In step S22: (1) the number of smoothing points is set to 15; (2) the wave numbers of the characteristic peaks of the selected key chemical structures are: 2924 cm corresponding to the symmetric stretching vibration of -CH3 and -CH2- -1 and 2870 cm corresponding to the asymmetric stretching vibration -1 , 1663 cm corresponding to carbonyl stretching vibration -1 , 1590 cm corresponding to the vibration of the benzene ring skeleton -1 and 1512 cm -1 , specifically Figure 2 shown.

[0053] In step S30, a three-point bending test specimen is cut from the CFRP every 1.5 cm, and the bending strength of the CFRP is obtained as its target output data through the three-point bending test. Figure 3 As shown. Among them, Figure 3 In the study, the mechanical properties of CFRP decreased significantly after lightning strike compared with those before lightning strike.

[0054] In step S41, the number of training set data sets is 40.

[0055] In step S42, the maximum number of iterations of the network is set to 500, and the minimum training target convergence error is set to 10 -5 .

[0056] Figure 4 This is a diagram of the integrity evaluation of CFRP after lightning strike in the test example of the present invention.

[0057] like Figure 4 As shown in Figure 3, the relative error of the residual strength predicted by the artificial neural network in this test case is less than 5%.

[0058] Figure 5 It is a neural network design diagram of the test example of the present invention.

[0059] like Figure 5 As shown in Figure 2, in this test case, the number of neural network layers and the number of hidden layer neurons after optimization are set to 3 and 12 respectively.

[0060] Those skilled in the art will appreciate that the present invention is not limited to the foregoing embodiments. The foregoing embodiments and descriptions are merely illustrative of the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for analyzing the integrity of CFRP after lightning strike, characterized in that: The following steps are involved: S10, by conducting high temperature exposure tests on CFRP at different temperatures, comparing the changes in its infrared spectrum to determine the evolution of chemical bonds in CFRP at different temperatures and to determine the key chemical structure; S20, after scanning the lightning-struck area of ​​the CFRP using an ATR-FTIR spectrometer to obtain historical infrared spectrum data thereof, select characteristic peaks of the key chemical structure therein and calculate the intensity thereof as target input data; S30, obtaining historical data of mechanical properties of CFRP after lightning strike through mechanical properties testing, and using the data as target output data; S40, establishing a mapping relationship between the target input data and the target output data using an artificial neural network, thereby training a CFRP post-lightning strike integrity prediction model; S50, after using the ATR-FTIR spectrometer to scan the CFRP lightning-struck area that needs to be actually analyzed to obtain its infrared spectrum historical data, the characteristic peaks of the key chemical structure are selected and the intensity is calculated and input into the prediction model, and the corresponding mechanical property data of the CFRP is output to realize the integrity analysis of the CFRP after the lightning strike.

2. The CFRP post-lightning strike integrity analysis method according to claim 1, characterized in that: in, In step S10, the CFRP is placed in a muffle furnace and ablated at a constant temperature of 100°C to 300°C in an air atmosphere. The infrared spectrum changes of the CFRP at different temperatures are compared to obtain the chemical bond evolution law of the CFRP at different temperatures to determine the key chemical structure.

3. The CFRP post-lightning strike integrity analysis method according to claim 1, characterized in that: in, In steps S10 to S20, the key chemical structure includes the CH stretching vibration of the methyl and methylene groups, the stretching vibration of the carbonyl group, and the skeletal vibration of the benzene ring.

4. The CFRP post-lightning strike integrity analysis method according to claim 1, characterized in that: in, In step S20, baseline adjustment and smoothing are performed on the infrared spectrum historical data.

5. The CFRP post-lightning strike integrity analysis method according to claim 1, characterized in that: in, In step S30, the mechanical properties test includes a three-point bending test. The mechanical property historical data includes bending strength data.

6. The CFRP post-lightning strike integrity analysis method according to claim 1, characterized in that: in, In step S40, constraints are set for the prediction models constructed under different training conditions, and the constraints include setting the maximum number of iterations and the minimum training target convergence error. The artificial neural network is a BP neural network.

7. The CFRP post-lightning strike integrity analysis method according to claim 1, characterized in that: in, In step S40, the prediction models under different training conditions are evaluated, and the accuracy of the prediction results is reflected by calculating the relative errors. If the relative error does not meet the evaluation requirements, the prediction accuracy of the prediction model is improved by expanding the training data sets of the target input data and the target output data and optimizing the structure of the artificial neural network until the relative error meets the evaluation requirements.

8. The CFRP post-lightning strike integrity analysis method according to claim 7, characterized in that: in, Optimizing the structure of the artificial neural network includes adjusting the number of layers, adjusting the number of neurons in the hidden layer, and optimizing the activation function.

9. The method for analyzing the integrity of CFRP after a lightning strike according to any one of claims 1 to 8, characterized in that: in, In step S50, the CFRP post-lightning integrity analysis also includes quantifying the concentration of the key chemical structure c The relationship between the characteristic peak intensity of the key chemical structure, The characteristic peak intensity is determined by the absorbance value in the infrared spectrum historical data. A Calculate the difference from the baseline. The concentration of the key chemical structure c : ， In the above formula, θ is the angle of incidence, n s and n c are the refractive indices of CFRP and ATR crystal, K is the molar absorption coefficient, K Depends on the material of CFRP and the wavelength of incident light λ .

10. A CFRP post-lightning strike integrity analysis system, characterized in that: The method for analyzing the integrity of CFRP after a lightning strike according to any one of claims 1 to 9 is used, comprising: A data input module, configured for a user to input the target input data and the target output data; a model building and training module, based on an artificial neural network and connected to the data input module, for establishing a mapping relationship between the target input data and the target output data, thereby training a CFRP post-lightning strike integrity prediction model; a prediction analysis module connected to the model building and training module, configured to receive input from a user of the intensity of characteristic peaks of the key chemical structure in historical infrared spectrum data of a lightning-struck region of a CFRP to be actually analyzed, and then output mechanical property data of the CFRP corresponding to the prediction model trained by the model building and training module; and The output module is connected to the prediction and analysis module and is used to output and display the mechanical performance data and generate an integrity evaluation report.

Images