Study On Damage Mechanism And Performance Prediction Of Fiber Composites Under Compression/ Shear After Impact

Composite structures are widely used in aircraft load-bearing components.However,the brittleness,low strength,and fracture toughness of composite in the thickness direction make it sensitive to impact damage.Therefore,it is very important to accurately evaluate and predict the damage forms of composite structures under low velocity impact(LVI)and the influence of LVI damage on the mechanical properties of composite structures.In this study,the damage modes of composite laminates under LVI and multi-zone impact of stiffened panels,as well as the progressive damage evolution process and failure mechanism under compression/shear after impact(CAI/SAI)were deeply studied by the combination of experimental mechanics and finite element numerical simulation.The equivalent impact damage model and artificial neural network structure model were established to predict the LVI response of composite laminates,the CAI buckling mode and load-bearing performance of stiffened panels.The main research contents of the dissertation include the following aspects:(1)The LVI and CAI responses of composite laminates under different impact energies were investigated by experiments and FE simulation.The damage states caused by different impact energies were explored by LVI tests,and the damage morphologies of the composite laminates were characterized using the ultrasonic phased array C-scan technique to quantify the damage degree.CAI tests were conducted on composite laminates with different impact damage,and the full-field displacement and strain distribution during entire experiments were analyzed via 3D-DIC technique.To reveal the complex damage and failure mechanism of composite laminates,a FE simulation method was explored,which introduced interlaminar delamination damage,intralaminar matrix damage,and intralaminar fiber damage into a 3D damage model.The maximum strain criterion was implemented to identify the initiation of the compressive and tensile fiber damage in the composite laminates.The failure criterion proposed by Puck was applied to the initiation and evolution of matrix damage.A cohesive element with a traction separation constitutive relation was incorporated into the FE model to capture the delamination of composite laminates under LVI and CAI loading conditions.A quadratic stress criterion was applied to determine the initiation of interface delamination damage under complex loadings.The power law criterion was adopted to determine the damage evolution rules of delamination.The combined results were used to analyze the impact damage modes and delamination morphologies of the composite laminates under different impact energies and their effects on the damage development and failure modes during compression.The overall mechanical response and damage distribution characteristics of the established finite element model were verified by experimental results.(2)The effects of multi-zone impact damage on the shear properties and failure modes of composite T-stiffened panels were studied experimentally and numerically.The morphology and size of LVI damage in each region were observed by ultrasonic phased array C-scan technology.The full-field displacement distribution and the full-field buckling mode evolution of stiffened panels were monitored by 3D-DIC technology and fringe projection profile(FPP)measurement system during the SAI test.A finite element model considering multi-zone impact and the coexistence of multi damage modes is established to reveal the interaction of multi-zone impact damage,complex damage evolution and failure mechanism of composite stiffened panels.Meanwhile,the progressive damage and shear nonlinear analysis were introduced into the FE model,and the impact damage modes of composite stiffened panels under multi-zone impact and its influence on the SAI damage propagation and failure modes were discussed in comparison with the experimental results.(3)Based on the established three-dimensional finite element progressive damage model,an artificial neural network method for efficiently and accurately predicting the LVI damage degree of composite laminates was proposed.The important factors that have a great influence on the LVI damage of composite laminates were fully considered.Python was used for secondary development to establish a finite element model set with impact/material factors(impact energy,impact position,impact hammer size,composite laminate thickness,ply sequence,etc.)as input and impact damage prediction(impact damage area,pit depth,and impact penetration degree)as output.Combined with enough LVI data,the artificial neural network training set and test set were constructed.A variety of neural network structures were selected for training and testing,so that they can predict the impact damage area,pit depth,and impact penetration degree under different input variable configurations within an acceptable accuracy range.The application of differential evolution algorithm was discussed.According to the average error of neural network model,the artificial neural network model with the best training result was selected to carry out the investigation of performance evaluation and generalization ability.Compared with the finite element results,the impact damage area and pit depth predicted by BP neural network are linearly distributed with the test set data,and the correlation coefficients are R~2=0.98639 and R~2=0.99739,respectively.The average errors are 1.48%and 1.03%,respectively.The prediction accuracy of BP neural network for predicting the impact penetration of laminates is up to 99.41%.The results indicated that the two neural network models established have good accuracy and high efficiency.(4)Based on the proposed artificial neural network method for predicting impact damage response,a phenomenological equivalent impact damage finite element model was proposed and established combined with the results of finite element simulation and ultrasonic C-scan.According to different degrees of damage in the impact area,an image recognition method was adopted to divide the entire composite laminate and stiffened panel with impact damage into regions.And the material stiffness and strength parameters were weakened in various levels.The pit deformation caused by impact was equivalent to the element node offset in the corresponding area of the model.The interface delamination and debonding induced by impact was equivalent to interface holes without any elements.The equivalent impact damage model was employed to evaluate the CAI buckling behavior and residual compressive strength of laminates and stiffened panels.By comparing the test results,it was pointed out that the difference between the ultimate load predicted by the equivalent impact damage model of laminates differs from the experimental and numerical simulation results less than2%.The difference between the buckling load obtained by the equivalent impact damage model of stiffened plates and the experimental results is only 0.32%,and the difference between the ultimate load is 2.32%.The accuracy and effectiveness of the established equivalent impact damage model are verified in predicting the compression buckling mode and strength of composite laminates under LVI and stiffened panels under multi-zone impact.Combined with the equivalent impact damage model,an artificial neural network method for predicting the CAI buckling mode and load-bearing performance of composite stiffened panels was proposed.An under-complete Auto-Encoder and a sparse Auto-Encoder were selected to extract the features and reduce the dimension of the input parameter set in the process of neural network training,respectively.The extracted dimensionality reduction feature parameters were used as the initial input parameters of BP neural network to deal with the regression problem(predicting buckling load and ultimate load)and the classification problem(predicting buckling mode)for training,verification,and testing.The CAI buckling load and ultimate load predicted by BP neural network are linearly distributed with the test set data,and the correlation coefficients are R~2=0.99841 and R~2=0.99908,respectively.The average errors are 1.23%and 1.01%,respectively.The prediction accuracy of BP neural network for predicting CAI buckling mode is up to 99.63%.The results indicated that the constructed BP neural network model can accurately predict the CAI buckling load and ultimate load of the stiffened panels,and meet the prediction accuracy of the CAI buckling mode of the stiffened panels.