Research On Responses Of Fiber Reinforced Laminates Subjected To Ice Projectile Impact And Inverse Methods Of Material Parameters

Aircraft,high-speed rail,cars and other vehicles often encounter soft material impact represented by birds and hail ice.The structure with high speed operation will face deformation and damage under impact load.With the continuous application of composite structure in various industrial fields,its resistance to soft material projectile impact has attracted increasingly more attention.However,the extensive,comprehensive and systematic research in this field is restricted by strict experimental conditions,complex data acquisition methods,difficult material models and difficult means of parameter determination.Therefore,this paper started with hail impact on carbon fiber reinforced composite(CFRP)laminates to explore the response and damage characteristics of new material structures under soft material impact load.In order to obtain the dynamic mechanical behavior of ice materials at high strain rates,the relationship between the compression and tensile strength of ice materials and the strain rate was obtained using the split Hopkinson bar and the dynamic Brazil disk experiment.According to the high divergence problem on experimental results,the completion of the stress equilibrium hypothesis has been thoroughly analyzed,and the necessity of three-wave method in processing data was discussed.Combined with the spacing calculation of compression specimens with the measuring bars,the influence of the roughness of the specimen surface induced by the moisture condensation in the air on the loading process was analyzed,and consequently the repeatability of the experiment was determined.The characteristics of crack initiation and crack propagation in tensile specimens were analyzed using high-speed camera and digital image correlation(DIC)techniques.The validity of the two experiments is verified.The influence of temperature effect on experimental results is also summarized.The impact force characteristics of ice projectiles were obtained by hollow aluminum tube measurement device,and the convergence of experimental results was analyzed in combination with the energy transfer process.The finite element model of ice projectile was established based on the results of compression and tensile strength of ice material obtained above.In order to achieve accurate impact simulation,the unknown material parameters were determined indirectly by inversion method.A long-short term neural network(LSTM)was introduced to establish an inversion model considering the continuous and intermittent impact loading of ice projectiles.The global convergence level of the model is determined by generalization ability verification.The effectiveness of inversion parameters and inversion method was verified by comparing the impact force curve and deformation process of projectile body in experimental and simulation results.The influence of tensile failure stress parameters related to the fragmentation mechanism of the projectile body on the deformation process was analyzed through parameter study.After comparing the training efficiency of unidirectional LSTM and bidirectional LSTM neural networks,the advantages of the inversion method adopted in this paper was revealed.After the impact behavior of ice projectile was determined,the impact test of CFRP laminates was carried out using a single-stage gas cannon and the global deformation data of the target plates were obtained by 3D-DIC method.The damage condition inside the target plate was obtained by observing the cross section and CSAM scanning,and the delaminated damage mode was determined.Based on the crack propagation mechanism and DIC global displacement data,an appropriate hypothesis was proposed and a damage prediction model based on the accumulated strain energy criterion of target plate was established.Based on grid sensitivity analysis,the influence of the facet size and calculation step size on the results of DIC measurement is discussed.The influence of strain rate effect on calculation results is discussed through parametric study.The damage strain energy threshold was determined based on the damage area of the target plate obtained from experimental scanning,and it was found that the threshold value was slightly higher than the critical energy release rate of type II crack in existing literatures,which verified the effectiveness of the method and revealed the application potential of DIC results in damage prediction methods.A finite element model was established to simulate the ice impact events on CFRP laminates.In order to solve the underfitting problem in inversely determining the material parameters of the cohesive model and CFRP layers,a deep learning method was introduced.The time cost of sample production was optimized combined with parameter study method and grid sensitivity analysis,and data augmentation method was used to expand sampling space.The influence of the size of convolutional neural network and the convergence efficiency on the training effect was discussed through hyperparameter analysis.By comparing the convergence results in the training set and the validation set,the generalization ability of the inversion network was verified,indicating that the under-fitting problem was effectively solved.After the simulation was carried out using the material parameters obtained from the inversion,the central deflection curve of the experimental target plate and the damage condition of the scanning result was compared,and the validity of the parameters and the reliability of the inversion method were verified.