Fatigue Analysis Of Two-dimensional Braided Ceramic Matrix Composites

Two-dimensional braided ceramic matrix composites have the advantages of high specific strength,high specific modulus,high temperature resistance,corrosion resistance and so on.Two-dimensional braided ceramic matrix composites subjected to fatigue load during service,and its fatigue analysis is more complex than traditional materials because of its complex microstructure and failure mechanism.Therefore,it is of great significance to develop the fatigue analysis method of two-dimensional braided ceramic matrix composites.In this thesis,the following work is carried out for the fatigue analysis of two-dimensional braided ceramic matrix composites:First,the fatigue behavior of two-dimensional braided ceramic matrix composites at ambient and elevated temperatures is studied.A submodel-based multi-scale normal temperature fatigue life analysis method for two-dimensional braided ceramic composites is proposed.Taking the finite element shape function as the interpolation function,the coupling analysis between the macro-scale finite element model and the meso-scale finite element model is realized.The fatigue failure criterion of ceramic matrix composites based on shear-lag theory realized the coupled analysis of meso-scale and micro-scale.Taking 2D Si C / Si C ceramic matrix composites as the research object,a mesoscopic unit cell finite element sub model is established and its fatigue life is analyzed.The analysis results are in good agreement with the experimental results.The fatigue life analysis of 2D Si C / Si C stiffened plate and the influence of the position of different sub models in the macro element on the calculation results are carried out.The analysis results show that the multi-scale fatigue life analysis method based on sub model proposed in this thesis has good applicability to two-dimensional braided composite structures,and the position of different submodels in the macro element has little influence.Aiming at the fatigue problem of two-dimensional braided ceramic matrix composites under service conditions,the high-temperature fatigue test was carried out.The test was carried out with a large loading cycle of heating-constant temperature fatigue loading-cooling.Based on the test results,the high temperature fatigue life curve(S-N curve)of two-dimensional braided ceramic matrix composites is obtained,and the variation laws of high temperature fatigue residual stiffness and hysteretic curve area are analyzed.The research results show that the residual stiffness decreases approximately linearly in the large cycle before failure,and the variation of hysteretic curve area has obvious periodicity.Secondly,aiming at the problems that it is difficult to obtain the meso parameters of the meso mechanical model and the macro phenomenological model needs a large number of experimental data,a two-dimensional braided ceramic matrix composite fatigue life analysis method based on neural network is proposed.Based on the multi-scale simulation fatigue life data of two-dimensional braided ceramic matrix composites,taking the loading parameters and material parameters as the input and the fatigue life as the output,GRNN network(general regression neural network),Elman and CNN network(revolutionary neural networks)are used for training.The research results show that Elman neural network and CNN neural network can obtain high-precision prediction results under the condition of using only 15 S-N curves.Taking the experimental data in the relevant literature as the data set,using Elman network and CNN neural network training,good prediction results are obtained under the condition that only 5 SN curves are used as the training set.Finally,the prediction method of fatigue residual stiffness based on β-Variational Autoencoder(β-VAE)and neural network ordinary differential equation(ODE)is proposed.The β-VAE is used to extract and isolate the latent features of the underlying fatigue behavior,and the neural network ODE is used to learn the latent dynamics corresponding to the evolution mechanism of fatigue residual stiffness.The prediction of fatigue residual stiffness of 2D C/Si C ceramic matrix composites at room temperature is carried out.A fatigue residual stiffness prediction method based on potential variable interpolation is proposed,and the prediction results are in good agreement with the calculation results of phenomenological model.A fatigue residual stiffness prediction method based on partial data retraining is proposed.Using only 10%of the initial fatigue residual stiffness data,the subsequent residual stiffness prediction results with high accuracy can be obtained.At the same time,the high-temperature fatigue residual stiffness prediction of 2D C/Si C ceramic matrix composites is carried out.The neural network can learn the stiffness degradation law from the high-temperature fatigue residual stiffness data,and realize the high-precision reconstruction of high-temperature fatigue residual stiffness.The retraining method based on partial data can better predict the subsequent residual stiffness only by using the first 30% of the residual stiffness data.