Data-driven failure probability criterion of composite material based on Micro-mechanics inspired failure criterion
Topic(s) : Material and Structural Behavior - Simulation & Testing
Co-authors :
Xiaodong WANG (CHINA), Zhidong GUANAbstract :
Composites materials are one of the most important structural materials in the application of aerospace, due to their excellent specific performance and designability. Compared with metal materials, the mechanical properties composites exabit obvious dispersion and anisotropic, which bring significant challenge on the accurate prediction of the failure behavior of composites structures, hindering the further application of composites in the aerospace engineering. The multi-constituents (fiber, matrix and fiber/matrix interface) and multi-scale (micro-scale, meso-scale, macro-scale) induce the multi-physical failure modes (fiber tension failure, fiber compression failure, inter-fiber tension failure, inter-fiber compression failure and etc.), causing complex failure envelope of composite under different stress states.
With the development of the failure criteria, the physical-based failure criteria show significant advantages. In this paper, the failure mechanics of the composites under biaxial stress state are studied by the Representative Volume Element (RVE). The fiber, matrix and fiber/matrix interface are considered in the model, with random fiber distribution in transverse model and fiber initial misalignment in longitudinal compression model. The anisotropic elastic-plastic behavior of matrix is modelled by the Drucker-Prager plastic model and Ductile criterion, the interface failure is simulated by the cohesive element considering friction effect. The simulate results indicate that the transverse failure is dominated by the local inter-fiber stress state and the inter-fiber properties (properties of matrix and interface), the longitudinal failure is dominated by two types of failure mode, fiber bending failure and kink-band failure.
Based on the micro-scale mechanics, a micro-mechanics based failure criterion is proposed. The local inter-fiber stress state is calculated by the Eshelby inclusion theory, and the effect of fiber initial misalignment on the stress state is derived based on the stress analysis of cosine shape fiber path under longitudinal compression stress. The relationship of micro-scale stress and macro-scale stress is adopted to determine the local micro-scale stress state of composites. Then the failure of composites is judged by the failure criterion of constituents. The proposed micro-mechanics based failure criterion is accurate on the prediction of composites under complex stress state. To establish the data-driven failure probability criterion, the failure strength of composites under complex stress state considering performance dispersion is generated as the data set. Then the construct neural network model is trained by the train set and validation set, then the neural network model is test by the test set. The trained data-driven failure probability criterion can give the failure probability of composite under certain stress-state, provided the method for failure probability analysis of the composite structures.
With the development of the failure criteria, the physical-based failure criteria show significant advantages. In this paper, the failure mechanics of the composites under biaxial stress state are studied by the Representative Volume Element (RVE). The fiber, matrix and fiber/matrix interface are considered in the model, with random fiber distribution in transverse model and fiber initial misalignment in longitudinal compression model. The anisotropic elastic-plastic behavior of matrix is modelled by the Drucker-Prager plastic model and Ductile criterion, the interface failure is simulated by the cohesive element considering friction effect. The simulate results indicate that the transverse failure is dominated by the local inter-fiber stress state and the inter-fiber properties (properties of matrix and interface), the longitudinal failure is dominated by two types of failure mode, fiber bending failure and kink-band failure.
Based on the micro-scale mechanics, a micro-mechanics based failure criterion is proposed. The local inter-fiber stress state is calculated by the Eshelby inclusion theory, and the effect of fiber initial misalignment on the stress state is derived based on the stress analysis of cosine shape fiber path under longitudinal compression stress. The relationship of micro-scale stress and macro-scale stress is adopted to determine the local micro-scale stress state of composites. Then the failure of composites is judged by the failure criterion of constituents. The proposed micro-mechanics based failure criterion is accurate on the prediction of composites under complex stress state. To establish the data-driven failure probability criterion, the failure strength of composites under complex stress state considering performance dispersion is generated as the data set. Then the construct neural network model is trained by the train set and validation set, then the neural network model is test by the test set. The trained data-driven failure probability criterion can give the failure probability of composite under certain stress-state, provided the method for failure probability analysis of the composite structures.