Artificial Neural Networks In The Elastic-plastic Finite Element Applications

The computation of element stiffness matrix is important and time consuming in the finite element method. In this paper, an approach which may realize the computation of elasto-lastic element stiffness matrix based on the BP net is proposed. Once the BP net is trained successfully, the computation is real梩ime. Through intensive study of primary theory of BP artificial nerual network and characters of triangular membrane element stiffiiess matrix, BP net is determined to finish the task. in addition, the least independent elements of the element stiffness matrix are proposed firstly. Some particular problems are also studied, such as the algorithm of BP net, the values of input and output of the BP net and the samples optimized to the BP net. Based on above work, the elasto-lastic triangular membrane element stiffness matrix is computed by the BP net successfully. A program in MATLAB is developed to implement the function mentioned above. A program simulating the elasto-lastic finite element analysis of the deformation theory of plasticity is also done in MATLAB so that compare the traditional finite element method to the one using BP net.