Integration modified wavelet neural networks for solving thin plate bending problem

In this paper, a modified wavelet neural network (MWNN), which is trained by chaos particle swarm optimization and whose activation function is fourth-order scaling function of spline wavelet, is first proposed for solving thin plate bending problem. The highest derivatives of variables in the governing equations are represented by the outputs of MWNN. The variables and the other derivatives are obtained by integrated outputs of MWNN. During the integration process, multiple boundary conditions can be implemented straightforward. It has been verified that the MWNN method can successfully solve various thin plate bending problems and it is convergent based on different distributions of scattered points.