Implementation of Finite Element Neural Network Computing Method Based on FPGA PDF

Keywords

Finite element; Neural network computing; Floating -point operations; FPGA; Hardware description language.

Abstract

Based on the research of neural network (NN) computing method of finite element equations for the structure of the discrete model, a complete implementation of the various parts of the computing method is presented. The computation module was described with hardware description language (Verilog HDL) at the RTL level, and functional simulation was executed by Quartus II 11.0 and ModelSim 6.5. Simulation results show that the implementation of neural network computing method based on field programmable gate array (FPGA) can satisfy the requirement of the parallel computing, which could further improve the speed of structural finite element computation, and ensure the speed and accuracy of the computation.