Numerical Simulation And Hardware Realization Of Finite Element Analysis With Neural Network Dynamical System

The structural optimization computation based on the finite element method. If it is large-scale and complex engineering structure, so then there are more design variables and constraint functions, the optimization computation will consume much time. Accordingly the people use the parallel computation method. The Hopfield neural network is one kind of the parallel optimization computation network. The question of the finite element method can be solved by the modified network.The finite element method of the elastic mechanics, it corresponds to the quadratic programming with equality constraint, further the computation can be translated into the quadratic programming with unconstraint by restricted reduction. In the paper, the question is solved by modified Hopfield neural network. There the energy function of the neural network equals to the objective function of the finite element method. Along with time evolution, the energy function of the network tends to minimum. Therefore calculate least value of finite element energy function transforms as the solution minimum point of network energy function, namely approach stable equilibrium point's process of the network system.In this work, the main productions obtained are following as:1. Have derived the circuit time formula based on finite element nerve computing system of Hopfield network, calculate this time that consuming time for the theory of this computing system.2. Have derived the circuit error estimate formula based on finite element nerve computing system of Hopfield network, namely has drawn the calculation error estimate method to cause because of the errors, such as component, or input, etc...3. Computer simulation analysis of Hopfield neural network computing system simulation circuit retrofitted.4. The hardware of Hopfield neural network computing system simulation circuit retrofitted realizes analyzing.The computer simulation and analog circuit experiment all show that the neural computing method of finite element based on above-mentioned Hopfield network is feasible, effective, and reliable.