Abstract: There is an unsolved contradiction between the computation and memory of the Von Neumann computer.So many problems of engineering mechanics haven’t breakthrough because of such reason as large computation scale，etc.Therefore，new theory and method of computational mechanics need to be developed to solve above problems.Solving method of finite element and neural network design are studied.The neural network of finite element solving is obtained on the basis of Hopfield neural network that is reformed and avoids using of high-gain transfer function in the circuit.Then the no error solving of finite element neural net computation is realized in theory.

摘要： 计算机本身固有的计算与存储之间是一对很难解决的矛盾，许多工程力学问题因计算规模大等原因还没有突破性进展，因此，需要发展新的理论与计算方法。通过对有限元求解方法和Hopfield神经网络的深入研究，在对Hopfield神经网络适当改造后，得到了有限元的神经网络计算方法，在电路实现中避免了采用高增益传递函数的假设，进而在理论上实现了有限元神经网络计算的无误差求解。