固体力学有限元计算问题可以采用神经网络进行实时求解，神经网络可用改进的Ｈｏｐｆｉｅｌｄ网络．求解之前必须确定信号网络的联接权矩阵（相应结构刚度矩阵）．分析了单元刚度矩阵的组成，提出用ＢＰ网络来实现单刚矩阵的实时计算方法．分别用传统有限元法和ＢＰ网络法求解了一三角形单元的单刚矩阵元素，结果表明ＢＰ网络法计算具有实时性，精度满足工程要求． The problem of solid mechanics finite element calculation can be solved by neural network in real time, and the improved Hopfield network can be used in neural network. Before solving, you must determine the connection matrix of the signal network (the corresponding structure stiffness matrix). The composition of element stiffness matrix is ​​analyzed and a BP network is proposed to realize the real-time calculation method of single-element matrix. By using the traditional finite element method and the BP network method respectively, the simple matrix elements of a triangular unit are solved. The results show that the BP network method has the real-time calculation and the accuracy meets the engineering requirements.