论文部分内容阅读
从神经网络的基本方程出发,在理论、数值计算方面对弹性力学神经计算的稳定时间和计算精度进行了研究。理论分析表明:偏置电容C、信号网络神经元的放大倍数β影响网络的稳定时间;偏置电阻R、约束网络神经元的惩罚系数α影响求解精度,但在工程上没有必要对R,C,β,α等进行严格的限制。数值模拟结果也证实了理论分析的结论。
Based on the basic equations of neural network, the stability time and calculation precision of elastic mechanics neural computation are studied theoretically and numerically. The theoretical analysis shows that the bias factor C and the amplification β of the signal network neurons affect the stability time of the network. The bias resistance R and the penalty coefficient α of the constrained network neurons affect the accuracy of the solution. However, it is not necessary in engineering to evaluate R, C , Β, α and other strict restrictions. Numerical simulation results also confirm the conclusions of theoretical analysis.