神经网络用于非线性动力学系统建模研究

文中利用先验知识和神经网络黑箱建模相结合的杂交建模方法对一类滞后非线性系统进行建模研究。由先验信息得到非线性动力系统模型的确定部分,神经网络对系统未确定部分进行建模。文中以系统恢复力网络模型和恢复力的杂交模型为例进行仿真研究,说明杂交模型物理结构明确,网络参数少,运算量小,精度较高。

Research on the Modeling of Nonlinear Dynamical System Using Neural Networks

Hybrid modeling method was used to research the modeling of one kind of hysteresis nonlinear system in this paper ,which combining the prior knowledge and neural networks black-box model. The determinate parts of the nonlinear dynamical system model were based on the prior knowledge, the uncertain parts, nonlinear restore force are modeled by the neural networks. Hybrid model have clearer physical structure, less network parameters, and smaller computation, higher precision by the simulation comparison studies between the restore force NN model and hybrid model of it.