论文部分内容阅读
本文从一般的多自由度离散系统动力学模型出发,建立了结构化神经网络模型。该模型将复杂的多自由度系统非线性特性识别问题分解为若干个单自由度系统非线性特性识别问题,简化了对于复杂系统的分析与求解。文中把模糊自适应BP算法应用于包装件缓冲垫层材料的非线性特性识别问题,在一定程度上提高了网络的训练速度,减少了对于训练参数的人为干预。在此基础上本文提出了引入变异机制的改进的模糊自适应BP算法,它进一步提高了算法的效率,增强了算法的自适应性。针对具有两种典型的包装件缓冲垫层材料的三自由度系统模型的模拟实验,表明了改进算法用于非线性识别问题的有效性。
In this paper, based on the general multi-degree-of-freedom discrete system dynamics model, a structured neural network model is established. The model decomposes the complex multi-degree-of-freedom nonlinear system identification problem into several non-linear characteristic identification problems, simplifies the analysis and solution of complex systems. In this paper, the fuzzy adaptive BP algorithm is applied to the nonlinear characteristic identification of package cushion material, which improves the training speed of the network to a certain extent and reduces the human intervention for the training parameters. On this basis, this paper proposes an improved fuzzy adaptive BP algorithm which introduces mutation mechanism, which further improves the efficiency of the algorithm and enhances the self-adaptability of the algorithm. A simulation experiment on a three-DOF system model with two typical package cushion materials shows the effectiveness of the improved algorithm for nonlinear identification problems.