基于神经网络的平面五杆可调机构的动力学解析模型

Dynamics models of planar five-bar adjustable mechanisms based on neural network

应用Kane动力学方法对平面可调五杆机构进行了动力学分析,建立了平面五杆机构的动力学解析模型。并研究了构件几何参数和物理特性对广义驱动力矩的影响规律。在此基础上,应用BP神经网络对可调五杆机构逆动力学模型进行了辨识,将整个动力学模型系统分为3个子系统,分别采用BP神经网络系统对广义质量矩阵、向心力哥氏力矩阵和重力矩阵进行学习,求解得到可调平面五杆机构的逆动力学模型问题。

The analytical kinematics and dynamics models of planar five-bar mechanism are established in this paper. The influence of corresponding parameters on the driving torques is investigated. Then the dynamic model system of planar five-bar adjustable mechanism is divided into three subsystems. The inverse dynamic model is investigated based on BP neural network system that is trained and the solutions to inverse dynamic model problem can be obtained. The method presented in this thesis can avoid tedious symbol derivation and reduce greatly computing work as compared with analytical model approach.