基于改进型BP神经网络的并联机构位置正解研究

摘要

A kind of improved back propagation (BP) neural network method is used to study the forward position kinematics of the 3-RSR parallel mechanism.First,the inverse kinematics solutions of the 3-RSR parallel mechanism were studied based on theoretical approach,and then the improved BP neural network was adopted to train the data obtained from inverse kinematics solution.It was shown that the present method was reasonable by comparing the results obtained from Neural Network training and theoretical result.Furthermore,it could be clearly seen that both of the methods had a very high computational efficiency and avoided the complicated derivation and calculation effectively,but improved BP neural network had higher precision by comparing the improved BP neural network and ordinary BP neural network.Therefore,the improved BP neural network method is more reasonable in the field of engineering which high accuracy is necessary.%采用了一种改进型的BP神经网络方法对3-RSR并联机构的位置正解进行了研究,首先通过理论方法对3-RSR并联机构的位置逆解进行求解,然后采用改进型BP神经网络对通过位置逆解得到的数据进行了网络训练,神经网络训练结果与理论结果对比,验证了该方法的合理性.而且,对比改进后的BP神经网络结果与普通BP神经网络结果,发现两种方法均具有非常高的计算效率,且有效避免了复杂的推导和演算,但改进型BP神经网络方法得到的结果精度更高,因而在高精度的工程领域,采用改进型BP神经网络方法更具合理性.