论文部分内容阅读
由大跨度柔性悬索拖动馈源舱运动来实现其精确定位的新型大射电望远镜(LT)具有变结构、非线性、大滞后、多输入多输出的特点,传统的建模方法已很难建立起其精确的数学模型,这里提出了采用BP神经网络进行辨识建模的思想,考虑到基于标准BP算法的神经网络收敛速度慢、易于陷入局部极小值的不足,提出了基于数值优化的L M(Levenberg-Marquardt)训练算法。应用Matlab6.x中的神经网络工具箱实现了系统的仿真,实验结果表明,采用这种方法可成功地建立舱索系统模型,无论其学习能力还是泛化能力都得到了很好的效果,且其收敛速度大大提高。
The new large radio telescope (LT) with its long-span flexible cable dragging the feed capsule to achieve its precise positioning has the characteristics of variable structure, nonlinearity, large hysteresis and multiple input and multiple output. The traditional modeling methods have been very difficult And set up its precise mathematical model. Here we put forward the idea of using BP neural network to recognize and model. Considering that the convergence rate of neural network based on standard BP algorithm is slow and easy to fall into local minimum, a numerical optimization based LM (Levenberg-Marquardt) training algorithm. The simulation of the system is realized by using the neural network toolbox in Matlab6.x.Experimental results show that this method can successfully establish the model of the navaglass system, and both the learning ability and the generalization ability have achieved good results. Its convergence speed is greatly improved.