基于混合遗传算法的磁控形状记忆合金驱动器磁滞模型优化

为了消除或减小磁滞非线性特性对磁控形状记忆合金驱动器定位精度的影响,应用BP神经网络建立了磁控形状记忆合金驱动器磁滞模型。针对BP网络算法存在的不足,以及网络结构、初始连接权值和阈值的选择对BP网络训练的影响很大等问题,提出一种混合遗传算法对神经网络磁滞模型的权值和阈值进行优化。将优化后的参数赋值给BP神经网络重新训练,结果表明,优化后的磁滞模型训练误差绝对值由25nm减小到5nm,有较好的收敛性。

Optimization of MSMA Actuator Hysteresis Model Based on Hybrid GA

In order to improve the positioning precision of MSMA actuator, a BP neural network hysteresis nonlinear model was built. For the shortcomings that BP neural network existed, and the differences of network structure and the choices of initial connection weights and thresholds effected  BP network training precision. For solving these problems, a hybrid algorithm of GA and BP  algorithm was established, the network weights and thresholds were optimized by using the GA, and BP neural network hysteresis nonlinear model was renewally trained by using the optimized parameters. Results show that the optimized neural network hybrid model has better convergence, absolute value of  training error is decreased from 25nm to 5nm.