基于神经网络的步进电机细分电流最佳设计

提出用神经网络来确定步进电机最佳细分电流的思想.介绍了最佳细分电流数据训练集的简易方法,设计了多层前向神经网络的学习和仿真程序并进行了仿真实验.在学习中使用了Bayes正则化算法,使得网络的推广能力得到提高,同时为了避免多层前向神经网络陷入局部极小点,使用了权值调整技术.文中首次给出了用神经网络获得的64细分、8bit存储时的最佳细分电流数据表,对同行研究有直接参考价值.

A Neural Network Based Method for Determining the Best Subdivision Current Data of Step Motor

The idea and method of how to use neural networks to determine the best subdivision current data are carried out in this paper and an easy method for preparation of training set for subdivision current data is introduced.The emulating MLFNN(multi-layer forward neural networks)and the training program are designed.By using Bayes regularization method in learning,the networks have a better generalization ability.The weight adjust technique is adopted for local minimum problem.The 64 subdivision current data form with an 8bit memory is given which could be of high value to related readers.