基于函数链神经网络的磁化曲线拟合方法  

Magnetization Curve Fitting Based on the Function-chain Neural Network

作　　者：吕新明[1] 孙汉旭[1] 

机构地区：[1]北京邮电大学自动化学院,北京100876

出　　处：《机电产品开发与创新》2007年第2期3-5,共3页Development & Innovation of Machinery & Electrical Products

基　　金：国家自然科学基金(50375020)

摘　　要：为了用简捷的方法实现高精度的磁化曲线拟合,并简化拟合公式以便于在电磁场分析计算的使用,提出了函数链神经网络进行磁化曲线拟合的方法。首先给出了函数链神经网络用于磁化曲线拟合的原理,分析了这种算法的特点;然后给出了一个具体的拟合算法实例,探讨了神经网络参数对于学习过程及拟合结果的影响;最后给出了拟合结果,证明了该算法拟合过程方便,拟合函数明晰,同时拟合精度能够满足工程应用的需要。To accurately fit the magnetization curve with a method as simple as possible, and to predigest the fitting expression for application in dectromagnetism field calculations, magnetization curve fitting based on the function-chain neural network is put forward. Firstly, the principle of the function-chain neural network is presented, and the specialty of the algorithm is analyzed. Then, an example of the fitting arithmetic is given, and the infection by the neural network parameter on the fitting result is researched. At hst, the fitting result is given, which proved that the function-chain neural network fitting arithmetic is convenience, the fitting function is clear, and the fitting precision can satisfy the need of engineering application.

关 键 词：函数链神经网络 磁化曲线 曲线拟合 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]