BP网在摩擦学系统建模和预测应用中的研究

本文在分析了摩擦学系统特点以及在BP神经网络的结构的基础上，采用BP神经网络对摩擦学系统进行了建模和预测，通过对结果的分析，给出了这种神经网络在不改变隐层节点数和训练精度的条件下，来提高训练速度和拟合精度的一种新的方法，并指出了要保证BP神经网络近似任意的连续非线性函数，其训练数据应具备的条件。它对BP神经网络在处理大数值的应用中，具有较大的指导意义。

Study of BP Network in Tribology System Modeling and Forecast Application

BP neural network was used to tribology system modeling and forecast in this paper, based on the characteristic of tribology system and structure of BP neural network A new method was presented through analyzing result The method can enhance training rate and imitating precision when latent node number and training error did not change A demand of training data was indicated in order to ensure BP network to imitate arbitrary continuous non linear function The method has biggish direction meaning when BP neural network deal with big data