故障诊断知识获取的一种神经网络理论方法

A CONNECTIONIST THEORETICAL METHOD OF KNOWLEDGE ACQUISITION FOR FAULT DIAGNOSIS

基于神经网络的竞争学习机制,提出了一种新的基于神经网络专家系统的自动化生产过程监控的知识获取理论方法.这种理论方法在故障诊断的知识获取上是通过竞争学习机制来实现的,与以往人们一般较常采用的BP学习算法相比,具有算法简单、易于实现及无需教师进行监督等特点.利用此方法,经在一个铣削加工过程监控系统上进行仿真研究表明:这种理论方法是非常有效的.

In connectionist expert system, knowledge is presented in the weight matrices of the system. The process of knowledge obtaining is such a process that weight matrices are gradually regulated according to training rules. In the past, BP learning algorithm has been often used as a rule, but it is difficult to realize in practice because of its complexity and the need of being supervised. On the basis of the competitive learning mechanism of neural network, this paper proposes a new theoretical method of automatic prodution monitoring. Compaed with the commonly used BP learning rule, this method possesses advantages of simple algorithm, easy performing and unsupervised learning. Simulation studies for milling machining process show that this theoretical method is very effective.