几种改进BP算法在电机故障诊断中的分析比较

近年来,人工神经网络以其独特的容错、联想、推测、记忆、自适应和处理复杂的多模式等优点,在许多学科中掀起了研究热潮,同样在故障诊断领域,其发展前景也十分乐观,它作为一种自适应的模式识别技术,并不需要预先给出有关模式的经验知识和判断函数,对于特定问题建立的神经网络故障诊断系统,可以从输入数据（故障症状）直接推出输出数据（故障原因）,从而实现故障的检测与诊断。但是传统的BP算法存在一些缺陷。所以采用一些改进的BP算法,把它们用于电机的故障诊断,通过实验对这些改进算法的分析比较,得出其中较合适的算法。这样不仅能发挥神经网络的泛化映射能力而且诊断速度也有提高,有较强的学习能力。

Several Improved BP Algorithm for Motor Fault Diagnosis

In recent years,artificial neural networks with its unique advantages of fault-tolerant,association,speculation,memory,adaptive,and dealing with complex multi-mode,etc. ,causes the study upsurge in many disciplines and also in fault diagnosis field and its development prospect is very optimistic. This network is an adaptive pattern recognition technology,so it is not necessary to give the knowledge and experience of model in advance. For the neural network fault diagnosis system that established by specific questions, it can get the output data directly from the input data （ failure symptoms） in order to achieve the fault detection and diagnosis results. However,the traditional BP algorithm has some shortcomings. Therefore,this paper uses some improved BP algorithm for the motor fault diagnosis,and through the analysis and comparison with these improved algorithms,it gets some appropriate algorithms. It will not only be able to play mapping neural network generalization ability and the speed of diagnosis is also improved,with.