多层感知器神经网络在机械故障诊断中的应用

针对ＢＰ网络用于复杂机械故障诊断时学习收敛慢，易陷于局部极小点等不足，提出了改进－－较大误差相关修正法，并对各系数进行了研究，结果表明，改进算法拓宽了各系数的取值范围、使网络性能更加平稳，且缩短了训练时间，适用于解决多输出节点的复杂故障诊断问题。

Application of multi layer perceptron neural network in mechanical fault diagnosis

A modified method for back propagation networks, named Higher Error level Correction, was proposed in accordance with the classical back propagation networks' shortages, such as slow convergence and usually getting trapped in local minima, etc. Coefficients are studied as well. The conclusion indicates that the modified method can improve the networks' performance, for example, it makes coefficients' range wider, makes the networks' behavior stabler and shorten the training time compared with the classical back propagation algorithm. The method is fit to solve complex diagnosis problems in which there are many nodes on the network's output layer.