基于椭球单元网络的旋转机械多故障同时性诊断

Multiple Fault Simultaneous Diagnosis for Rotating Machine Based on Ellipsoidal Unit Neural Networks

阐述了椭球单元（ＥｌｉｐｓｏｉｄａｌＵｎｉｔ）网络的原理及其结构，研究了网络权重初始化方法和网络的训练算法，借助这种高阶网络泛化的有界性，针对大型旋转机械多故障同时性诊断问题，构造了一种由多个子网络组成的分级诊断网络（ＨＤＡＮＮ）。测试结果表明：用基于椭球单元网络的ＨＤＡＮＮ网络分级诊断策略解决大规模故障诊断问题是合理有效的，且具有较高的诊断精度，可用于旋转机械工况实时监测和诊断场合。

To overcome the limitations of the standard feedforward neural networks,a new sort of high order neural networks (i.e.ellipsoidal unit networks)has been proposed recently,which is very suitable for fault diagnosis applications due to its bounded generalization and extrapolation.In this paper,the theory and structure of such networks are described,a method for initializing hyperellipsoids is proposed,and the training algorithm is given based on standard backpropagation algorithm.Then, based on such networks,a hierarchical diagnosis network(HDANN) is proposed with respect to multiple fault simultaneous diagnosis for rotating machines.HDANN consists of several subnetworks,and aims at dividing a large pattern space into several smaller subspaces,so that the subnetworks can be trained in subspaces respectively,and the whole networks is capable of multiple fault simulatnecus diagnosing .The results show that HDANN based on ellipsoidal unit networks can obtain more accurate and efficient diagnosis results ,and is available for real time condition monitoring and diagnosis of rotating machine.