基于改进Hu不变矩和SVM的转子故障诊断方法

Abstract

在转子发生故障时，对转子进行准确地故障诊断一直是工程领域研究的重点。针对这一问题，本文提出了一种基于改进Hu不变矩和支持向量机(SVM)相结合的转子多故障诊断方法。该方法首先通过仿真合成4种常见故障的轴心轨迹图，并利用改进Hu不变矩分别对其进行特征提取，然后将得到的Hu不变矩值作为特征向量输入有向无环图SVM进行识别分类，最终实现转子的多故障诊断。经实验表明，该方法能够准确地完成转子的多故障诊断，同时通过与原始Hu不变矩比较，证明改进Hu不变矩在识别准确率上明显优于原始Hu不变矩。 The accurate diagnosis of fault conditions of steam rotor when faults occur in rotor has been the re-search focus in the field of engineering. To solve this problem, this study proposes a multiple fault diagnosis method for steam rotor based on the improved Hu invariant moments and SVM. First, 4 kinds of common faults’ Shaft orbits are synthesized by simulation in this method, and extract fea-tures by the improved Hu invariant moments, then, the Hu invariant moment values will input to the directed acyclic graph SVM as feature vectors to recognize and classify, and ultimately achieve the multiple fault diagnosis of rotor. The experimental results show that this method can accurately complete rotor multiple fault diagnosis. Meanwhile, to compared with the original Hu invariant moments, it is proved that the improved Hu invariant moments are better than the original Hu in-variant moments in the recognition accuracy.