基于最小二乘支持向量机的振动传感器故障诊断

Fault Diagnosis of Vibration Sensor Based on Least Squares Support Vector Machines

针对目前机械故障诊断中,难以获得大量的故障数据样本以及诊断知识获取困难等不足,提出了专门针对有限样本的新一代机器学习的算法——最小二乘支持向量机(LS-SVM),它能够得到现有信息下,不仅是样本数趋于无穷大时的最优解,因此,在样本很少的情况下具有较好的泛化能力,比较适合解决故障诊断小样本情况的实际问题。本文介绍了LS-SVM的基本原理和分类方法,并利用其对振动传感器的常见故障进行诊断,结果表明了LS-SVM对设备故障具有良好的分类效果。

It is difficult to acquire lots of fault data and diagnostic knowledge in current mechanical fault diagnosis. The paper presents a new machine- learning algorithm - least squares support vector machines （LS - SVM）, which can acquire the most optimal solution on the limited sample data instead of that on the infinite sample data. LS- SVM specially aims at the small - sample cases, so it has better generalization ability when the data is few, and it is fit to solve the practical problems contained few sample data. The paper introduces the basic theory and classification methods of LS- SVM,and diagnoses the common faults of the vibration sensor using the method. The result shows that LS - SVM really has preferable ability of classification.