基于支持向量机的提升机制动系统故障诊断

针对提升机制动系统中常见的卡缸故障，利用支持向量机（SVM）这一新的机器学习方法进行智能诊断．在某一闸系统正常时获得2组信号，卡缸时获得6组信号，采用3层小渡包对闸瓦间隙-时间信号进行分解，以各频带的能量为元素构造特征向量，形成故障诊断样本，在Matlab6．5环境下用SVM工具箱进行编程，建立SVM故障分类器并时测试样本进行测试，从而实现提升机制动系统卡缸故障诊断．实验结果表明，在不到0．1S时间内，就建立了SVM故障分类器，该分类器对测试样本的诊断正确率达到了100%；当训练样本由6组减少至4组时，SVM故障分类器仍可以有效地实现对卡缸故障的诊断．因此，SVM方法对于少样本的故障诊断有较强的适应性，非常适合于矿井提升机这种安全运行要求很高，但又不具备大量故障样本的系统．

Mine Hoist Braking System Fault Diagnosis Based on a Support Vector Machine

The blockage piston in cylinder, a typical fault of mine hoist braking system, was intelligently diagnosed with a support vector machine(SVM) method. Gathering two sets of signals while in normal working and six sets of signals while in blockage piston in cylinder, using a wavelet package of 3 levels to decompose signal curves of brake distance-time, and construct feature vectors with the energy of all individually frequency bands signal, and form the fault diagnosis samples. Based on SVM toolbox in Matlab6. 5 system, SVM fault classifier was constructed for testing samples and the blockage piston in cylinder fault diagnosis was completed. Experimental results show that the SVM fault classifier was constructed within 0. 1 second and the testing samples diagnosis correct rate of this classifier is up to 100%. It can also effectively accomplish the blockage piston fault diagnosis of braking system when the training samples are decreased from 6 sets to 4 sets. Hence, the SVM method has a high adaptability for fault diagnosis in the case of smaller number of samples and is especially suitable for mine hoist.