为提高汽轮机故障诊断的准确率,本文提出一种基于小波包能量与隐马尔可夫模型相结合的汽轮机故障诊断方法.对汽轮机振动信号进行小波包分解,将小波包能量作为特征集;分别对每种故障状态的样本训练HMM(Hidden Markov Model),并构建故障诊断知识库;最后利用训练好的HMM对待测样本进行故障诊断.通过对汽轮机常见故障的诊断分析表明,基于隐马尔可夫模型的故障诊断方法的准确率优于BP神经网络、SVM(Suppvrt Vectwr Machine)等方法.%In order to improve the diagnosis precision,a fault diagnosis method turbine generators is proposed based on wavelet packet energy and hidden Markov model (HMM).Firtsly,the wavelet packet decomposition is used to generate feature sets.Then,an HMM model is trained for working conditions on turbine generators.Finally,the trained HMMs are applied to calculate the probabilities of testing samples as inputs,whereas the maximum value determines corresponding class.Therefore,it is proven from experimental results that the performances of the proposed method are better than those of BP network and SVM method.
展开▼
