因子隐Markov模型(Factorial Hidden Markov Models,FHMM)是模拟语音特征的概率统计模型。它也广泛地应用在动态时间序列概率模型的学习中。由于旋转机械振动监测过程的某些动态因素类似于语音中的动态模式,因此,将FHMM引入到旋转机械升降速过程中,通过实验模拟了旋转机械在升速过程中的动态行为,并对振动数据进行了适当的处理之后,建立了各种典型故障的FHMM模型。利用这些模型进行故障分类,实验结果表明该方法是十分有效的。 Factorial Hidden Markov Models (FHMM) are probabilistic statistical models of simulated speech features. It is also widely used in the learning of dynamic time series probabilistic models. As some of the dynamic factors in the vibration monitoring process of rotating machinery are similar to those in speech, FHMM is introduced into the process of raising and lowering speed of rotating machinery. The dynamic behavior of rotating machinery during its rising process is experimentally simulated, After the data has been properly processed, various typical fault FHMM models have been established. Using these models for fault classification, the experimental results show that this method is very effective.