中文摘要：

为了充分利用先验信息和实测信号，提高故障识别率，根据Bayes方法和序贯决策的思想，将实测信号分段，将前一段信号的诊断后验信息作为后一段信号的先验信息，提出了一种基于隐Markov树（hidden Markov tree，HMT）的序贯故障诊断模型。给出了诊断模型的建模步骤、HMT模型的建立方法和Bayes后验概率的计算方法。将模型应用于某型减速器故障诊断的结果表明，对于有先验信息和无先验情况，该序贯模型都可以有效地提高故障识别率。

英文摘要：

To fully utilize prior information and on-line signals in order to enhance fault recognition rate, following Bayesian method and the sequential decision thought, a sequential fault diagnosis model based on Hidden Markov Tree （HMT） is presented in this paper. On -line signals are divided into segments, and posterior of one segment is used as the prior for next segment. The diagnosis modeling procedure, modeling method for HMT and computational methods for Bayesian posterior probability are also given. The application of the presented methods to a gearbox fault diagnosis shows that the sequential model can effectively improve fault identification rate in situations either having prior or with no prior information.