基于HMM的设备故障预测方法研究

故障预测是设备实施基于状态维修的1个重要内容，是实现真正意义上精确维修的前提和基础。隐马尔可夫模型（HMM）作为1种统计分析算法，在设备的故障诊断中获得了成功应用。但对于故障的预测，传统隐马尔可夫模型存在很多缺陷，因此研究相关的改进算法，构建了基于隐马尔可夫的故障诊断和预测框架，使设备的故障诊断和预测能够同时进行。最后通过对滚动轴承实测数据的仿真验证，表明该算法具有较高的故障识别率并且对设备的剩余寿命能进行有效的预测。

Study on the Equipment Fault Prognosis Based on Hidden Markov Model

Fault prognosis is the important aspect in a condition-based maintenance which indeed actualize the maintenance in a proper time. As a statistic analysis algorithm, the Hidden Markov model （HMM） has a successful application in equipment fault diagnosis. But the effect on prognosis using tradition HMM is not explicit. So some improvenent for the traditional model is introduced. The algorithm based on improved HMM for performing both diagnosis and prognosis in a unified framework is then presented. As a result, the diagnosis and prognosis can be implemented at the same time. Finally, the proposed algorithm was carried out through a roller beating experiment. The result indicated that the algorithm was effective to classify classical fault and have the ability to predict the remaining useful life （RUL）.