基于隐半马尔可夫模型设备退化状态识别方法研究

机械设备从正常到故障往往经历一系列退化状态,正确识别与估计设备当前所处的退化状态,对预防设备进一步退化和故障的发生具有重要意义。隐半马尔可夫模型(Hidden Semi-MarkovModels,HSMM)是隐马尔可夫模型(hidden Markov models,HMM)的一种扩展模型,克服了因马尔可夫链的假设造成HMM建模所具有的局限性,比HMM具有更好的建模能力和分析能力。由状态识别和HMM本质上的相通性,将HSMM引入到机械设备的状态识别中,提出了一种基于HSMM状态识别方法,描述了该模型的拓扑结构和主要参数以及相应的训练和识别算法。最后通过滚动轴承试验系统验证了方法的有效性。

On Equipment Degradation State Recognition Using Hidden Semi-Markov Models

The failure process of mechanical equipments usually consists of a series of degradation states;correctly recognizing and estimating the current state of equipment is important for preventing equipment from further degradation and occurrence of failure.Hidden semi-Markov models(HSMM) is an expanded model of hidden Markov models(HMM);it can overcome the modeling limitation of HMM due to the Markov property and therefore is better for modeling and analyzing.Because state recognition was similar to HMM in nature,a new state recognition method based on HSMM is proposed in this paper.The topology of this model and its parameters are described in detail.Finally,the new method is tested with experimental data collected from a roller bearing experimental system and the results demonstrate that it is very effective for recognizing the degradation states of the roller bearing.