论文部分内容阅读
文中通过对隐式Markov模型(HMM:hidden Markov model)假设条件的松弛研究,提出了基于自回归隐式半Markov链(AR-HSMM:auto-regressive hidden semi-markovmodel)的设备健康诊断和预测新方法.与传统的HMM相比,AR-HSMM具有3个优点:一是将传统HMM所假设的隐藏状态分布改进为显式Gauss分布,因此能够用于设备性能衰退预测;二是改进了传统HMM中各观测变量相互独立的假设,通过自回归建立各观测变量之间的依赖关系,从而使之更加符合实际情况;三是AR-HSMM不必服从不现实的Markov链条件,因而具有更强的建模和分析能力.文中定义了新的“前向-后向”变量,给出了改进的“前向-后向”算法.通过一个实例对所提出的方法进行评价与验证.实验结果表明,基于AR-HSMM的设备健康诊断和性能衰退预测新方法是有效的.
In this paper, the device health diagnosis and prediction based on AR-HSMM (auto-regressive hidden semi-markov model) are proposed based on the relaxation of the assumption of implicit Markov model (HMM) Method.Compared with the traditional HMM, AR-HSMM has three advantages: one is to improve the hidden state distribution assumed by the traditional HMM to an explicit Gauss distribution, so it can be used for the prediction of equipment performance degradation; the other is to improve the traditional HMM The independent variables in the observation variables, the autoregression to establish the dependence between the observed variables, so as to make it more in line with the actual situation; Third, AR-HSMM does not have to follow the unrealistic Markov chain conditions, which has a stronger built Modalities and analytical abilities, a new “forward-backward” variable is defined in this paper and an improved “forward-backward” algorithm is given. The proposed method is evaluated and verified through an example. Experimental results show that AR-HSMM based device health diagnosis and performance degradation prediction of new methods is effective.