Research On Machinery Condition Assessment Based On Age-dependent Hidden Markov Model

Machinery Maintenance Management is an important part of company management,which plays a vital role in reducing the cost of company's running and enhancing company's competition ability.After Periodical Maintenance,Breakdown Maintenance and Emergency Maintenance,Condition-based Maintenance?CBM?has become the prominent method for machinery maintenance.Prognostics and Health Management technology?PHM?now is a key technique for CBM,and Machinery Health Condition Assessment is the basic supportive step of PHM,which has important academic value and research meaning.Recently the research about machinery health condition assessment excessively focuses on fault diagnosis,which ignores that before machines fail to work their performance may become degraded.Hidden Markov Model?HMM?has been applied in the area of machinery condition assessment,because of its powerful function,and has shown good performance.To depict the degraded process of machine's health condition more accurately,some researchers endeavour to adapt traditional HMM.This paper proposes an adapted HMM in which the transition probability is considered to be fixed i.e.aging factor is introduced to the traditional HMM to build an age-dependent HMM.To be specific,this paper completes these following contents:?1?Reviewing the machinery condition assessment process based on traditional HMM systematicallyTraditional HMM includes three parts:data collection and processing,HMM building and state evaluation.About data collection and processing,this part introduces the methods of original data pre-processing,feature extraction and descending dimension.HMM building includes state classification,and initial parameters confirmation and training algorithm.For state evaluation,the Maximum a posteriori estimation method is used to confirm machine's health state.?2?Introducing aging factor to traditional HMM and develop an algorithm to estimate new parametersThis paper introduces aging factor to HMM and build a model for health condition assessment based on age-dependent HMM.This paper develops a new double EM algorithm to estimate parameters:initial states transition probability array A₀ and aging factor.?3?Building the condition assessment process framework based on the adapted HMMThe framework for health condition assessment based on the adapted HMM embraces three parts:date collection and processing,age-dependent HMM building and state evaluation.this paper proposes the evaluation methods based on the age-dependent HMM.Finally,this paper uses a numerical case to perform the framework,and the effect and accuracy of the framework can be validated through this case study.To some extent,the method proposed in this paper can be validated to support the current research for machinery condition assessment.