摘要
剩余寿命预测是作出正确的状态维修决策的基础和前提,是设备退化状态识别的重要内容。隐马尔可夫模型(HMM)是一种具有较强模式分类能力的统计分析算法,但是它不能直接用于剩余寿命的预测,而且考虑到隐马尔可夫模型的局限性和剩余寿命预测模型的可解释性,应用隐半马尔可夫模型(HSMM)进行建模和预测。针对HSMM的训练算法极易陷入局部极值点的问题,提出了基于改进微粒群优化算法(MPSO)进行修正。实验结果证明了该方法在设备剩余寿命预测研究上的有效性和可行性。
Prediction of equipment residual life based on the recognition of degradation is the important aspect in a condition-based main- tenance which indeed actualizes the maintenance in a proper time. As a statistic analysis algorithm, the Hidden Markov Model (HMM) with well capability in pattern classification has a successful application in identification of equipment degradation state. But HMM cannot be directly used to prognosticate residual life. In this paper, considering the limitations of HMM and the explanation of remaining life pre- diction model, apply the Hidden Semi-Markov Model (HSMM) for modeling and forecasting. In view of problems that HSMM training algorithm can easily fall into local extreme point ,the algorithm based on Particle Swarm Optimization (PSO) is proposed to improve. Ex- perimental results show that the method on the residual life prediction of equipment has effectiveness and feasibility.
出处
《计算机技术与发展》
2014年第1期184-187,191,共5页
Computer Technology and Development
基金
国家自然科学基金资助项目(41272374)
关键词
隐半马尔可夫模型
微粒群优化算法
剩余寿命
预测
hidden simi-Markov model (HSMM)
particle swarm optimization (PSO)
residual life
forecast