Abstract: To improve the accuracy and validity of equipment health diagnosis and prognosis,a method of equipment management integrating multi-sensor information based on Adaptive Hidden Semi-Markov Model (AHSMM) was proposed.The basic algorithms of AHSMM were modified to decrease computation and space complexity,and the modified HSMM with multi-sensor information was applied,in which the hidden degradation process could be seen as the system state.The maximum likelihood linear regression transformations method was used to train the output and duration distributions to re-estimate all unknown parameters.The AHSMM could also be used to obtain the transition probabilities among health states and health state durations.The main results were verified by a case study,and the results showed that it had a better performance for health diagnosis and prognosis than HSMM.

Key words: adaptive hidden semi-Markov model, health diagnostics, residual useful lifetime, maximum likelihood linear regression, multi-sensor information

摘要： 针对设备健康诊断与寿命预测问题,提出一种基于自适应隐式半马尔可夫模型(AHSMM)结合多传感器信息的设备健康预测方法。提出了AHSMM的前向—后向算法、Viterbi算法和Baum-Welch算法,有效降低了模型的计算复杂性。利用最大似然线性回归训练对输出概率分布和驻留概率分布进行自适应训练,处理多传感器信息间的差异性,进行有效的多传感器信息融合,以更加准确地进行设备健康诊断与寿命预测。利用失效率理论建立了对设备剩余使用寿命进行预测的基本步骤。通过美国卡特彼勒公司液压泵的状态识别和健康预测实际案例对所提出的方法进行评价与验证,实验结果表明,基于AHSMM的设备健康诊断和性能衰退预测方法比传统的隐式半马尔可夫模型(HSMM)更有效。

关键词: 自适应隐式半马尔可夫模型, 健康诊断, 剩余有效寿命, 最大似然线性回归, 多传感器信息