A Two-Stage HSMM Model for Diagnosis and Prognosis of Gearboxes

作者：Jia Xisheng; Teng Hongzhi^*; Hu Qiwei

来源：Information-An International Interdisciplinary Journal, 2012, 15(12B): 5819-5828.

• Diagnosis
• Prognosis
• Choi-Williams distribution
• Test-to-failure experiment
• Remaining useful life
• Hidden semi-Markov model

The implementation of condition-based maintenance (CBM) can lead to significant benefits for industry. Diagnosis and prognosis are considered as one of the main steps in the CBM process. Hidden semi-Markov model (HSMM) can be used to identify the characteristics of each stage of the failure process and describe the failure process which is the basis of using HSMM for diagnosis and prognosis. In this paper, a two-stage HSMM model for diagnosis and prognosis of gearboxes is proposed. The diagnosis process is separated into two stages. The failure mode is identified in stage 1 by selecting the classifier of HSMM which maximizes the log-likelihood function of a given observation. Stage 2 is to recognize the health level of a given failure mode which was identified in stage 1. The remaining useful life is estimated by HSMM. A method to extract the features of gearbox signals is studied. Feature extraction from vibration signals is carried out by time-frequency analysis. Through a test-to-failure experiment of a gearbox, the feasibility and effectiveness of this model are verified.

• 出版日期2012-12