Integration of machinery condition monitoring and reliability modeling: A prelude to predictive maintenance.

摘要

Condition Based Maintenance (CBM) is a philosophical approach that uses the most cost effective methodology for the performance of machinery maintenance. The idea is to ensure maximum operational life and minimum downtime of machinery within predefined cost, safety and availability constraints. When machinery life extension is a major consideration the CBM approach usually involves predictive maintenance. In this research a two-level approach for predictive maintenance has been defined: (1) to develop a Condition Monitoring and Diagnostic System (CMDS) for machine fault detection and maintenance suggestion, and (2) to develop a machine performance estimation model for machine reliability modeling and failure rate analysis. The objective is to provide a new and practicable solution for condition-based predictive maintenance.;In this research artificial neural network (ANN) technologies and analytical models have been investigated and incorporated to increase the effectiveness and efficiency of CMDS. Several advanced vibration trending methods have been studied and used to quantify machine operating conditions. An on-line, multi-channel condition monitoring procedure has been developed and coded. The major technique used for fault diagnostics is a modified ARTMAP neural network. In the second part of this research a new method of obtaining maintenance information has been developed. A Cerebellar Model Articulation Controller (CMAC) neural network has been employed to estimate and quantify machine performance. By combining reliability theory with a real-time, on-line CMAC Performance Estimation Model (CMAC-PEM), machine reliability statistics such as failure rate and mean time between failures (MTBF) can be calculated. CMAC-PEM may provide a practicable solution for condition-based predictive maintenance since it estimates machine reliability measures on-line. In addition, Weibull Proportional Hazards Model (WPHM), has been implemented as a proven tool to verify CMAC-PEM results. Real-world data obtained from a bearing fault experiment and a bearing deterioration process were provided to test the proposed methodologies.;Essentially, this research presents an innovative method to synthesize low level information, such as vibration signals, with high level information, like reliability statistics, to form a rigorous theoretical base for condition-based predictive maintenance.