Application of sensor fusion and signal classification techniques in a distributed machinery condition monitoring system

摘要

Abstract: A new paradigm for machinery maintenance is emerging aspreventive maintenance strategies are being replaced bycondition-based maintenance. In condition-basedmaintenance, machinery is repaired or serviced onlywhen an intelligent monitoring system indicates thatthe system cannot fulfill mission requirements. Theimplementation of such systems requires a combinationof sensor data fusion, feature extraction,classification, and prediction algorithms. In addition,new system architectures are being developed tofacilitate the reduction of wide bandwidth sensor datato concise predictions of ability of the system tocomplete its current mission or future missions. Thispaper describes the system architecture, data fusion,and classification algorithms employed in adistributed, wireless bearing and gear healthmonitoring system. The role and integration ofprognostic algorithms - required to predict futuresystem health - are also discussed. Examples areprovided which illustrate the application of the systemarchitecture and algorithms to data collected on amachinery diagnostics test bed at the Applied ResearchLaboratory at The Pennsylvania State University.!7