Induction motor fault diagnosis using industrial wireless sensor networks and Dempster-Shafer classifier fusion

This paper presents a novel induction motor fault diagnosis system using industrial wireless sensor networks (IWSNs), in which on-sensor feature extraction and fault diagnosis approaches are investigated to address the tension between the higher system requirements of IWSNs and the resource constrained characteristics of sensor nodes. Classifier fusion using Dempster-Shafer theory on the coordinator is then explored to increase diagnosis result quality. Three kinds of motor operating condition - normal, loose feet, and mass imbalance - are monitored to evaluate the proposed system. Experimental results show on-sensor feature extraction and fault diagnosis could effectively reduce payload transmission data, and decrease node energy consumption, while Dempster-Shafer classifier fusion significantly improves fault diagnosis accuracy compared with using local neural network classifiers alone.