2018 10th International Conference on Electronics, Computers and Artificial Intelligence (ECAI 2018)

1 / 2018-03-13 13:33:29

Motor Fault Detection Based on Multi-Agent Classifier System

Multi-agent,Data Classification,fault detection

An early detection of component faults is crucial for motors. In this paper, we present a novel framework to detect and identify fault conditions of three-phase induction motors using an ensemble of classifiers. The Q-Learning Multi-Agent Classifier System (QMACS) is a multi-agent system, which uses the trust-negotiation-communication (TNC) reasoning scheme. Hybrid models of the Q-learning and online neural networks (NNs) are used as learning agents of the multi-agent system. The effectiveness of QMACS for detecting and identifying motor faults is evaluated through experiments. Time-domain statistical features are extracted and fed into QMACS for classification. Experiment results demonstrate that QMACS is able to achieve a superior performance with a decision fusion of its constituents.