Abstract: One motor condition recognition system based on multi-agent decision fusion was proposed.Six classifiers were used to classify motors condition by system inputs:vibration and current signals.In the system,each classifier was considered as an agent,which independently completed recognition task,then exchanged information among classifiers to improve classification accuracy.Sensor fusion and classifier selection were put into the system,and this method was much better than one-single signal and no classifier selection.The best recognition result of the proposed system achieved 98.9%.