Detection and Estimation of Extremely Small Fault Signature by Utilizing Multiple Current Sensor Signals in Electric Machines

This paper presents a novel motor current signature analysis algorithm for accurate detection and precise estimation of extremely small fault signature by utilizing multiple-phase-current signals in an electric machine. The two most critical challenges in developing a reliable diagnosis method are the detection of extreme small signals under extremely harsh environment, and the formulation of adaptive threshold under dynamic operation and noise variation of a motor. These challenges are addressed in this paper by logically and statistically utilizing multiple existing current sensors and current signals to make an accurate detection, and statistical/adaptive threshold for precise detection and decision making. The performance of the proposed algorithm has been mathematically and theoretically proved under modeled Gaussian noise conditions. Proposed theory is also experimentally verified on a 3-kW five-phase permanent-magnet-assisted synchronous reluctance motor.