A New Fault Detection and Isolation Algorithm Applied to Dc Motor Parameters Supervision

Abstract. This paper presents a fault detection and isolation algorithm applied to the monitoring of DC Motor parameters. Fault Detection and Isolation (FDI) schemes are implemented as real-time algorithms whose inputs are plant output observations. They are basically used for a) fault detection: to decide whether the plant is in a normal operating condition or in a faulty one and b) fault isolation: to point out and identify the kind of the fault (if present) among a given fault set. DC Motors suffer from diverse possible critical failures that could compromise its performance and cause severe gear damage, such as, armor coil opening, field coil opening, armature static converter short circuit, field static converter short circuit, armature coil short circuit, field coil short circuit, cooling system failure, lack of bearings and bushing lubrication, armature current sensor failure, field current sensor failure, speed sensor failure. The proposed algorithm uses the singular values of a Hankel matrix built from output measurements to detect and isolate DC Motor parameter failures. The main feature of the proposed algorithm is that it does not rely on the plant model identification. Having obtained a nominal plant image through the singular values of the Hankel matrix, this image can be used to determine, by comparison, any value drift of the plant parameters. Two functional levels of the procedure are distinguished, namely alarm generation and alarm interpretation. At the alarm generation level (detection), the algorithm naturally displays plant failure through the change of the singular values structure and values and at the alarm interpretation level (isolation), the algorithm delivers an image of the plant parameters through the singular values allowing the identification of the faulty parameter. Simulation examples are presented to illustrate the performance of the proposed algorithm.