A Novel Approach to Fault Modeling, Diagnostics, and Prediction in Motor Drive Systems

Marquette University, Milwaukee WI

Investigators

Abstract

The goal of this project is to develop novel and effective approaches to modeling, diagnosing, and predicting faults in electric motor drive systems. Such systems are used throughout industry. Failure of motor drive systems has serious impact, including plant shutdown, negative environmental effects, and worker endangerment. Accurate diagnostics and fault prediction will increase the reliability of such systems and minimize the problem of failures in the field. We address three fundamental problems in developing a motor drive diagnostic system: 1. The difficulty and cost of obtaining large amounts of motor drive field failure data. 2. The difficulty in developing accurate, specific, robust, and fast diagnostic techniques. 3. The lack of methods for predicting future motor drive failure. Thus, there are three primary objectives: 1. Generate large datasets of high fidelity simulations, which requires significant advances in the Time Stepping Coupled Finite Element-State Space techniques. 2. Develop reconstructed phase space modeling techniques for characterizing motor drive waveforms. The resulting models will form a comprehensive database for diagnosing motor failures. 3. Develop methods to predict the onset of faults through a novel system tracking approach based on Extended Kalman Filter banks. The transition between incipient and full-fledged faults will be tracked and predicted. With our industrial partner, Rockwell Automation, a sample of fault simulations and fault signatures will be verified on actual motor drives. A larger set will be verified using a configurable motor drive system. A prototype diagnostic system will be developed and tested on a centrifugal pump test bed.