Date of Award

1-1-2011

Document Type

Campus Access Dissertation

Department

Electrical Engineering

First Advisor

Charles Brice

Abstract

In this dissertation, a general methodology for drive diagnostics, using decision trees populated with data from numerous techniques, will be presented. Older, large drive systems lack the diagnostic abilities to report back issues and possible sources. It is not a cost effective solution to change out these multi-million dollar systems due to this shortcoming, nor is it practical to monitor every data point because of hardware cost and the availability of expert personnel to interpret those recordings. Quite often, the technical experts are not in the United States to assist with resolving issues. A recorder with triggered alarms for torque or current thresholds would be insufficient due to system dynamics. The ideal solution is an intelligent monitoring system; however, no system exists that addresses the entire process. There are a number of recent technical articles about modeling and simulation, but very few actually addressed the issues of industrial systems. The system must monitor these high power controllers and report abnormal events or anomalies based on the process dynamics and based on the rules defined within the decision tree. Such a system needs to operate in real time and report back so that scheduled repairs can be made to prevent costly unexpected failures from occurring. The system developed in this work is unique in its capabilities. This dissertation outlines the general methodology for drive diagnostics that are driven from various inputs provided to a decision tree. Such a technique offers the flexibility for the system to be modified for a family of drive systems, such as dc drives or induction machines. This pooling of information into the decision tree encapsulates and mimics the trouble-shooting steps that an expert would take, even to the point of testing for rotational harmonic disturbances in the system. This methodology allows for the continual improvement of the rules of the decision tree, techniques to pool the input data, and the flexibility to be modified for a family of drives and motors.

Share

COinS