Nonlinear observer design and failure diagnostics for thermofluid systems
Abstract
The increasing availability of low cost microprocessor technology has prompted the investigation of on-board process diagnostics for improved appliance maintenance and performance. In this research project, two failure detection strategies are developed for monitoring thermofluid systems. A model-free limit and trend checking scheme directly monitors the system signals for transgressions of fixed operating thresholds. A model-based innovations failure detection scheme uses a system model, nonlinear observer, and statistics to monitor the process operation. The performance of several nonlinear observer design strategies has been studied on a single finned-tube heat exchanger, and refrigerator compressor and condenser model. The declaration of a system failure by either detection technique activates the failure identification scheme. A robust multiple-hypothesis isolation strategy has been developed using control theory, pattern recognition, and multivariate statistics. The use of an experimental a priori database minimizes the modeling activities associated with each hypothesized failure mode. A series of anomalies are experimentally simulated on a household refrigerator to demonstrate the feasibility of applying failure detection and identification techniques to thermofluid systems.
Degree
Ph.D.
Advisors
Shoureshi, Purdue University.
Subject Area
Mechanical engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.
