Fault Detection and Diagnosis for Uncertain Nonlinear Systems Using Set-Based State Estimation

点击翻译按钮获取中文摘要

This project will contribute to the economic prosperity and safety of the nation by advancing automated methods for detecting and diagnosing component malfunctions and other abnormal events (such as, faults or failures) in complex engineering processes. Many key US industries are increasingly reliant on complex, integrated, and highly automated systems, including manufacturing, chemicals, energy, transportation, and even medicine. Failures in such systems are inevitable, and those that go undetected or misdiagnosed can be responsible for massive economic losses, catastrophic environmental damage, and loss of life. This project will develop methods for detecting the occurrence of faults and determining their root causes with significantly increased speed and accuracy, thus enabling rapid corrective actions. The educational component of the project will include undergraduate curriculum development and inclusion of undergraduate researchers in implementation of the developed methods on the South Hinson Chiller plant at Clemson University.The research objective of this project is to develop and experimentally validate advanced set-based fault detection and diagnosis (FDD) algorithms with greatly enhanced detection speed and accuracy for uncertain nonlinear systems, by combining state-of-the-art set-based FDD algorithms for linear systems previously developed by the principal investigator with a powerful new approach for bounding the reachable sets of nonlinear systems under uncertainty. Current industrial fault detection systems most often operate by comparing process measurements to historical data. Consequently, their ability to distinguish faults from normal disturbances, and to identify root causes, depends critically on the applicability of the historical data to the current conditions. Thus, these methods are susceptible to false alarms and misdiagnoses when confronted with frequent transients, highly variable or uncertain inputs, nonlinear dynamics, or first-of-a-kind faults. The developed methods will exploit process models to characterize the process outputs consistent with normal operation and to provide a rigorous basis for fault detection and diagnosis. This project specifically considers methods that furnish a guaranteed enclosure of these outputs using set-based state estimators, which eliminates the possibility of false alarms.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.