Set-Based Fault Diagnosis and Dynamic Optimization of a Cellulose Oxidation Reactor
Author(s)
Mu, Bowen
Advisor(s)
Editor(s)
Collections
Supplementary to:
Permanent Link
Abstract
This dissertation focuses on control methods for state-space models to solve two specific problems: fault detection and diagnosis in uncertain nonlinear systems, and optimal control of a cellulose oxidation reactor.
The first portion of the thesis concerns fault detection and diagnosis (FDD) in uncertain nonlinear chemical systems. In modern chemical processes, faults such as leaks, sensor failures, and other equipment malfunctions happen often. Quickly detecting and diagnosing them is critical for safe operation. Classical data-driven FDD algorithms are simple and widely implemented in industry. However, they often suffer from high false alarm rates and misdiagnoses in particular cases. In this thesis, we developed a new set-based FDD algorithm considering the state-space model. Given a dynamic model with bounded uncertainties, this method can predict rigorous bounds that accurately enclose all possible outputs of the model in the next time step. These bounds are then used to test for faults in the real process by checking whether the measured outputs are within the computed bounds.
The second portion of the thesis focuses on the modeling and dynamic optimization of a cellulose oxidation reactor. Oxidized regenerated cellulose (ORC) is widely used in surgical procedures to control bleeding. However, its production from regenerated cellulose in the oxidation reactor is in low efficiency in industry. In this thesis, we improved the ORC production process by building a first-principles model of the reactor and optimizing it dynamically. For the modeling work, the reaction kinetics, mass and heat transfer, and vapor-liquid equilibrium within the reactor were considered. The model predicted reactor performance under different operating conditions. For the optimization work, an optimal control problem was formulated to find the optimal oxidant and heat input profiles. The results provided key insights into the reactor operating strategy, guided experiments in the lab, and demonstrated their value in real-world ORC production.
Sponsor
Date
2024-11-15
Extent
Resource Type
Text
Resource Subtype
Dissertation