Date of Award
12-2025
Document Type
Dissertation
Degree Name
Doctor of Philosophy (PhD)
Department
Mechanical Engineering
Committee Chair/Advisor
Dr. Javad Mohammadpour Velni
Committee Member
Dr. Phanindra Tallapragada
Committee Member
Dr. Ardalan Vahidi
Committee Member
Dr. Umesh Vaidya
Abstract
Data-driven predictive control enables designing controllers directly from data, making it attractive for complex systems with hard-to-model dynamics. However, practical deployment is challenged by modeling inaccuracies and changing operating conditions. This dissertation develops predictive control frameworks that incorporate robustness and adaptability to address these issues in uncertain nonlinear systems.
The first part employs the Linear Parameter-Varying (LPV) framework, which represents nonlinear dynamics through simple linear form representation. To characterize the plant-model-mismatch often caused by limited data and numerical calculations, Bayesian Neural Networks (BNNs) are used, and their uncertainty estimates are integrated into two robust control approaches. The first is a scenario-based MPC (ScMPC) scheme that performs multi-stage optimization to account for uncertainty. The second is an LMI-based robust controller that embeds the BNN-defined uncertainty into a polytopic set to ensure reliable tracking.
The second part introduces a Neural Parameter-Varying Data-enabled Predictive Control (NPV-DeePC) framework for systems with rapidly changing conditions. Within the behavioral systems theory setting, a neural network represents nonlinear dynamics in an affine form compatible with DeePC, while a hypernetwork updates the parameters online to capture operating variations. This enables adaptive multi-step prediction and effective control.
The frameworks are validated through high-fidelity simulations of a Reactivity Controlled Compression Ignition (RCCI) engine and an Atmospheric Pressure Plasma Jet (APPJ). The LPV-ScMPC and LPV-LMI controllers effectively handle model uncertainty, and the NPV-DeePC method achieves superior temperature tracking and thermal dose delivery. Overall, this dissertation introduces three complementary data-driven control frameworks that provide robust and adaptive solutions for uncertain nonlinear systems.
Recommended Citation
GhafGhanbari, Pegah, "Robust Data-driven Predictive Control of Nonlinear Systems Under Modeling Uncertainty" (2025). All Dissertations. 4154.
https://open.clemson.edu/all_dissertations/4154