Student Name: Mariam Emara

Advisor: Dr. Dimitri Mavris

Milestone: PhD Thesis Final Examination (Defense)

Degree Program: Aerospace Engineering

Title: A Multi-fidelity modeling approach for vehicle performance tuning

Abstract: Simulation and virtual testing are central to modern vehicle dynamics analysis, where accurate performance evaluation is required across a wide range of operating conditions. As reliance on digital models increases, engineering teams are faced with the challenge of balancing predictive accuracy with computational cost. High fidelity models offer detailed and accurate predictions at a high expense, while low fidelity models are computationally efficient but may lack reliability. In practice, model fidelity selection is often subjective or constrained by available tools, leading either to unnecessary cost or unreliable predictions. This work addresses the research question: For vehicle dynamics models, what level of model fidelity best satisfies accuracy and computational cost requirements for reliable and efficient vehicle performance evaluation? A systematic methodology is developed to support model fidelity selection by integrating multi-fidelity modeling methods with uncertainty quantification and decision-making frameworks. Multi-fidelity methods are used to trade accuracy and cost by combining information from models of varying fidelity, while epistemic uncertainty quantification enables a comprehensive assessment of model accuracy and reliability beyond error metrics alone. The proposed approach begins with the construction of a library of multi-fidelity model alternatives, including surrogate-based methods, hierarchical fidelity-switching simulations, and mixed-fidelity component models. Model accuracy is evaluated using a fidelity metric that incorporates both prediction error and uncertainty, and multi-attribute decision-making techniques are applied to balance fidelity, computational efficiency, and usability for vehicle performance evaluation. Application to an automotive vehicle dynamics use case demonstrates gains in computational cost while maintaining adequate predictive accuracy. The methodology provides a transparent and repeatable process for identifying suitable model simplifications, selecting among multi-fidelity modeling strategies, and determining the appropriate fidelity level under varying requirements and driving scenarios. Although demonstrated for car dynamics, the approach is broadly applicable to other dynamic systems, supporting knowledge transfer across engineering domains.

Date and time: 2026-02-12, 10:00 am

Location: Collaborative Design Environment (CoDE) - Weber Space Science and Technology Building (SST II)

Committee:
Dr. Dimitri Mavris (advisor), School of Aerospace Engineering
Prof. Daniel Schrage, School of Aerospace Engineering
Prof. Graeme Kennedy, School of Aerospace Engineering
Prof. Shreyes Melkote, School of Mechanical Engineering
Dr. Michael Balchanos, School of Aerospace Engineering
,