Research
Research Mission
As engineering systems continue to increase in complexity, there is a growing need for rigorous and systematic approaches to analyze, evaluate trade-offs, make decisions, and design these systems to meet stakeholder needs. My research focuses on developing computational methods and tools that accelerate the design of next-generation aerospace vehicles while ensuring adherence to safety standards set by regulatory bodies.
State-of-the-art methods are normally used at later stages of the design process, and by then critical decisions in terms of configuration and size may have been made, which translates into correcting efforts that are costly and time-consuming. Thus, the methods and tools must be practical for early-stage conceptual design.
Three significant challenges exist:
High-fidelity tools require knowledge of detailed parameters that are not known early in the design, especially for advanced configurations
My work addresses this challenge through multidisciplinary design optimization (MDO).Large runtime of simulations involved in the structural weight estimation of advanced configurations
My work applies physics-based and data-driven model reduction techniques to lower simulation runtime, thereby reducing design-cycle time.Slow and costly response to rapidly changing requirements during the design of advanced configurations
My work develops design processes and efficient decision-making through the use of model-based systems engineering (MBSE).
Multidisciplinary Design Optimization (MDO)
Physics-based methods often rely on detailed parameters that are typically unknown during the early design phase, especially for advanced configurations.
Multi-Disciplinary Optimization (MDO) provides a viable approach to incorporating physics-based methods in the early stages of aircraft design. MDO treats these unknown parameters as design variables within an optimization problem, enabling the use of physics-based tools in early-stage design.
A bottleneck in MDO is the challenge of obtaining derivatives of the objective function and constraints in relation to design variables. My research draws on symbolic differentiation methods to present physics-based models as computational graphs thereby automating derivative computations.
Model Reduction for Structural Weight Estimation
A key activity in the design of aircraft is estimating the structural weight. This requires consideration of geometric and material nonlinearities and dynamic aeroelastic effects. State-of-the-art methods typically combine high-fidelity CFD analysis with high-fidelity shell-based structural analysis. In MDO, these methods are run hundreds or thousands of times to find an optimal design that ensures the structural integrity of the aircraft. However, these high-fidelity methods may take days or weeks to run, thus making their use prohibitively expensive.
Model reduction creates approximate fast-to-evaluate models that can be used in MDO. My research explores two forms of model reduction: (1) physics-based methods to create approximate aerodynamic and structural representations of the aircraft, and (2) data-driven methods to create multi-fidelity reduced order models (ROM).
Model-Based Systems Engineering (MBSE)
MBSE, as defined by INCOSE SE Vision 2020, involves formalized modeling techniques applied throughout the system development lifecycle, encompassing system requirements, design, analysis, verification, and validation.
This work recognizes the inherent limitations of MBSE, such as the lack of analytical and numerical models for performance prediction and optimization with a large number of design variables. To overcome these limitations, the research explores the integration of MDO.