Optimal Design of System Architectures: From Formal Methodologies to Crowdsourcing Games
A. Emrah Bayrak, PhD
The Polytechnic School
Ira A. Fulton Schools of Engineering
Thursday, Feb. 25, 2016
Santan (SANTN) 220, Polytechnic campus [map]
Free to attend
A long-time challenging problem in design optimization is the representation and optimization of system architectures. Layout of the building blocks that constitute a system and interactions among these building blocks define the architecture of a system. This seminar presents research approaches to two sets of design problems namely, the optimal design of powertrain architectures for hybrid electric vehicles, and the design of modular architectures for fleets of vehicles. In this research, hybrid electric powertrain design starts with generating the feasible architecture alternatives with a general architecture representation. Optimal architecture design among these feasible alternatives must be determined together with powertrain component sizing and a control strategy distributing the power demand to powertrain components because component sizes and controller design impact vehicle performance and fuel economy, and hence the architecture selection. This research uses a decomposition-based design formulation for the complete hybrid electric architecture design problem. In design of modular vehicle architectures, the problem is defining a set of modules to create a fleet of modular vehicles that maximize the mission performance. Design decisions made at the module-level must be evaluated with a detailed fleet operation accounting for various aspects such as vehicle performance, maintenance/repair operations and life-cycle cost. This research uses a function-based representation of a vehicle fleet to design modules in an optimization framework. This seminar also presents an online game-based crowdsourcing approach to system design that relies on extracting design heuristics from gamers. Such a crowdsourcing approach can be an alternative when formal design methodologies become computationally intractable.
A. Emrah Bayrak is currently a post-doctoral research fellow working at the Optimal Design Lab in the University of Michigan. He received his B.S. degree in mechatronics engineering in 2011 from Sabanci University in Istanbul. He, then, moved to the University of Michigan where he received his M.S. in 2013 and PhD in mechanical engineering in 2015. His research focuses on developing integrated system design approaches to engineering systems including applications to automotive industry. He is particularly interested in the optimal design and control of powertrain system architectures, design of modular platforms for a fleet of vehicles. In addition to the formal design methodologies, he also works on the application of online games as a crowdsourcing medium to solve such system-level design problems. He also teaches a graduate-level design optimization course at the University of Michigan.