Date
Event Location
3540 Engineering

Mechanical Engineering Faculty Candidate

Tuesday, April 17, 2018

10:15 a.m., 3540 Engineering Building

Refreshments Served at 10:00 a.m.

Faculty Round Table Discussion

Room 2555D Engineering Building

Leveraging nonlinear dynamics and data-driven control towards higher energy efficiency

Ashkan Haji Hosseinloo

Ph.D. Candidate

MIT

Abstract:

As the complexity and connectivity of the structures and the machines we build increase, conventional analysis, design, and control methods become less effective. In my research I leverage nonlinear dynamics and data-driven controls for analysis and control of complex dynamical systems ranging from vibratory energy harvesters and smart buildings to autonomous vehicles and robotics. In this talk I will present how inclusion of nonlinearity and data-driven controls in presence and absence of an accurate model, respectively, could be employed towards higher technology efficiency. In the first part of the talk I focus on the fundamental power limits of nonlinear vibration energy harvesting and techniques for approaching these limits. To this end, I will first present a generic model-based framework for calculation of the power limits. As a byproduct of this analysis, I then characterize a universal nonlinear optimal law, termed the Buy-Low-Sell-High (BLSH) strategy that maximizes the harvested power. I will conclude the first part of the talk by proposing three mechanisms to implement this strategy: latch-assisted harvesters, adaptive bistable harvesters, and harvesting via structural instabilities.

In the second part of the talk I will present my recent work on data-driven and autonomous climate control of next-generation smart buildings to concurrently improve occupants’ comfort and energy efficiency. Buildings account for nearly 40% of the total energy consumption in the United States, about half of which is used by the HVAC systems. Traditional rule-based and model-based strategies for the control of HVAC systems are often far from optimal due to the complexity of building thermal dynamics and stochasticity in the environment. To remedy this issue, I employ reinforcement learning techniques which are tailored to find adaptive and more efficient control policies while relying only on a small data-set. The general ideas and techniques presented here are applicable to other areas of research including autonomous vehicles, robotics, and predictive maintenance.  

About the Speaker

Ashkan Haji Hosseinloo is a Ph.D. candidate at MIT in the department of Mechanical Engineering where he works with Konstantin Turitsyn on nonlinear vibration energy harvesting and data-driven climate control in buildings. Ashkan’s research is in the field of dynamical systems and controls with a focus on structural dynamics and vibration, energy harvesting, vehicle dynamics, and machine learning. He holds a Master’s degree in Mechanical Engineering from NTU of Singapore where he worked on a number of applied industrial projects in dynamics and vibration in collaboration with ST Kinetics, Defense Science Organization (DSO) and SMRT Corporation of Singapore. Ashkan is the recipient of a number of awards including Den Hartog Award in Mechanics, Martin Fellowship in Design, and Graduate Exploration Fellowship from MIT and ST Kinetics and DSO scholarships from Singapore.