Daniel Egan Email & Phone Number

Southfield, Mi, Us

Southfield, Mi, Us

Novi, Michigan, United States

Greenville, South Carolina Area

Accelerating the Derivation of Optimal Powertrain Control Strategies Using Reinforcement Learning and Virtual PrototypesModern powertrain complexity has reached a threshold where classical control and calibration techniques are no longer feasible. In place of classic control techniques deep reinforcement learning (DRL) can be used to generate a control strategy. DRL agents can bypass the curse of dimensionality and learn from complex models directly with minimal human oversight. DRL can accelerate our understanding of the performance potential of advanced powertrain concepts and the complexity of controlling them. Research includes:• Development of a DRL-based framework to create rapid and fair comparisons of novel powertrain concepts for hybrid electric vehicles. Designed to accept any set of constraints (thermal or mechanical) and control the system without constraint violation. Framework can handle mixed discrete-continuous action spaces.• Use of DRL to control a organic Rankine cycle (ORC) waste heat recovery system (WHR) for class 8 vehicles over entire drive cycles. Control oriented moving boundary models of ORC systems break down when all three phases (liquid, mixed phases, and superheated gas) are not present within the WHR boiler. DRL allows for control strategies to be learned from high fidelity finite volume models with the trained agents capable of running in real-time. • Enabling real-time non-linear model predictive control (nMPC) using artificial neural networks (ANNs) with modern internal combustion engines. nMPC linearization and computation time can be reduced by modeling complex dynamics within artificial neural networks.

Terre Haute, Indiana Area

- Oversees day-to-day activities, relations, and finances - Teaches and advises on manufacturing methods and techniques- Leads design of unsprung mass components (SolidWorks)- Utilizes Finite Element Analysis to verify designs (SolidWorks & Ansys)

Auburn Hills

- Researched current industry knock metrics and investigated relationships between them- Analyzed the feasibility of new knock metrics and new knock threshold determination methods- Designed and conducted test plans to gather knock information on a variety of engines- Coordinated research efforts with a global team of engineers

Indianapolis, Indiana Area

- Validated and submitted new calibration values for upcoming software release- Conducted software validation test plans of new software features in development vehicles and on simulators- Traced several software calculation values to their root cause and recommended solutions - Interpreted transmission controller software to update test plans to current format specification

Terre Haute, Indiana Area

- Distinguished Intern Nominee- Developed machine setups and procedures for the manufacture of prototype parts- Advised interns on how to design for parts for manufacturing- Manufactured one-off parts for prototype designs- Communicated design ideas and processes to interns

Doctor Of Philosophy - Phd, Automotive Engineering

Dissertation: Accelerating the Derivation of Optimal Powertrain Control Strategies Using Reinforcement Learning and Virtual Prototypes

Master Of Science - Ms, Automotive Engineering

Bachelor Of Science (Bs), Mechanical Engineering