C.R. (Bob) Koch /* Add links in reverse order to display on website due to CSS float:right formatting.*/ Publications Teaching Research

C.R. (Bob) Koch

  Professor  
  Mechanical Engineering, University of Alberta  
  Office: 10-232 Donadeo Innovation Centre for Engineering  
  Tel. (780) 492-8821 Fax: (780) 492-220
  Ph.D. (Stanford University, 1986-1991), P.Eng  
  Email: mailto:bob.koch@ualberta.ca  
  Google Scholar: Koch Google Scholar  
  ORCID ORCID-iD_icon-16x16.png https://orcid.org/0000-0002-6094-5933  

Research Interests

Control to energy systems with a focus on internal combustion engines, fuel cells and fluid systems. The applications of my research are control and diagnostics of H2/Diesel and H2 truck engines, Solid Oxide Fuel Cells, and active control of mixing and drag reduction of non-reacting fluid systems.

Key Words

Internal Combustion Engines, Control Systems, Machine Learning, Fluid Mechanics, Low Temperature Combustion, Exhaust Gas Sensors, Exhaust Gas Aftertreatment, Carbon Free Fuels, Biofuels, Mechatronics, H2/Diesel Dual Fuel, Model Based Diagnostics, Machine Learning Control, Physics Inspired Neural Networks

Motivation

One goal is to develop control methods that reduce CO2 in Internal Combustion Engines (ICE) powered transportation by optimizing the engine/aftertreatment system during dynamic operation and using a variety of low or zero carbon fuels. It is proposed to combine Model Predictive Control (MPC) and Machine Learning (ML) with domain knowledge to solve these complex constrained problems.

Develop a fundamental understanding and methods of combining MPC with ML to improve the performance of complex engineering problems is a longer term goal for the research.

Brief Description

Model Predictive Control (MPC) is a systematic way to provide the optimal solution for control problems while considering constraints. MPC has challenges, such as high computational cost and the requirement for high accuracy plant models and this is where there is an exciting opportunity to combine it with Machine Learning (ML).

Integrated high efficiency multi-mode combustion and aftertreament has emerged as a very promising technology to simultaneously reduce fuel consumption and harmful emissions for ICEs but requires control. To achieve high thermal efficiency and low emissions, advanced ICE combustion modes are used and these modes often require control during operation and when switching and the combustion/control must be co-optimized with the fuel.

The combustion timing of Homogeneous Charge Compression Ignition (HCCI) is controlled in realtime, cycle-by-cycle using solenoid intake/exhaust valves and has been controlled using MPC. Significant progress has also been made in controlling HCCI during the combustion where a physics based reaction kinetics combustion model (clear box) is reduced to a grey box model and run in realtime at 0.1 degree crank angle on an FPGA. Engine knock is avoided using water injection while variable valve timing and high energy spark are used to avoid misfire with this in-cycle control. This work will allow very novel control to enable use of a variety of bio-fuels and carbon free fuels such as H2.

Fast responding feedback sensors that are inexpensive and reliable are essential to achieve Real Driving Emissions (RDE). The focus of my work has been on understanding low cost electrochemical production sensors and using this understanding to: diagnosis sensor failures, understand NOx ammonia cross sensitivity, and to change the sensor operation point to measure HC emission. Using the existing NOx sensor to measure another species (HC or NH3) was possible through detailed understanding of the sensor and dynamic control of the sensor operation. This sensor knowledge is combined with ML in a grey-box model.

Control of distributed parameter systems, although challenging, is relevant to flow and combustion problems. Fundamental aspects of discrete time model based control is examined. Applications include: jacket tubular reactors where stability of model based control in a distributed setting is examined.

Progress has been made applying modeling and control of thin films, particle separation and drag reduction in flows.

Financial Support

EDI

  • My aim to provide the highest quality training within an inclusive, equitable, and supportive environment
  • In my job postings I encourage all qualified applicants and encourage applications from racialized persons/visible minorities, women, Indigenous persons, persons with disabilities, ethnic minorities, and persons of minority sexual orientations and gender identities

Job Postings

Research Assistant

  • Working Title: Hydrogen for Transportation
  • Position Type: Trust Research Academic Staff (TRAS)
  • Start Date: May 1, 2025 or when suitable candidate has been found
  • Length: 2 Years with possible option for 2 year extension

This job posting is for a Senior Researcher in the Department of Mechanical Engineering, focusing on low-carbon transportation research. The role requires expertise in the transportation industry and managing research facilities.

PhD

  • Faculty/Department: Engineering/Mechanical
  • Project Title: Physics-Informed Machine Learning Control for Hydrogen Maximization in Heavy-duty Engines
  • Position Type: Doctorial Student
  • Start Date: September 1, 2025 or when suitable candidate is found

This position is for a PhD student in the Mechanical Engineering Energy Control Lab (MEECL), focusing on integrating advanced machine learning (ML) networks for hydrogen internal combustion engine control. The research will explore Physics-Informed Neural Networks (PINNs) and Kolmogorov-Arnold Networks (KANs) to improve Model Predictive Control (MPC), enhancing prediction accuracy while reducing computational demands. The developed ML-based controllers will be experimentally validated in a state-of-the-art testing facility for heavy-duty transport engines.

Post-doc

  • Working Title: Machine Learning-based Control Development for Hydrogen Energy Conversion Systems
  • Position Type: Post Doctoral Fellow
  • Start Date: May 1, 2025 or when suitable candidate is found
  • Length: 2 Years with potential for 2 year extension

This position is for a Postdoctoral Fellow in the Mechanical Engineering Energy Control Lab (MEECL), focusing on developing Machine Learning Control (MLC) for zero-carbon energy conversion systems, specifically 100% hydrogen internal combustion engines (ICEs). The research will aim to enhance engine efficiency, durability, and low emissions, with potential future applications in fuel cells. The role involves experimental testing, physics-based modeling, and data-driven control strategies.

This position is part of the low-carbon transportation research team, working closely with graduate students, research associates, and principal investigators to develop cutting-edge control strategies for sustainable energy solutions.

Directions

Conferences/Links/News