|
1 |
Hydrogen engine
Global warming is becoming increasingly serious, and regulations on carbon dioxide (CO₂) emissions are tightening worldwide. Spark-ignited engines, which primarily run on gasoline, are being developed with improved thermal efficiency to reduce CO₂ emissions. In addition to these advancements, researchers are exploring new combustion concepts and alternative fuels. Hydrogen is considered a promising fuel for stable combustion due to its excellent flammability and the fact that it produces no CO₂ emissions upon combustion.
In our lab, hydrogen combustion experiments are being conducted using a single-cylinder engine, which offers flexibility for hardware modifications.
Several studies are underway to investigate in-cylinder flow characteristics and ignition energy, both of which influence the stability of lean hydrogen combustion. Additionally, the effects of engine parameter variations—such as fuel injection method, compression ratio, and valve timing—are being analyzed with the aim of improving thermal efficiency.
 - Experimental setup of single-cylinder hydrogen direct injection engine
Dual-fuel engine
Hydrogen–diesel dual-fuel systems use the high ignitability of diesel to help control abnormal combustion of hydrogen, achieving both high efficiency and low emissions. The experiments are conducted on a single-cylinder engine, which allows flexible changes to hardware. Current research focuses on hydrogen–diesel reactivity controlled compression ignition (RCCI), dual direct injection, and related strategies.
One advantage of this setup is that it can run spark ignition, dual-fuel, and RCCI modes using the same engine. This is possible because the cylinder head is equipped with a spark plug, a diesel injector, and a hydrogen injector. This design allows direct comparison between different combustion strategies under consistent conditions. While low-reactivity fuels like gasoline are prone to knocking under certain conditions, hydrogen’s high auto-ignition temperature and high octane number make it more resistant to knocking, making it well-suited for advanced combustion modes.
 - Experimental setup of single-cylinder hydrogen-diesel dual-fuel engine
Application of reinforcement learning on engine calibration
Engine calibration and mapping are essential tasks to find optimal driving points of the engine. Various factors such as torque, efficiency, and emission, which are in trade-off relationships, must be calibrated. Also, as number of HEVs are increasing, powertrain system of vehicle is getting more complicated, making calibration tasks harder. To deal with this, concept of automating engine calibration and mapping is arising.
Reinforcement learning, a type of machine learning method that has learning process closest to human or animals, is one of the most powerful tools for control and optimization tasks. In this research, various reinforcement learning algorithms are tested for engine calibration and mapping automation. The training results showed that the model well calibrates an engine for single map case. In the future, this will be expanded to simultaneous calibration of several maps.
 - The structure of reinforcement learning model
|
