Session: 04-01: Powertrain Diagnostics and Control Strategies

Paper Number: 141703

141703 - A Generalized Additive Modeling Framework Approach for RCCI Combustion Phasing Prediction 

Abstract: 

Reactivity Controlled Compression Ignition (RCCI) represents a promising combustion strategy in internal combustion (IC) engines, offering high efficiency alongside remarkably low nitrogen oxides (NOx) and soot emissions. The distinctive dual-fuel approach of RCCI demands precise control over combustion timing, necessitating accurate prediction of combustion phasing to meet stringent emissions standards while optimizing engine performance. While various data-driven and machine learning (ML) techniques have emerged to address the complexities of RCCI combustion control, some may lack reliability in the face of uncertainties due to limited insights into their underlying models. Integrating computational fluid dynamics (CFD) and other phenomenological models may add complexity and computational burden, rendering RCCI combustion control approaches impractical. Achieving an optimal balance between model complexity, prediction accuracy, and computational cost is crucial for cycle-by-cycle control on advanced RCCI engine platforms and diverse fuel types. In this study, we propose a novel approach utilizing Generalized Additive Model (GAM) for transparent modeling of CA50 (the crank angle at which 50% of the fuel energy is released) in RCCI engines, facilitating reliable prediction for precise combustion phasing control. The GAM framework accommodates complex nonlinear relationships and interactions among key parameters without extensive prior knowledge of underlying physical and chemical processes. Additionally, the GAM structure offers transparency and interpretability, aiding in engine design understanding and ensuring reliable RCCI combustion control without significant computational overhead. The proposed GAM model for CA50 prediction is validated with experimental data across various steady-state engine operating conditions in Gasoline/Diesel (G/D) and 85%Ethanol (E85)/20%Biodiesel (B20) RCCI regimes, demonstrating high predictive accuracy greater than 90% Pearson Product-Moment Correlation Coefficient (PPMCC) and fast computational runtime of less than 500 milliseconds (ms) per query. Thus, the GAM model holds great potential for real-time model predictive control (MPC) in RCCI applications.

Presenting Author: King Ankobea-Ansah Illinois Institute of Technology

Presenting Author Biography: King Ankobea-Ansah is a PhD candidate at the Illinois Institute of Technology- Chicago. His research is focused on the modeling and control of fuel-flexible multi-cylinder internal combustion engines. Specifically, on developing control methodologies for advanced combustion strategies like the Reactivity Controlled Compression Ignition (RCCI); and on the analysis & control of Dimethyl Ether (DME) and Propane blends-fuelled compression ignition engines. He is recipient of the 2023 SAE/Yanmar scholarship award for his work on improving diesel engines. He is a member of the ASME Automotive and Transportation Systems Technical Committee, the National Society of Black Engineers (NSBE), the Society of Automotive Engineers (SAE), and the Order of the Engineer. Ankobea-Ansah has won numerous awards for his transformative leadership and service including the Illinois Tech 2023 Clinton E. Stryker award.

Authors:

King Ankobea-Ansah Illinois Institute of Technology
Carrie Hall Illinois Institute of Technology