Machine learning-assisted optimization of thermodynamics and combustion in hydrogen argon power cycle engines

The Hydrogen Argon Power Cycle (HAPC) presents a promising route towards ultra-high thermal efficiency and net-zero emissions in internal combustion engines. This study develops a comprehensive one-dimensional (1D) modeling and optimization framework using GT-SUITE, targeting efficiency maximization through a combination of physics-based simulation and data-driven machine learning (ML) techniques. Key engine parameters, compression ratio (CR = 4-20), argon mole fraction (Ar% = 78-94), and equivalence ratio (ER = 0.4-1.2) are systematically explored via a Design of Experiments (DoE) approach. Surrogate models based on Kriging and ML algorithms are trained on the simulation results to enable efficient prediction and global optimization of indicated thermal efficiency (ITE). Three fuel injection strategies, including a premixed combustion mode with port hydrogen injection, hydrogen direct injection (DI), and DI combined water co-injection were comprehensively investigated. These strategies are assessed for improving combustion phasing, suppressing heat losses, and expanding the high-efficiency operating envelope. The results highlight critical trade-offs among thermodynamic parameters, injection strategies, and heat transfer behavior, while underscoring the effectiveness of surrogate modeling in accelerating the design and optimization of HAPC engines. Notably, the combination of DI with water co-injection achieves a peak ITE of 60.2%, following with the DI-only strategy (54.6%) and the premixed hydrogen combustion mode (53.2%). This demonstrates the synergistic benefit of thermal management and combustion control enabled by dual-injection strategies in argon-diluted environments.