Development and validation of an artificial intelligence platform for characterization of the exergy-emission-stability profiles of the PPCI-RCCI regimes in a diesel-methanol operation under varying injection phasing strategies: A Gene Expression Programming approach

Abstract

The present study leverages Gene Expression Programming as the AI platform to characterize the performance-emission-stability profiles in the parametric interaction realms of PPCI-RCCI and its interspersed regimes of a diesel engine operating at a constant speed of 1500 rpm (±2%) under varying injection phasing and methanol-induced reactivity strategies. Exergy efficiency, NOx, UHC, Opacity(%) and COVimep were chosen as the responses to be emulated with Overall reactivity, pilot-mass percentage, pilot-injection angle and main injection angle as the input parameters. The competency of GEP technique is examined through diverse correlation, absolute and relative error-based goodness-of-fit metrics. The quality of system information gained by the meta-models was evaluated through Theil Uncertainty(II)and Kullback-Leibler divergence metric, while generalization competency was evaluated across a test dataset, kept blind to the training phase. Analysis demonstrated significantly low footprints of error metrics along with negligible to low information loss across the entire response spectra emulated during both the training and testing phases. The GEP technique manifested commendable precision and reliability in adaptation to the actual experimental domain knowledge which endows it with a promising potential for deployment in model-in-loop based EMS frameworks for real-time decision-making in the complex control paradigms of such emerging engine combustion strategies.