Machine learning enabled real time adaptive fuel injection control for optimized engine performance

The current study investigates the effects of hydrogen content, di-tert-butyl peroxide (DTBP), and algae residue carbon nanoparticles (ACN) in a compression ignition engine fueled with a 30 % Pongamia biodiesel blend (PBD30). A series of test conducted across wide range of load conditions varying between 25 % and 100 %. Hydrogen gas injected via air intake manifold at the concentration of 10 and 20 L per minute (LPM). Based on the series of test we found an optimized fuel system with PBD30, 1.5 % DTBP, 3 % ACN and 20 LPM of hydrogen. An above blend increased the brake thermal efficiency from 25.06 % to 32.0 % and brake specific fuel consumption decreased by 8.2 % at full load than neat diesel. Combustion analysis revealed significant enhancements in ignition characteristics, with ignition delay reduced from 12.06° to 9.48° crank angle (CA). Meanwhile the combustion duration shortening from 37.88° to 30.75° CA. This clearly implies the superior heat transfer by the fuel combination compared to neat diesel. While exhaust gas temperatures increased from 327.3 °C to 375 °C due to the intensified combustion. With regard to the emissions, carbon monoxide decreased by 54.7 %, hydrocarbons by 38.6 %, and smoke opacity by 37.6 % at full load. The multi-additive method demonstrated a superior effectiveness compared to individual blend. When hydrogen added with the biodiesel blends with the presence of DTBP and ACN the overall ignition was massively improved owing to the ACN catalytic activity and DTBP energy carrying capacity. Hence optimizing these based on the loading condition was key to attain an economic viability of this blend. The adaptive fuel control technology was modelled and simulated via machine learning model. Based on the simulation it is clear that the effectiveness of the engine can be widened further by having an adaptive control of fuel injection compared to convention injection model. The machine learning modelled was performed for each load and its respective optimized fuel blend was proposed.