Machine learning based hydrogen fuel approach: A detailed experimental study on CRDI engine performance, combustion, and environmental characteristics

Hydrogen stands out as a promising candidate among alternative energy sources, both as a fuel additive and/or an energy carrier. It is anticipated to become a primary alternative fuel, particularly to comply with rigorous environmental regulations, thanks to its carbon-free composition. Its application spans internal combustion engines, gas turbines, and the aerospace sector, owing to its non-toxicity, odorlessness, high energy content, and wide combustion temperature range. Additionally, hydrogen is recognized as an eco-friendly and sustainable energy source. The study investigates the impact of different hydrogen levels on combustion, efficiency, and emissions in a dual-fuel diesel engine, with tests at the various engine speeds of 1500, 2000, and 2500 rpm, and at the various loading conditions of 5–25 Nm with the intervals of 5 Nm covering various operational conditions. The introduction of hydrogen was varied, with flow rates set between 20.1, 27.3, 34.5, 41.6, and 49.2 lpm, tailored to match each specific load condition. Moreover, the paper also models the various performances of an Artificial Neural Network (ANN). Network performance is evaluated by calculating the Mean Absolute Percentage Error (MAPE) between simulated and actual outputs. The MAPE value for the best network is calculated to be 4.55 %, 1.49 %, 1.53 %, and 15.47 % for BSEC, BTE, Energy consumption, and NOx, respectively. The results indicate that hydrogen flow rates of 20.1 and 41.6 lpm significantly influence the engine's coefficient of variation and overall performance. Additionally, these flow rates notably enhanced CO, CO2, and smoke emission levels. Brake thermal efficiency significantly increased at a hydrogen flow rate of 20.1 lpm due to reduced combustion time and optimized combustion phasing. This underscores the vital role of hydrogen as a green energy source for the future.