Application of machine learning for performance prediction and optimization of a homogeneous charge compression ignited engine operated using biofuel-gasoline blends

Homogeneous Charge Compression Ignition (HCCI) engine is a prospective technology that effectively utilizes net carbon-neutral biofuels to achieve ultra-low nitrogen oxides (NOx) and smoke emissions with minimized indicated specific energy consumption (ISEC). The HCCI engine characteristics vary significantly with the biofuel type, requiring an investigation for the most suitable fuel composition and properties for enhanced fuel economy and lower exhaust emissions. In the present investigation, the critical fuel parameters of molecular weight, hydrogen-to-oxygen ratio, carbon-to-oxygen ratio, research octane number, energy content, and cooling potential were chosen based on their impact on the HCCI combustion load range limits. The support vector machine (SVM) regression models were developed to predict the HCCI engine characteristics, including combustion phasing, ISEC, and regulated emissions, using the above-mentioned critical fuel parameters as inputs. Experiment data obtained by running the HCCI engine using seven biofuels blended in gasoline consisted of 147 data points that trained the SVM models. The chosen biofuels belonged to distinct oxygenated organic functional groups of alcohol, ester, ether, and ketone. The relative importance of each key fuel parameter in predicting the investigated engine parameters was estimated. The robust SVM models were used in multi-objective Pareto-search optimization to find Pareto optimal solutions. TOPSIS was used to select the best alternative among 70 Pareto-optimal solutions for minimum ISEC and regulated emissions. The optimized fuel composition in the HCCI mode reduced the ISEC by 18% and NOx by 76% than the conventional diesel combustion mode. Also, the HCCI mode produced almost zero smoke emissions (< 0.0007 g/kW-h). The optimal fuel parameters could be achieved by tuning the biofuel proportions in gasoline. The present work demonstrated that using optimized fuel composition for HCCI combustion could tackle the significant performance and emission drawbacks of conventional light-duty diesel engines. •A broad HCCI engine experiment database was generated using seven biofuels.•The dynamic fuel composition variation extended the HCCI engine load range.•Support vector machine regression was used to model the HCCI engine parameters.•Pareto-search and TOPSIS ensemble were used for HCCI engine optimization.•The fuel parameter values for optimal HCCI engine performance were established.