Advanced Machine Learning and Response Surface Methodology-Based Optimization of Diesel Engine Performance and Emissions Using 2-Ethylhexyl Nitrate in Mahua Methyl Ester Biodiesel Blends

Abstract This study presents a detailed experimental and numerical investigation of the performance, combustion, and emission characteristics of a Common Rail Direct Injection diesel engine fueled with a ternary blend comprising 30% BD30, 70% diesel, and 1000 ppm of 2-ethylhexyl nitrate (2EHN) as a cetane improver. The engine was evaluated under varying load conditions, compression ratios, and exhaust gas recirculation (EGR) levels. Response Surface Methodology (RSM) with Central Composite Design (CCD) was employed to develop predictive models and perform multi-objective optimization of critical responses including brake thermal efficiency (BTE), brake-specific fuel consumption (BSFC), brake power (BP), exhaust gas temperature (EGT), carbon monoxide (CO), and nitrogen oxides (NOx). The BD30/2EHN blend exhibited superior BTE and BP with a significant reduction in BSFC and CO emissions, attributed to improved ignition kinetics and enhanced combustion propagation. However, elevated NOx and EGT levels were observed due to increased in-cylinder temperatures. The quadratic models yielded high statistical accuracy with R2 values exceeding 0.96 for all responses. Machine learning–based validation identified Random Forest as the most robust predictive model. The optimized operating region was determined at 75% load, compression ratio of 17.6, and 5% EGR, balancing performance and emission tradeoffs. The findings affirm the feasibility of BD30/2EHN blends for high-efficiency, low-emission diesel engine applications.