Performance, emission, and combustion analysis of ternary lemon peel oil–mahua biodiesel–diesel blends in a YSZ-coated CI engine: Hybrid ML prediction, AGE-MOEA optimisation, and TOPSIS ranking

This paper investigates the effect of three lemon peel oil (LPO)-based ternary blends on combustion, emissions and performance characteristics of a CI engine using a Kirloskar TV1 direct-injection CI engine (5.2 kW, 1500 rpm, CR 17.5:1) with Mahua biodiesel and mineral diesel as test fuels: B1 (80D:10 M:10LPO), B2 (70D:15 M:15LPO) and B3 (60D:20 M:20LPO). The study is undertaken with both uncoated and yttria-stabilised zirconia thermal barrier-coated (YSZ-TBC) piston configurations, incorporating graphene oxide (GO) and cellulose nanocrystal (CNC) carbon-based organic nano-additives at concentrations of 25 and 50 ppm across five load conditions. To the best of the authors' knowledge, this study is to simultaneously integrate a ternary citrus–Mahua–diesel system with carbon-based organic nano-additive comparative evaluation, YSZ-TBC piston modification, hybrid transformer–boosting ML prediction, AGE-MOEA Pareto optimisation, and multi-scenario TOPSIS decision making within a unified experimental–computational framework. There were 130 experimental observations collected, and the machine learning training and comparisons of XGBoost and FT-Transformer models were assisted with SHAP explainability analysis. Six-objective Pareto optimisation was performed using the AGE-MOEA evolutionary algorithm with trained surrogates and TOPSIS to rank solutions over three different weighting scenarios. The results revealed that B3 mixed with GO at a dosage of 50 ppm for the TBC engine operating at 75% load gave rise to a brake thermal efficiency (BTE) of up to 35.99% which is 10.6% better than the clean diesel baseline, supplemented by cuts in hydrocarbons (66.7%), carbon monoxide (53.7%), and smoke (63.8%). The XGBoost exhibited R2 values above 0.998 for six of the seven targets, and FT-Transformer performed best on CO prediction (R2 = 0.956). Further, the TOPSIS analysis indicated that the state composed of B3, GO and TBC at a 75% load was identified as the highest-ranked condition across all three TOPSIS weighting scenarios (closeness coefficient = 0.814).

Actions (login required)

View Item