Design and Experimental Validation of an Enhanced Fuel Management Strategy using ANN-Based Model Predictive Control for Biodiesel-Fuelled CI Engines

Abstract

This paper presents the design, implementation, and experimental validation of an Enhanced Fuel Management Strategy (EFMS) for compression-ignition (CI) engines operating on waste cooking oil (WCO) derived biodiesel blends. The EFMS integrates an Artificial Neural Network (ANN) surrogate-based Model Predictive Controller (MPC) with real-time common-rail direct injection (CRDI) adaptation, exhaust gas recirculation (EGR) scheduling, and closed-loop emission feedback to simultaneously maximise energy efficiency and minimise pollutant output. Experiments were conducted on a 4-cylinder, 2.5-litre turbocharged direct-injection CI engine over the four-load European Stationary Cycle (ESC) points using B0, B20, B40, and EFMS-controlled B20 fuel configurations [1,2]. The EFMS delivered a peak brake thermal efficiency (BTE) of 36.8% at 75% load, representing an improvement of 9.2% over the B0 baseline and 11.2% over unmanaged B20, while simultaneously achieving reductions in NOx of 17.6%, CO of 26.9%, and particulate matter (PM) of 37.2% relative to the B20 baseline [3,4]. Combustion analysis via in-cylinder pressure diagnostics confirmed a reduction in ignition delay from 7.9°CA (B20) to 6.8°CA under EFMS, with lower heat release rate peaks indicating smoother premixed combustion. The ANN surrogate model achieved prediction accuracy (R²) of 0.983–0.997 across all target variables, confirming its suitability for real-time deployment. These outcomes substantiate the EFMS as a high-impact, technically implementable strategy for sustainable CI engine operation [5,6].