Neural Model of a Spark-Ignition Engine Fuelled with Gasoline with a Biofuel Additive

Due to the depletion of global crude oil reserves and the unstable socio-political conditions in the regions with the highest production, there is a growing need to reduce oil consumption, more than 50% of which is associated with the transportation sector. One approach to decreasing oil demand is the incorporation of bio-additives into petroleum-derived fuels. However, modifications to fuel composition inevitably alter engine operating characteristics. Consequently, adjustments to the control system settings are required to maintain optimal engine performance. This study investigates the application of an artificial neural network to model the influence of bio-additive concentration on the operating parameters of a spark-ignition engine under varying control system configurations. Measurements were conducted over an engine speed range of 1200–6000 rpm. Key engine operating parameters were identified, relevant control settings requiring correction were selected, and experimental measurements were carried out on an engine test bench. The resulting dataset was evaluated for accuracy and statistical correlation. An artificial neural network model was subsequently designed, trained, and optimized using the dataset. For comparison, alternative models based on polynomial approximation and Partial Least Squares (PLS) were also constructed. The neural model achieved a coefficient of variation in relative root mean square error (CVRMSE) in the Leave-One-Condition-Out (LOCO) procedure of 7.04%, whereas the polynomial and PLS models exhibited errors of 8.52% and 7.39% accordingly. The findings of this study demonstrate the suitability and effectiveness of artificial intelligence methods for addressing the operational problem under investigation.