Multi-objective optimization of virtual formula powertrain design for enhanced performance and fuel efficiency

Abstract

Formula 1 is considered an important motorsport event that has driven the development of cutting-edge automotive technology. Similarly, virtual formula competition is introduced at the university level to harness technological advances for automotive development among universities. Performance and fuel efficiency have always been two conflicting objectives in determining the optimal powertrain design. In this paper, state-of-the-art multi-objective evolutionary algorithms (MOEAs) are employed to optimize the vehicle powertrain design based on competing event and the ontrack performance through software in-the-loop optimization approach. Unlike conventional powertrain optimization approach where the powertrain design is optimized again standard urban driving profiles to achieve optimum driving performance and fuel consumption which is not suitable to be employed to optimize race car powertrain design. The powertrain design is optimized and benchmarked in multiple dynamic test events i.e. Acceleration, Autocross and Endurance. Each event measures different performance parameters (i.e. lap time and fuel consumption) A total of nine design parameters are optimized in this study. Three design parameters (employed to formulate the progressive gear calculation of six gear ratios), five gear shifting rpm (revolutions per minute) and an aerodynamic downforce scale value (for front and rear downforce). Two stages of optimization solution are proposed to handle such complicated design requirement. NSGA-II was identified as the most efficient algorithms among the MOEAs employed in this study; especially in terms of computational time, Pareto distribution and Pareto design. Overall, the optimized powertrain design showed significant improvement in all events i.e. energy efficiency was increased by 13.7% for the endurance event while lap time was reduced by 0.7 and 1.2% in the acceleration and autocross events, respectively.

Ensemble of multi-objective metaheuristics for multiprocessor scheduling in heterogeneous distributed systems: a novel success-proportionate learning-based system

“…The set of all non-dominated solutions extracted by the algorithm is known as optimal Pareto-front. A solution is considered as non-dominated solution if it is better than other solutions in at least one objective and it is not worse in terms of other objectives [3,4]. Figure 1 represents the optimal Pareto-front for a minimization problem consisting of seven non-dominated solutions for two objective functions F1 and F2.…”