A novel sub-objective progressive trade-off optimization method and application in the preliminary design of opposed-piston two-stroke diesel generator sets

The preliminary design of complex systems involves multiple tightly coupled sub-objectives and subsystems, which complicates the coordination required for achieving a comprehensive optimum. Traditional methods, such as the MultiDisciplinary Feasible (MDF) approach, directly optimize the system-level problem but often encounter multiple local optima. Avoiding these local optima requires extensive global exploration, which inevitably increases computational time. To enhance optimization accuracy while reducing computational time, a novel “Sub-Objective Progressive Trade-off (OPT)” optimization method is proposed and applied to optimize power generation fuel consumption rate of an opposed-piston two-stroke diesel generator set. It decomposes the system-level optimization into three stages, where three sub-objectives—indicated thermal efficiency, mechanical efficiency, and power generation efficiency—are progressively involved in the trade-off in sequence. Local optima caused by one sub-objective’s optimality can be avoided sequentially. The results show that, compared to the MDF method using a simulation model, the OPT method reduces the optimal fuel consumption rate by 0.34 g/kWh and computational time by 67%. When compared to the MDF method using a surrogate model, the OPT method reduces the optimal fuel consumption rate by 1.28 g/kWh and computational time by 65.5%. Moreover, regardless of initial solution differences, all three efficiencies corresponding to each optimal solution from the OPT method remain nearly consistent, verifying that local optima caused by a relatively high efficiency are avoided. Computational time statistics indicate that progressive optimization and variable allocation reduce the time consumption of the multi-stage optimization process. This study presents an efficient optimization method for the preliminary design of complex systems.