Abstract
In this paper, we describe a strategy for modeling the feasible set of power-split continuously variable transmission (CVT) system designs for retrofitting rear-wheel-drive consumer automobiles. A design is considered feasible if it produces a higher fuel economy than the stock vehicle’s fuel economy rating. Towards modeling the feasible set of designs, we first model a vehicle with a power-split CVT system taking into account the system’s mass and CVT efficiency. In this model, the effective design variables are the mass of the transmission system, the CVT functional efficiency, and effective gear ratio defining the allowable power split through and around the CVT. We formulate the set of feasible design solutions utilizing a multiobjective optimization problem to define the boundaries of the maximum allowable system mass, minimum allowable efficiency, and minimum allowable effective gear ratio. We solve this multiobjective optimization problem using NSGA-II and fit a quadratic model to the NSGA-II results to define a surrogate model of the feasible design set. We show that this surrogate modeling approach is sufficient for predicting the feasibility of a candidate transmission design.
Approximate dual and approximate vector variational inequality for multiobjective optimization
