A Local-Diffusion Genetic Algorithm for Disjoint Pareto-Optimal Problems With Application to Vehicle Suspension

This paper presents a multi-objective design optimization study of a vehicle suspension system with passive variable stiffness and active damping. Design of suspension systems is particularly challenging when the effective mass of the vehicle is subject to considerable variation during service. Perfectly maintaining the suspension performance under the variable load typically requires a controlled actuator in order to emulate variable stiffness. This is typically done through a hydraulic or pneumatic system, which can be too costly for small/medium pickup trucks. The system in this paper employs two springs with an offset to the second spring so that it engages during large deformation only, thereby providing passive variable stiffness without expensive hydraulics. The system damping is assumed to be controlled via variable viscosity magnetizable fluid, which can be implemented in a compact, low-power setup. Simulation of the suspension system is performed by numerically solving the nonlinear equations of motion for a quarter-vehicle mode subject to excitation from a road profile over a set period of time. A performance index from the literature is evaluated for the suspension system for the cases of minimum and maximum weight, and the two indices values are regarded as objectives in a multi-objective problem. As the individual objectives are prone to having local optima, the multi-objective problem is prone to having a disjointed Pareto-space. To deal with this issue, a modification is proposed to a multi-objective genetic algorithm. The algorithm performance is investigated via analytical test functions as well as a design case of the suspension system. Results show a reduction in the system’s spring size, with the Pareto point obtained from the proposed diffusion model, without compromising the system’s performance.Copyright © 2012 by ASME