Optimal Multi-Objective Elasto-Kinematic Design of a Suspension System by Means of Artificial Intelligence

Abstract In this paper, two multi-objective optimization methods are presented. These methods address the optimal design of multi-body systems, whose complexity arises from the high number of design variables and objective functions involved. k-optimality metric, which introduces a hierarchical sorting of the Paretooptimal design solutions, is adopted as additional ranking. The two methods differ in the exploration of the design space. The first method generates new suspension configurations within spherical neighborhoods around a reference one, iteratively updated to converge towards regions where optimized solutions are located. The second method employs a Deep Deterministic Policy Gradient (DDPG) algorithm to explore the design space and identify configurations that improve the considered optimality metric. These methods were applied to a MacPherson suspension – acting on hard points locations and rubber bushings characteristics – demonstrating a clear advantage over Parameter Space Investigation method in achieving solutions with higher optimality levels with comparable computational effort.