Surrogate-based modeling strategy for design optimization of passenger car suspension system

Abstract

The dynamic response of a Low-Fidelity (LoFi) vehicle model exhibits a discrepancy when compared to a High-Fidelity (HiFi) vehicle model. HiFi model construction involves complex state-space equations, a high degree of freedom, and requires a huge quantity of early data to completely define this model. This causes a delay and makes the computation process less efficient. On the other hand, the LoFi model developed using simpler state-space equations is faster and computationally cheaper. However, the response accuracy of this model is lower than that of HiFi. Due to this competence mismatch, it constrains the ability and integration of LoFi model or HiFi model applications in vehicle dynamics research. In previous researches, the proposed surrogate model has been completely replaced any physics-based model for subsequent engineering applications once it has been generated. However, this model has limitation to perform fine tuning either on LoFi or HiFi models. The primary aim of this research was to formulate a surrogate-based modeling strategy by tuning LoFi model for optimizing the design of the passenger car suspension system. The study began with the development of HiFi and LoFi models in Matlab, and their performances were verified by comparing the results produced by MSC Adams software. The LoFi model was used to determine the overall relationship between the suspension system's main elements, namely spring stiffness (Ks) and damper rate (Cs), and the design criteria, namely Body Acceleration (BAcc), Dynamic Tire Load (DTL), and Suspension Workspace (SWS). Based on the Design Criteria Space (DCS) map and recommendations from the literature, the Design Objective Space (DOS) map for a passenger car suspension system was established. Following that, three approaches to formulating surrogate models were introduced, namely the Response-Based Approach (RBA), the Variable-Based Approach (VBA), and the Parameter-Based Approach (PBA). The VBA for the Quadratic Transformation Scheme (QTS) was found to be the most suitable for the proposed newly surrogate model. Next, the surrogate model was linked to an optimization strategy to tune the suspension elements. Finally, a single optimal solution was obtained using the Min-Max method. The optimal tuning for the suspension elements of the chosen passenger car was Ks = 12535.6 N/m and Cs= 1416.7 Ns/m which increased the BAcc by 12.6% but at the expense of DTL performance by 6.4%, and keeping the SWS below the 7 mm restriction. In conclusion, the proposed surrogate-based modeling strategy could be a potential tool for optimizing the design of a passenger car suspension system.