This paper evaluates new optimization algorithms for optimizing automotive suspension systems employing stochastic methods. This method is introduced as an alternative over the conventional approach, namely trial and error, or design of experiment (DOE), to efficiently optimize the suspension system. Optimizations algorithms employed are the multi-objective evolutionary algorithms based on decomposition (MOEA\D), and non-sorting genetic algorithm II (NSGA-II). A two-degree-of-freedom (2-DOF) linear quarter vehicle model (QVM) traversing a random road profile is utilized to describe the ride dynamics. The road irregularity is assumed as a Gaussian random process and represented as a simple exponential power spectral density (PSD). The evaluated performance indices are the discomfort parameter (ACC), suspension working space (SWS) and dynamic tyre load (DTL). The optimised design variables are the suspension stiffness, K s and damping coefficient, C s. In this paper, both algorithms are analyzed with different sets of experiments to compare their computational efficiency. The results indicated that MOEA\D is computationally efficient in searching for Pareto solutions compared to NSGA-II, and showed reasonable improvement in ride comfort.

The conventional approach in vehicle suspension optimization based on the ride comfort and the handling performance requires decomposition of the multi-performance targets, followed by lengthy iteration processes. Suspension tuning is a time-consuming process, which often requires the benchmarking of competitors’ vehicles to define the performance targets of the desired vehicle by experimental techniques. Optimum targets are difficult to derive from benchmark vehicles as each vehicle has its own unique vehicle set-up. A new method is proposed to simplify this process and to reduce significantly the development process. These design objectives are formulated into a multi-objective optimization problem together with the suspension packaging dimensions as the design constraints. This is in order to produce a Pareto front of an optimized vehicle at the early stages of design. These objectives are minimized using a multi-objective optimization workflow, which involves a sampling technique, and a regularity-model-based multi-objective estimation of the distribution algorithm to solve greater than 100-dimensional spaces of the design parameters by the software-in-the-loop optimization process. The methodology showed promising results in optimizing a full-vehicle suspension design based on the ride comfort and the handling performance, in comparison with the conventional approach.

In an effort to reduce cost involving repetitive prototype build-test cycles, it is inevitable that simulation on full vehicle will be carried out during the product development stage. Desired suspension kinematic profiles of a given vehicle parameter are often unknown at the initial design stage. This paper demonstrates a simple methodology to obtain optimized kinematic characteristics against quality of handling performance using this model as predictive model in earliest design stage. A full vehicle model that is inclusive of suspension kinematic profiles and nonlinear damper profiles has been derived to enable the engineer to study the characteristics of the nonlinear elements against the vehicle performance when only limited vehicle data are available in the initial stage. Results suggest that the handling characteristics of a vehicle are sensitive to the changes in suspension kinematic profile. Additionally, the proposed vehicle model is able to provide satisfactory handling objective when measured in transient handling and frequency response compared to other vehicle models. A robust prediction model of the vehicle responses in frequency domain is proposed. It is coupled with the vehicle model employed as predictive model to optimize front toe angle profile against vehicle quality of handling performance measured in frequency domain. Keywords 10-degree-of-freedom full vehicle model, suspension kinematic profiles, design of experiment, vehicle handling Date

No abstract

No abstract