Abstract
KEYWORDS – Chassis Optimization, Quality Function, Weighting Factors, Objectification, Characteristic Curves
ABSTRACT
This study deals with chassis optimization using characteristic curves. To find an optimal configuration of a chassis, either it could be searched manually or automatically using an optimization tool with help of quality functions. Increasing number of parameters, non-linearity of the vehicle characteristics and mutual correlation of different chassis settings on characteristic values make it more challenging to find an optimal configuration manually. For this reason, under the assumptions that the lowest value of quality function corresponds to expected target values or expected tendency and the optimizer finds a global optimum, automated search would be preferred without any doubt. The problem lies on the fact that the lowest value of quality function does not always corresponds to expected optimization result. The optimization result is profoundly dependent on its form of quality functions and the weighting factors. In this study, four different form of sub-quality functions (\_ , _/ , \_/and \/) will be introduced and its gradients will be determined based on the perception sensitivity of a driver or a value from reference vehicle. Secondly, it will be dealt about how the weighting factors can be objectively determined, which enables to reach the target values by possible lowest quality function values. The weighting factors determined by subjective priority or importance might lead to undesirable results, in case where certain characteristic values are intensively conflicting to each other or if the probability of realization is not considered for a determination of the weighting factors. In general, the weighting factors are higher rated to reflect the importance or to improve the fulfillment of the characteristic values, but there are also some cases that are not able to be fulfilled at all, no matter what kind of chassis configurations are applied or how large the weighting factor is. These effects should be considered in determination of weighting factors by applying simulation results from Latin Hypercube sampling. An example of optimization results using a genetic algorithm based on double-track model with different weighting factors will be compared and discussed. As a final result, a methodology to determine the weighting factors will be proposed, which enables to produce a repeatable and desirable optimization results.