Design and design validation of vehicle alloy wheels using optimisation techniques and neural networks

There are various optimization methods developed for the solution of optimization problems in various disciplines. Especially in the automotive sector, the use of optimization methods in product development and design processes is having more importance day by day due to challenging targets and increasing competition. Many optimization methods are being developed and applied in order to shorten the vehicle design processes and to reach values closer to the global optimum design results with less resources. Natural frequency and weight values of aluminium alloy wheels are the parameters that directly affect the driving comfort of the vehicle. In the rim design process, preventing overlap of the rim's 1st mode natural frequency value with the frequency values of other vehicle parts directly affects the noise performance of the vehicle. Besides, reducing the rim weight improves the vehicle's parameters such as fuel consumption, emissions and road holding. In this context, with the proposed study, a new approach and artificial neural network (ANN) based parameter estimation method and parametric design optimization method were applied to improve the rim design process. The initial data set was obtained through finite element analysis with the experimental design model. The initial data set obtained was used in the training of multiple linear regression (MLR) and ANN models, and both trained models were applied as a surrogate model in the multi-objective genetic algorithm (MOGA) optimization process. Optimization results of 1st mode frequency and weight parameters of the optimization performed with ANN and MLR based estimation methods are revealed. When the proposed new method is compared with the well-known wheel design optimization methods in the literature, it is seen that more successful results are obtained in terms of target values and values closer to the global optimum results.