서강대학교

초록/요약

Recently, the development of electric vehicles is getting attention, because of the environmental and energy issues. In order to increase the energy efficiency of an electric vehicle, it is essential to reduce the weight of the vehicle and solve issues regarding vehicle noise and vibration. The NVH performance of the suspension module determines the quality of the ride comfort of the vehicle. Therefore, the frequency response analysis of a suspension module is essential for the vehicle design and analysis of NVH performance. In general, the finite element method has been used to analyze the vibration of the suspension module. In this study, we propose a new method that can quickly and accurately predict the frequency response data of a suspension module without using the finite element method. The purpose of this study is to develop an optimal prediction model that can predict the frequency response analysis data of the suspension module according to the change in bush dynamic stiffness by using an artificial neural network with a multi-layered perceptron structure. I propose the predictive model to achieve the best prediction performance through changes in learning data features, hidden layers, neurons, and activation function conditions constituting the artificial neural network with multi-layer perceptron structure. For the prediction validation, the accuracy of the predictive model is quantified through the Pearson correlation coefficient and RMSE. The Bush dynamic stiffness data set and the finite element analysis results are used to generate the training data set. As the verification method, k-fold cross-validation method is used. The validity of the prediction model is evaluated by determining whether the proposed artificial neural network model shows underfitting or overfitting phenomenon..