Evaluation of machine learning techniques for the NVH design of electric vehicle battery

This paper explores the structural dynamic behavior of electric vehicle (EV) batteries and its impact on driver vibration comfort, especially at low frequencies. Being an additional bulky element of the EVs, the battery pack has a significant impact on the vehicle body dynamics. Yet, its design optimization for vibration purposes is challenging given the complexity of its stiffness distribution and coupling to the vehicle body. In this paper, a workflow employing machine learning (ML) techniques to assess battery NVH performance is described. The methodology starts by creating a conceptual finite element (FE) based parametric model of the battery pack. For efficient computation purposes, the battery model is coupled to a reduced order model of the vehicle body using an FE sub-structuring approach. This model forms the foundation for stablishing a database used to train ML algorithms to create a surrogate model for vibration response estimation. The test set is used to evaluate the accuracy of the model and the size of the training data required. It is considered the first step to enhance the design process of EV with advanced ML algorithms.