Designable Data Augmentation-based Domain-adaptive Design of Electric Vehicle Considering Dynamic Responses

Abstract

To address environmental pollution problems, electric vehicles (EVs) are attracting attention as future mobility vehicles. However, an increase in the number of advanced systems coupled with such vehicles imposes a limit on the development of EVs. The conventional design methods require a large amount of experimental and simulation data to satisfy the target performance of the system. Therefore, it takes time to arrive at the desired design solution. Hence, we propose a new design method using domain-adaptive designable data augmentation (DADDA). DADDA is a deep learning-based generative model that applies an inverse generator and domain adaptation concept to the data augmentation algorithm. This model aims to rapidly provide a design solution for a new system with a performance level similar to that of the existing system by adapting the domain of the existing system when the design information for a new system is insufficient.