비지도학습을 이용한 브레이크 시스템 역설계

다국어 초록 (Multilingual Abstract)

It is well known that braking performances are target performances which must be considered for vehicle development. Apparent piston travel (APT) which is the distance piston travels when the driver pushing on the pedal. In this work, we propose a deep learning-based inverse design model for APT to reduce time and cost of iterative optimization methods. we can get the optimal design instantly, that satisfies APT performance. In particular, in order to have various optimal design values for a single performance target, an inverse designing technique using unsupervised learning such as Conditional GAN(CGAN) and Conditional VAE(CVAE), is presented. This approach will be extended to various vehicle systems in the future.