1: \begin{abstract}

2: Traditional TCAD simulation has succeeded in predicting and optimizing the device performance; however, it still faces a massive challenge — a high computational cost. There have been many attempts to replace TCAD with deep learning, but it has not yet been completely replaced. This paper presents a novel algorithm restructuring the traditional TCAD system. The proposed algorithmpredicts three-dimensional (3-D) TCAD simulation in real-time while capturing a variance, enables deep learning and TCAD to complement each other, and fully resolves convergence errors.

3: \end{abstract}

4: