摘要详情

ID / 提交时间

98 / 2023-09-19 17:06:28

标题

The output characteristic model of SiC DSRD is established by BP neural network

关键字

pulsed power device, DSRD, neural network, efficiency

主题及专题

Special Sessions > SS06 - Artificial Intelligence for Semiconductor Modelling and Applications

状态

终稿

作者

Ruiheng Tian / Xidian University

Xiao-Yan Tang / Xidian University

Jingkai Guo / Xidian University

Lejia Sun / Xidian University

ZHANG Yuming / Xidian University

摘要

Abstract—Drift step recovery diode (DSRD) is a kind of highly nonlinear pulse power devices that exhibits complex inner physics. Traditional modeling approaches often fail to provide accurate and efficient models due to the complexity of its internal mechanisms. In this report, a neural network is employed to establish the relationship between the output characteristics and its trigger condition. By introducing the feature points on the output curve, a prediction model for the output features is established. This approach significantly improves modeling efficiency in engineering applications, circumventing issues such as time-consuming and computationally intensive processes encountered in traditional modeling. The neural model is been established which keeps the average error below 5%.