人工智能在高速电路电磁兼容中的应用进展  

Review of Artificial Intelligence Applied to Electromagnetic Compatibility for High-speed Electronics

作　　者：刘恩校 Liu En-Xiao

机构地区：[1]新加坡高性能计算研究院

出　　处：《安全与电磁兼容》2024年第5期22-34,43,共14页Safety & EMC

摘　　要：人工智能及机器学习与电磁兼容及信号电源完整性呈现互为促进的关系。一方面,电磁兼容为人工智能硬件系统提供了保障;另一方面,人工智能技术为解决电磁兼容问题提供了新手段、新思路。文章首先对人工智能及机器学习进行了简要梳理,回顾发展简史,介绍基本组成要素、经验定律,总结主要的模型算法及架构等。然后,综述了机器学习特别是深度学习在解决电磁兼容及信号电源完整性问题中的应用及进展,涵盖替代模型、数值计算、逆问题、优化及生成式设计等方面。最后,展望了所面临的挑战与机遇。Artificial intelligence(AI)including machine learning(ML)benefits from electromagnetic compatibility(EMC)including signal and power integrity(SI/PI),and vice versa.On one hand,EMC/SI/PI ensure the compatibility and reliability of the AI/ML hardware;on the other hand,AI/ML offers potential new methodology for solving EMC/SI/PI issues.The first part of this paper presents a concise overview of AI/ML,covering its history,basic elements and laws,and a summary of major algorithms and architectures.Built on such foundation knowledge,it proceeds to survey the application of AI/ML including deep learning to addressing EMC/SI/PI problems,with an emphasis on surrogate models,computational algorithms,inverse problem,optimization and generative design.Finally,an outlook is presented about open issues,challenges and opportunities.

关 键 词：人工智能 机器学习 深度学习 电磁兼容 信号电源完整性 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TN03[自动化与计算机技术—控制科学与工程]