【作者】 何佳俊； 王仕斌； 徐长卿； 刘毅； 廖新芳； 沈佳先； 杨银堂；

【Author】 HE Jiajun;WANG Shibin;XU Changqing;LIU Yi;LIAO Xinfang;SHEN Jiaxian;YANG Yintang;Xidian University;Xidian University Guangzhou Institute of Technology;

【机构】 西安电子科技大学； 西安电子科技大学广州研究院；

【摘要】 本文针对宇航集成电路传统加固方法在性能、功耗、面积与抗辐射能力间的权衡局限,提出系统性解决方案。主要贡献如下:首先,建立融合性能、功耗、面积和软错误率的PPRA多目标优化模型,构建抗辐射设计的量化评估框架,支持设计者高效选择最优方案,并在闭环优化过程中自适应调整算法参数,提升抗辐射设计的综合性能和自动化程度;其次,设计BP-SPEA2智能优化方法,通过路径优化筛选关键触发器,结合SPEA2进化算法与BP神经网络实现多目标动态优化——前者基于Pareto支配关系维护高收敛性解集,后者加速功耗、面积、软错误率的联合评估,显著提升优化效率;最后,实验验证表明,该方法在保障电路性能的前提下可实现综合优化,软错误率降低75.3%,突破了传统单维度加固的性能瓶颈。本研究为高可靠宇航集成电路提供了兼具智能性与工程实用性的抗辐射设计方法。更多还原

【Abstract】 In this paper,we propose a systematic solution to address the inherent limitations of traditional radiation hardening methods in balancing performance,power consumption,area, and radiation tolerance in aerospace integrated circuits.The main contributions of this paper are summarized as:First,we establish a PPRA multi-objective optimization model integrating Performance,Power,Area,and Soft Error Rate,and,introduce a quantitative evaluation framework for radiation-hardened design.This model enables efficient solution selection and enhances optimization adaptability through closed-loop parameter adjustment.Second,we design a BP-SPEA2 intelligent optimization method,which is combined with path optimization and sensitivity analysis to identify critical flip-flops.This approach synergizes the SPEA2 evolutionary algorithm with BP neural networks to boost optimization efficiency.Experimental results demonstrate that our methods achieve a 75.3% SER reduction while maintaining circuit performance,thereby overcoming traditional single-objective hardening limitations.The proposed method offers an intelligent and practical solution for high-reliability aerospace IC radiation hardening design.更多还原