Scholarlink

<p>In aerospace applications, the demand for highly reliable systems necessitates advanced dynamic fault mitigation strategies to achieve the trade-off between system reliability, performance, and power efficiency in dynamic space radiation environments. However, traditional fault mitigation methods fall short in simulating space radiation dynamics, particularly during solar events, thus hindering the effective optimization of dynamic systems. This paper presents a novel methodology that integrates space radiation-driven fault injection with machine learning-enhanced dynamic Parity per Byte with Duplication (PBD) fault mitigation strategies. The methodology utilizes historical space radiation data, creating time-series fault datasets for precise simulation of dynamic space environments. These datasets are integrated into fault injection platforms and a fault prediction model training framework to assess the effectiveness of dynamic error correction strategies. The proposed approach was validated using several historical solar events on dynamic PBD-based RAM scrubbing. The results demonstrate the dynamic PBD method’s effectiveness in mitigating fault accumulation during SPEs, significantly enhancing RAM scrubbing performance by up to 13 times compared to five traditional static methods.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Dynamic Fault Mitigation for Space Radiation Using Fault Injection and Machine Learning

  • 期刊论文

  • 发表日期：

    2025年6月12日

查看全文
  • Junchao Chen,
  • Li Lu,
  • Marko Andjelkovic,
  • Fabian Luis Vargas,
  • Milos Krstic

摘要

In aerospace applications, the demand for highly reliable systems necessitates advanced dynamic fault mitigation strategies to achieve the trade-off between system reliability, performance, and power efficiency in dynamic space radiation environments. However, traditional fault mitigation methods fall short in simulating space radiation dynamics, particularly during solar events, thus hindering the effective optimization of dynamic systems. This paper presents a novel methodology that integrates space radiation-driven fault injection with machine learning-enhanced dynamic Parity per Byte with Duplication (PBD) fault mitigation strategies. The methodology utilizes historical space radiation data, creating time-series fault datasets for precise simulation of dynamic space environments. These datasets are integrated into fault injection platforms and a fault prediction model training framework to assess the effectiveness of dynamic error correction strategies. The proposed approach was validated using several historical solar events on dynamic PBD-based RAM scrubbing. The results demonstrate the dynamic PBD method’s effectiveness in mitigating fault accumulation during SPEs, significantly enhancing RAM scrubbing performance by up to 13 times compared to five traditional static methods.