Scholarlink

Finding non-conformities, such as physical failures causing electrical malfunctioning of a device, in modern semiconductor devices is challenging. Highly qualified employees in a failure analysis (FA) lab typically use sophisticated and expensive tools like scanning electron microscopes to identify and locate such non-conformities. Given the increasing complexity of investigated devices and very limited resources, labs may struggle to deliver analysis results in time. This paper proposes an approach to optimize the usage of FA lab resources by combining constraint programming with stream reasoning enabling situation-dependent monitoring of the lab’s conditions and schedule maintenance. Evaluation results indicate that our system can significantly improve the tardiness of real-world FA labs, and all its computational tasks can be finished in an average time of 3.6 s, with a maximum of 15.2 s, which is acceptable for the lab’s workflows.

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

Monitoring and Scheduling of Semiconductor Failure Analysis Labs

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Elena Mastria,
  • Domenico Pagliaro,
  • Francesco Calimeri,
  • Simona Perri,
  • Martin Pleschberger,
  • Konstantin Schekotihin

摘要

Finding non-conformities, such as physical failures causing electrical malfunctioning of a device, in modern semiconductor devices is challenging. Highly qualified employees in a failure analysis (FA) lab typically use sophisticated and expensive tools like scanning electron microscopes to identify and locate such non-conformities. Given the increasing complexity of investigated devices and very limited resources, labs may struggle to deliver analysis results in time. This paper proposes an approach to optimize the usage of FA lab resources by combining constraint programming with stream reasoning enabling situation-dependent monitoring of the lab’s conditions and schedule maintenance. Evaluation results indicate that our system can significantly improve the tardiness of real-world FA labs, and all its computational tasks can be finished in an average time of 3.6 s, with a maximum of 15.2 s, which is acceptable for the lab’s workflows.