Scholarlink

<p>With the advancement of transistor technology reaching scales as small as 5 nm, thermal management of electronics remains a critical challenge, particularly with hotspot formation in microprocessors. These hotspots, resulting from uneven heat generation, vary in location and intensity depending on operational conditions. Managing these dynamic thermal patterns is essential to avoid performance degradation and potential damage due to excessive heating. Hence, this study employs an inverse analysis using the Kalman Filter (KF) to estimate spatially dependent heat sources in real time, allowing for precise hotspot detection. A two-dimensional unsteady diffusion model, solved with the finite volume method (FVM), simulates the temperature distribution along the lid of a personal computer’s microprocessor, capturing the dynamic behavior of the heat flux applied by the silicon dies. The KF adjusts heat source estimates based on transient temperature data measured using infrared imaging, enabling continuous updates to the thermal profile. In order to do that, the experimental setup features a thermal camera that records transient temperature profiles of the processor under full load conditions. Comparisons with synthetic data and experimental measurements demonstrate the effectiveness of the KF in tracking and estimating the applied heat flux in real time. Verification using synthetic data was performed with four heat flux profiles: two featuring single sources and two with composite sources. The inverse analysis with experimental data resulted in temperature residuals consistent with the standard deviation of the experimental measurements, confirming that the proposed method operates within acceptable error margins.</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

Experimental investigation and Kalman filter-based estimation of spatial-dependent heat sources in microprocessor hotspots

  • 期刊论文

  • 发表日期：

    2025年1月15日

查看全文
  • Vítor Fernandes Egger,
  • Mateus Zuma Medeiros Bria,
  • Kleber Marques Lisboa,
  • Krishina Dasa Alves Da Costa,
  • César Cunha Pacheco,
  • Isabela Florindo Pinheiro

摘要

With the advancement of transistor technology reaching scales as small as 5 nm, thermal management of electronics remains a critical challenge, particularly with hotspot formation in microprocessors. These hotspots, resulting from uneven heat generation, vary in location and intensity depending on operational conditions. Managing these dynamic thermal patterns is essential to avoid performance degradation and potential damage due to excessive heating. Hence, this study employs an inverse analysis using the Kalman Filter (KF) to estimate spatially dependent heat sources in real time, allowing for precise hotspot detection. A two-dimensional unsteady diffusion model, solved with the finite volume method (FVM), simulates the temperature distribution along the lid of a personal computer’s microprocessor, capturing the dynamic behavior of the heat flux applied by the silicon dies. The KF adjusts heat source estimates based on transient temperature data measured using infrared imaging, enabling continuous updates to the thermal profile. In order to do that, the experimental setup features a thermal camera that records transient temperature profiles of the processor under full load conditions. Comparisons with synthetic data and experimental measurements demonstrate the effectiveness of the KF in tracking and estimating the applied heat flux in real time. Verification using synthetic data was performed with four heat flux profiles: two featuring single sources and two with composite sources. The inverse analysis with experimental data resulted in temperature residuals consistent with the standard deviation of the experimental measurements, confirming that the proposed method operates within acceptable error margins.