Scholarlink

<p>Detecting risks promptly and accurately is vital for safeguarding the safety and longevity of road and bridge structures. That said, conventional monitoring systems centered on cloud computing often run into issues with bandwidth constraints, delays, and data privacy, especially across scattered networks designed for structural health assessment. Here, we introduce Edge-TimeLSH, an edge computing framework that leverages a time-sensitive version of Locality-Sensitive Hashing (LSH) for immediate, on-location risk identification. Our technique transforms unprocessed sensor data from various sources into hash codes that incorporate temporal factors, facilitating fast lookups of analogous past structural conditions right on the edge hardware. We then assemble a matrix highlighting temporal similarities to pinpoint repeating patterns in mechanical performance, which in turn support a non-parametric forecasting of structural behavior, weighted by those similarities. The system forwards just condensed hash codes and critical notifications to the cloud, thereby upholding privacy while minimizing bandwidth use. Evaluations using real-world datasets reveal that Edge-TimeLSH delivers precise, understandable risk evaluations even on resource-limited edge devices. Overall, these outcomes highlight the framework’s potential as a feasible and adaptable approach for ongoing risk monitoring in road and bridge projects.</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

Edge intelligence for on-site risk detection in road and bridge engineering

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年2月7日

查看全文
  • Feng Xu,
  • Qin Wang,
  • Hossein Khosravi

摘要

Detecting risks promptly and accurately is vital for safeguarding the safety and longevity of road and bridge structures. That said, conventional monitoring systems centered on cloud computing often run into issues with bandwidth constraints, delays, and data privacy, especially across scattered networks designed for structural health assessment. Here, we introduce Edge-TimeLSH, an edge computing framework that leverages a time-sensitive version of Locality-Sensitive Hashing (LSH) for immediate, on-location risk identification. Our technique transforms unprocessed sensor data from various sources into hash codes that incorporate temporal factors, facilitating fast lookups of analogous past structural conditions right on the edge hardware. We then assemble a matrix highlighting temporal similarities to pinpoint repeating patterns in mechanical performance, which in turn support a non-parametric forecasting of structural behavior, weighted by those similarities. The system forwards just condensed hash codes and critical notifications to the cloud, thereby upholding privacy while minimizing bandwidth use. Evaluations using real-world datasets reveal that Edge-TimeLSH delivers precise, understandable risk evaluations even on resource-limited edge devices. Overall, these outcomes highlight the framework’s potential as a feasible and adaptable approach for ongoing risk monitoring in road and bridge projects.