期刊
STRUCTURE AND INFRASTRUCTURE ENGINEERING
卷 13, 期 5, 页码 604-624出版社
TAYLOR & FRANCIS LTD
DOI: 10.1080/15732479.2016.1171883
关键词
Wireless sensor network; bridges; structural monitoring; machine learning; classification; Gaussian process regression; anomaly detection; statistical process control; solar-powered
资金
- National Institute of Standards and Technology (NIST) and Technology Innovation Program (TIP) [70NANB9H9008]
- National Science Foundation [CCF-0910765, CMMI-1362513, ECCS-1446521]
- US Department of Transportation [OASRTRS-14-H-MICH]
- Div Of Electrical, Commun & Cyber Sys
- Directorate For Engineering [1446521] Funding Source: National Science Foundation
The purpose of this study is to advance wireless sensing technology for permanent installation in operational highway bridges for long-term automated health assessment. The work advances the design of a solar-powered wireless sensor network architecture that can be permanently deployed in harsh winter climates where limited solar energy and cold temperatures are normal operational conditions. To demonstrate the performance of the solar-powered wireless sensor network, it is installed on the multi-steel girder bridge carrying northbound I-275 traffic over Telegraph Road (Monroe, Michigan) in 2011; a unique design feature of the bridge is the use of pin and hanger connections to support the bridge main span. A dense network of strain gauges, accelerometers and thermometers are installed to acquire bridge responses of interest to the bridge manager including responses that would be affected by long-term bridge deterioration. The wireless monitoring system collects sensor data on a daily schedule and communicates the data to the Internet where it is stored in a curated data repository. Bridge response data in the repository are autonomously processed to extract truck load events using machine learning, compensate for environmental variations using nonlinear regression and to quantitatively assess anomalous bridge performance using statistical process control.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据