Dissertation

Structural Health Monitoring Meets Data Mining

Promotor: Prof.dr. J.N. Kok, Co-promotor: Dr. A.J. Knobbe

Author: S. Miao
Date: 16 December 2014
Links: Thesis in Leiden Repository

With the development of sensing and data processing techniques, monitoring physical systems in the field with a sensor network is becoming a feasible option for many domains. Such monitoring systems are referred to as Structural Health Monitoring (SHM) systems. By definition, SHM is the process of implementing a damage detection and characterisation strategy for engineering structures, which involves data collection, damage-sensitive feature extraction and statistical analysis. Most of the SHM process can be addressed by techniques from the Data Mining domain, so I conduct this research by combining these two fields. The monitoring system employed in this research is a sensor network installed on a Dutch highway bridge, which aims to monitor dynamic health aspects of the bridge and its long-term degradation. I have explored the specific focus of each sensor type under multiple scales, and analysed the dependencies between sensor types. Based on landmarks and constraints, I have proposed a novel predefined pattern detection method to select traffic events for modal analysis. I have analysed the influence of temperature and traffic mass on natural frequencies, and verified that natural frequencies decrease with temperature increases, but the influence of traffic mass is weaker than that of temperature.