Data-driven methods for structural damage detection
Speaker: Anne Kiremidjian, Stanford University
Series: CEE Departmental Seminars
Location: Bowen Hall Auditorium
Date/Time: Thursday, January 16, 2014, 4:30 p.m. - 6:00 p.m.
Structural health monitoring relies on the deployment of instruments and interpretation of the data collected by the instruments to predict the state of the structure. Majority of techniques use the data to verify the structural model and then predict changes in structural parameters. Another approach is to use the data directly to determine if changes have occurred in the structure based on changes in the data signature. These methods have been motivated by the development of wireless sensing technologies that enable analysis and evaluation on board of the sensor itself.
This presentation focuses on recent developments of structural damage detection algorithms that use statistical pattern recognition, machine learning and pattern classification methods. A brief review of progress on wireless sensing capabilities will be summarized first. Then, three damage-detection algorithms will be discussed including (a) autoregressive modeling, (b) wavelet transforms and (c) simple rotation estimations. Examples from the National Science Foundation Network for Earthquake Engineering Simulations (NSF-NEES) and California Department of Transportation (CALTRANS) - supported laboratory tests will be used to illustrate each method.