Unsupervised neural computation for event identification in structural health monitoring systems

This thesis report explores the use of unsupervised neural computation for event identification (EID) in structural health monitoring (SHM) systems. EID techniques are useful in SHM systems for minimizing the size of SHM data sets, and the costs associated with analysing, transmitting and storing SHM data. The approach to EID explored is adaptive, self-configuring and does not require detailed information about the structure being monitored.;A frequency sensitive competitive learning (FSCL) technique is used to model the output of an SHM system. SHM system output states which disagree with the model are deemed “novel” and identified as SHM events. The EID system is implemented in PERL and operates on SHM data stored by a database server running the MySQL DBMS software.;The EID system is evaluated with SHM data from three structures including the Taylor Bridge, the Portage Creek Bridge and the Golden Boy Statue. The EID system is able to identify strain gauge events of 0.75μϵ, 12.5μϵ, 1.25μϵ or smaller in the SHM measurement data from the Taylor Bridge, the Portage Creek Bridge, and the Golden Boy respectively. The EID system is able to identify accelerometer events of .0045g, 0.0020 g or smaller in the SHM measurement data from the Portage Creek Bridge, and the Golden Boy respectively.;The EID system is compared to a simplified event identification (S-EID) system, which does not use power spectral density estimation or unsupervised neural computation. The S-EID system is shown to be effective but less sensitive than the EID system to SHM events. The EID system is capable of adapting to noisy environments.;Some example SHM events, believed to be the result of seismic activity, from the Portage Creek Bridge are presented and discussed.