'Concrete Building, Monitoring Data and Bayesian Epistemology'

An important challenge for widespread application of structural health monitoring (SHM) in civil engineering is the creation and implementation of algorithms for automatic and reliable detection of unusual structural behavior. Branko, the lead role and the second author of this paper has been in charge of data analysis of a 19storey tall building. Observation of the data from the instrumentation has over the years, convinced Branko that there is an ongoing differential settlement of one of the base columns, in apparent contrast with his initial expectations. This conclusion matured gradually not only as a consequence of the monitoring results, but also based on verbal information received from a design engineer. Thus, besides the quantitative data provided by the monitoring system, including in the data analysis algorithms the engineer’s knowledge and experience has also been of value. In this study we propose an approach based on Bayesian logic as an effective tool to allow such a blend of field knowledge and SHM results. We show how the whole cognitive process followed by Branko can be suitably reproduced using Bayesian logic. In particular, we discuss to what extent the prior knowledge and potential evidence from inspection, can alter a perception of building behavior based on SHM data alone.