Bayesian model updating of a damaged school building in Sankhu, Nepal

This paper discusses Bayesian model updating of a damaged four-story masonry-infilled reinforced concrete structure using recorded ambient vibration data. The building, located in Sankhu, Nepal, was severely damaged during the 2015 Gorkha earthquake and its aftershocks. Ambient acceleration response of the structure was recorded with an array of 12 accelerometers following the earthquake. An output-only system identification method is deployed to extract modal parameters of the building including natural frequencies and mode shapes from the collected ambient vibration data. These dynamic properties are used to calibrate the finite element model of the building which is used to simulate the response during the earthquake. The initial three-dimensional finite element model is created using in-situ inspections. The goal of the Bayesian model updating procedure is to estimate the joint posterior probability distribution of the updating parameters, which are considered as the stiffness of different structural components. The posterior probability distribution is estimated based on the prior probability distribution of these parameters as well as the likelihood of data. The error function in this study is defined as the difference between identified and model-predicted modal parameters. The posterior distributions are estimated using the Markov Chain Monte Carlo stochastic simulation method. Ultimately, the stiffness values are estimated using the Bayesian model updating approach are compared with those from deterministic model updating conducted previously.