Vibration-based model updating of a timber frame structure

Abstract

This research paper presents the implementation of Bayesian updating to an innovative post-tensioned timber frame structure, which has recently been erected at ETH Zürich as the main lateral load carrying system of the ETH House of Natural Resources (HoNR).In this innovative timber frame, the columns and beams are solely connected via a straight tendon running through the beams at mid-height, in this way creating a moment-resisting frame structure. A numerical model of the post-tensioned frame structure is set up and updated on the basis of modal vibration data offering a deeper insight into the system’s behavior and corresponding parameters. The modal testing data is derived from laboratory testing of a 2D-frame set-up, investigated under different support conditions. The dynamic acceleration response of the structure was processed by means of subspace identification methods for inferring the modal characteristics of the structure (frequencies, mode shapes, and damping ratios). Parallel to the modal vibration tests, an a priori numerical model of the structure has been established, which nonetheless relies on a number of assumptions on defining structural properties, including material, support, and rigidity parameters. Bayesian model updating is adopted for the updating of these parameters by exploiting the value of information contained in the testing data. The finally rendered updated model features parameters of reduced uncertainty, which is attested by the superior fit to the experimental data. The updated model may now be employed for investigating further the dynamic behaviour of the existing frame system as implemented in the HoNR.