| Popis: |
For the development of sustainable building solutions, research on innovative structures figures a key contribution. In Switzerland, recent developments pursue investigations on innovative structures in hardwood, as hardwood is available in large quantities, without a sustainable, large-scale field of application. An implementation of hardwood in the building sector would be optimal, as large cross sections are processed and its excellent strength and stiffness properties render it a competitive material. Within this context, a pilot building, displaying the implementation of structural elements with hardwood, was designed and constructed at ETH Zurich. This research work aims at investigating the structural behaviour of the innovative structural systems implemented in this prototype building, named ``House of Natural Resources (HoNR)''. The building functions as a living lab, i.e., it unites the dual functionality of a research building and a regular office building. The main objective of this work is to extend the knowledge basis on the structural behaviour of the innovative timber-hybrid structural systems and to lead towards an exploitation of the full potential of the promising hardwood material. To investigate the structural behaviour, forced vibration tests are performed, combined with an extensive long-term monitoring campaign. The focus of this work is the development of data evaluation methods to process the large amounts of acquired dynamic and long-term monitoring data. First, a thorough background on existing data processing methods is provided. For the evaluation of the forced vibration and operational dynamic data, these methods stem from the domain of system identification and operational modal analysis methods. Furthermore, model updating and optimal sensor placement methods are presented. Model updating techniques are commonly implemented to update numerical models of the structure under investigation based on the results from the forced or ambient vibration tests. Optimal sensor placement methods are employed for the design of vibration test set-ups to determine the optimal positioning of a limited number of dynamic response measurement sensors. For the evaluation of the long-term monitoring data, structural health monitoring (SHM) methods are investigated. Additionally, several application examples to timber and timber-hybrid structures are presented. Subsequently, two data processing frameworks are developed. The first framework proposes an automated modal analysis, aiming at the automated processing of a large number of forced and ambient vibration test series. The framework implements three different system identification methods: a non-parametric method (nonpar), the poly-reference least squares complex frequency domain method (pLSCF) and the stochastic subspace identification method (SSI). The automated modal analysis procedure delivers the modal parameters of the structure. The framework implements an automated modal analysis method developed by Reynders et al. (2012) \cite{reynders_fully_2012} for the SSI method and is enhanced with adapted automated modal analysis procedures for the pLSCF and nonpar methods within this thesis. Additionally, the developed framework proposes a quantification of the uncertainty on the derived modal parameters, through a clustering-based combination of the results from all tests conducted on the same specimen. The second framework implements prediction models for the development of an SHM platform, which is able to automatically supervise the long-term performance of the monitored structural system. Subsequently, the case-study building HoNR is described and results from the forced and ambient vibration tests are presented together with the long-term monitoring data from the construction phase and the first two and a half years of the operation phase. The data is evaluated with the two proposed frameworks. From the data evaluation, the following conclusions could be derived. The fundamental frequencies of the building at the different construction stages could be derived with high precision. The precise quantification of damping ratios, however, was not possible. The predictive models, developed for the key parameters, i.e., the tendon forces in the post-tensioned timber frame and the fundamental frequency of the building, delivered promising results. The models predict maximal tendon force losses of 30\% for a service life of 50 years. This value is lower than the design value, which was determined from small-scale laboratory tests prior to the construction of the building. Within this thesis, the knowledge basis on the structural behaviour of the innovative structural systems implemented in the HoNR could be significantly extended. An actionable framework is delivered for effectuating the validation of the innovative structural systems, implemented in the HoNR in ETH Zurich, and fostering the adoption of novel data evaluation technologies into engineering practice. |
