Year: 2013
Type: dissertation
Abstract: The past decades have seen a renewed interest in timber research related topics. As instance, timber engineering is not only confined to the design of new construction with innovative wood based products, but also an increasing awareness for the preservation of existing timber structures is visible in nowadays society. Therefore, arises the need for a better understanding of timber performance as a construction material aiming at a better safety assessment of existing timber structures and possible necessary actions to maintain its integrity. It is important to provide methods and tools to assess the safety level of existing timber structures by describing and evaluating each of its components. However, timber is a rather complex material and its mechanical properties present large variation within and between elements of the same structure. In that scope, this work was proposed regarding the assessment of existing timber elements with the purpose of establishing a methodology for structural safety evaluation. To that aim, initially a multi-scale experimental campaign was made regarding 20 timber chestnut (Castanea sativa Mill.) beams taken from an old building's floor. After, the database resulting from that experimental campaign allowed for the correlation between mechanical properties of timber between and within elements and in different size scales. In a second phase, the results of non-destructive tests and local mechanical sampling are used to predict and infer about reference key properties of timber. The results of visual grading and local bending tests were used to predict the global stiffness of a structural size element. Random sampling was also taken into account in the prediction models. Moreover, the use of Bayesian methods was considered to update the mechanical properties of timber based on non-destructive tests and different levels of belief. Posterior distributions for bending stiffness and strength were also obtained, through use of Bayesian Probabilistic Networks with different combinations of prior information. The proposed networks consider information from different size scales and results from different types of tests. Example cases were considered to illustrate the different methods and to demonstrate their applicability and value in the context of safety evaluation. The main outcomes of the present work are related to the prediction and inference of reference properties of timber, using a hierarchical model that combines information from visual grading and non-destructive testing with local mechanic testing information, attending to possible onsite conditions and available data. The work presented in this thesis was carried out at the Civil Engineering Department of University of Minho, Portugal.
Author: Hélder S. Sousa
Language: English
Subfield: Building and Construction
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W57600164