Clancy, Paul (1999) Time and probablity of failure of timber framed walls in fire. PhD thesis, Victoria University of Technology.

Abstract

The general aim of the research for this thesis was to develop models to predict the performance of fire resistant light-timber framed walls, in relation to requirements of new performance-based building fire regulations being introduced around the world. The performance bases that were adopted in the modelling, were time and probability of failure. The models that were developed for a wide range of wall types, and have been specifically validated for ordinary cavity walls and double framed walls. Validation involved numerical checks for convergence and energy conservation, and the undertaking of eight full scale wall furnace experiments with well controlled conditions. Values of thermal properties of materials in the walls were obtained with simple independent experiments. These experiments helped to clarify appropriate values from the wide ranges which have been previously published. Model predictions were evaluated against results published in the literature. The models comprise specific models for fire severity, heat transfer, structural response and probability of failure. This research has developed the heat transfer model named ADIDRAS, a probability model, and undertaken all linkages of models to produce the time and probability of failure models. A model developed by a colleague, Young, was adopted for the structural response model. The time and probability of failure models can incorporate most fire severity models. In this research, two fire severity models were incorporated; one was the standard fire, and the other was the real fire severity model of Kawagoe and Lie. ADIDRAS, in several respects, is an advance on previous heat transfer models for structures containing cavities. It analyses thermal diffusion with alternating direction implicit finite difference analysis which enhances numerical stability without the loss of computation speed of explicit procedures. It uses the discrete radiation method which enables analysis of radiant heat transfer through smoke and cavities of any shape. Previous models have analysed only cavities without re-entrant corners. It was deduced that when the interfaces between studs and gypsum board on the fire side reach temperatures greater than 400°C, the interfaces open to form gaps due to shrinkage in these materials. These gaps prevent much of the heat transferring directly from gypsum board to the studs, and divert the heat to the cavity by radiation. This diversion explains some anomalies between experimental results and predictions of previously published heat transfer models. It is shown that temperatures in studs are insensitive to the transmissivity of radiant heat in smoke. The structural model and the time of failure model have been applied, to obtain for the first time, relationships between the time of failure and a number of variables including vertical load, elastic modulus of timber, strength of timber, height of wall, size and spacing of studs, initial crookedness of studs and thickness of gypsum board. The time of failure model has established failure, induced by the vaporisation of moisture, is critical for walls that are either higher than walls commonly built (higher than 3.6 metres), or are more heavily loaded than is currently permitted in timber structural engineering codes. It was found that the time of failure is most sensitive to fire temperature, vertical load, elastic modulus of timber in compression, and density and specific heat of timber. It is sensitive also to the thickness and thermal properties of gypsum board, and the sectional dimensions of studs; however, these variables are unlikely to vary much for walls built to specifications in common building construction. It is considered that the probability of failure model is the first theoretically-based reliability model for light-timber framed structures exposed to fire. Several applications of the model have been demonstrated. It was shown that the variabilities of thermal properties of gypsum board and wood are low compared with the large variabilities of mechanical properties inherent in timber. In the applications demonstrated, typical variations of all variables led to a coefficient of variation (Co V) in the time of failure of 0.12 for typical walls subjected to standard (ASI530A) fire. Exposure to a real fire did not significantly affect this CoY. Unlike the standard fire, the real fire did not necessarily lead to collapse. There is much scope for reducing fife resistances of walls without significantly increasing the risk to life.

Search Google Scholar

Repository staff login