This research investigated the fire performance of unprotected timber floors, focussing on composite joist floors, composite box floors and timber-concrete composite floors. The study of these floors was conducted using the finite element software ABAQUS using a thermo-stress analysis in three dimensions, and with experimental fire tests of floor assemblies. The major goal of this research was to develop a simplified design approach for timber floors, validated against the numerical and experimental work.

Four furnace tests were conducted on unprotected timber floor systems in the full-scale furnace at the BRANZ facilities in New Zealand. The tested floors were one-way strip floors with pinned support conditions exposed to the ISO 834 standard fire for varying durations of 30 – 105 minutes. The floors were loaded under standard office loading conditions of 3.0kPa live and 1.0kPa superimposed dead loading. The charring rates of the LVL timber members were found to range from 0.66 – 0.86 mm/min across all specimens. When designed to resist a similar load level both the composite joist and box floor types had a similar response to the fire loads, however the joist floors exhibited increased upward burning through the beam members in the latter stages of testing which may contribute to earlier failure times for smaller floor geometries.

A sequentially coupled thermal-stress analysis was conducted to determine the effects of a fire on floor assemblies under load. Firstly a thermal analysis was performed to determine the temperature profile of the floor assemblies for the duration of modelling, and then a stress analysis was performed using the temperature profile as input into the structural model. With regards to the thermal modelling, a proposed set of effective values was used to account for the mass transfer processes occurring in the timber. The thermal modelling predicted the charring damage of the floors tested in the experiments to within a few millimetres of precision, and the simplified assumptions made in relation to fire inputs, boundary conditions, mesh refinement and effective material parameters were accurate to the desired level of precision. A sensitivity study was conducted comparing different mesh sizes, time step sizes, material model approaches and software suites to determine any shortfalls which may be encountered in the analysis. It was found that a material model adopting a latent heat approach was the most adequate for modelling timber in fires using these effective values, and mesh sizes of up to 6 mm produced relatively precise results.

The structural modelling predicted the displacement response and failure times of the floors to within 20% of the experimental data, and the simplified assumptions made in relation to fire inputs, boundary conditions, mesh refinement and effective material properties were once again accurate to the desired level of precision. A modification to the reduction in tension strength at elevated temperatures was proposed to better predict the observed behaviour. A sensitivity study concluded that the material model definition plays a vital role in the output of the modelling. Non-standard fire exposures were also modelled for completeness.

A simplified design method to estimate the fire resistance of unprotected floor assemblies was also developed. The method uses a bi-linear charring rate the assumption of a zero strength layer in the timber. The method was compared to the experimental data from this research and others around the world. The results were also compared to other charring rate methodologies from around the world.