Structural behaviour of cross-laminated timber elements in fires
Cross-laminated timber (CLT) is a comparatively novel engineered material consisting of timber boards that are arranged in layers of alternating directions and bonded with adhesives. It is increasingly the material of choice to realise mid height to tall timber buildings; however, because timber is a combustible material there are concerns with respect to its fire safety when used for tall buildings. As an engineered material the configuration of CLT can be varied in multiple ways, with currently unknown implications for its fire performance. While a vast body of knowledge exists for the structural behaviour of sawn timber, it is unknown whether this knowledge can directly be applied to CLT, which is utilised in the form of factory produced large floor slabs and wall panels. Structural fire resistance for CLT panels has so far mostly been assessed in furnace testing following standard temperature time curves, and there remains a paucity of data on the structural fire performance of CLT walls. The understanding of how specified adhesive types and board configurations influence the fire performance of CLT elements is lacking. The research presented in this thesis studies how two adhesive types and two ply configurations may influence the response and load bearing capacity of CLT walls (and other elements) through all stages of a fire. Three experimental series have been completed, all investigating different loading and heating conditions at different scales for CLT that was sourced from one producer but with two variants in adhesive type and ply lay-up. This results in four different CLT configurations being considered. The first series assessed the axial compressive strength of CLT at small scale when heated to non-charring temperatures of up to 220 °C. The median strength retention upon heating in small scale compression experiments was observed to be 9 % less for CLT specimens bonded with the polyurethane (PU) adhesive type than for those bonded with melamine formaldehyde (MF) when only transient experiments or those with remaining moisture in the timber were considered. The influence of migrating moisture was observed to be the driving mechanism weakening timber in compression, and ultimately inducing failure. The second experimental series investigated the flexural behaviour of CLT beams under slow transient heating to non-charring temperatures of up to 150 °C at two applied load ratios, and illuminated the significance that shear stresses and the resulting shear strains at the adhesive bonds exert on the deflections of the different CLT configurations. Mean heat induced deflections were observed to be 40 % higher for specimens which utilised a polyurethane adhesive type compared to those using melamine formaldehyde; this effect was amplified for CLT made up of three plies compared to five plies. Through heating and subsequent cooling these experiments also demonstrated that heat induced deflections in bending for CLT are dominated by creep, and are irrecoverable. A third experimental series subjected CLT wall strip elements to a radiative heat flux, representative of conditions that might occur in a real fire, under equal constant load ratios. Walls that were bonded with the PU adhesive type failed significantly earlier (20 percentage points) than the mean failure time for all specimens. Samples consisting of three plies exhibited shorter (17 percentage points) failure times than the overall mean failure time. These occurred at similar char depths and with good repeatability, and were attributed to accelerated deflections due to increasing shear deflections in these samples causing exponentially increasing bending moments from P-Delta effects and subsequent buckling failure. In addition, heating and additional cooling of samples highlighted the propensity of CLT walls to fail in the decay phase of a fire, due to the tendency of heat to transfer deeper into the timber even after the primary heat source is removed. Numerical simulations of the wall strips with a wide range of mechanical input parameters highlighted the importance of accounting for weakening glue lines at elevated temperatures to predict failure in fire numerically. The simulations also showed that the failures in the fire decay phase could not be explained solely by redistribution of elevated temperatures through the cross-section; instead they were likely caused by creep due to changes in moisture content. The findings in this thesis are novel and have significant potential to aid manufacturers and designers to assess and improve their products' response to fire in the production stages. The results and their analysis presented herein raise questions regarding the applicability of standardised fire resistance furnace tests in qualifying CLT for use in tall timber buildings; instead more holistic approaches are advocated. Overall this thesis describes novel experimental set-ups that have been used to assess the response of CLT in previously unknown detail with a unique materials stock that allowed variation of two experimental variables whilst keeping the underlying timber quality and manufacturing constant. The results consistently demonstrated that adhesive type had no discernible influence on the load bearing capacity at ambient reference temperatures, but the PU adhesive type was found to significantly reduce the structural performance of CLT exposed to heat or fire. This was especially evident for CLT walls, where additional deflections caused by weakening of the bond lines led to earlier global buckling failures. Failure of CLT walls in the fire decay phase, as presented and investigated in this thesis, should be a concern for all engineers that work on tall timber buildings. The potential for structural collapse after burnout of the fuel load is shown as a real possibility and engineers and architects working with and advocating for tall timber buildings should be aware of the limits of their own knowledge and the limits of the state of the art on these issues.