Full Scale Testing

Full scale testing is the most thorough method available to study structural behaviour. Early studies by Boughton (1983) and Wolfe and LaBissoniere (1991b) used full-scale tests on roof assemblies and entire houses to determine load-sharing behaviour. Notable full-scale studies are outlined in the following section.

Boughton and Reardon (1982) and Boughton (1982),(1983) conducted extensive full scale testing on two complete houses. These studies examined load sharing of lateral and uplift loads, load redistribution and the effect of non-structural elements to the load-path. For the houses studied, large displacements at roof to wall connections were required for load to be redistributed to neighbouring trusses.

Under uplift loading, the redistributed load was carried to elements surrounding elements that had failed. This caused overloading of neighbouring connections and a spread of failure. This was observed in the case of batten to rafter connections that span continuously across rafters: The loss of one connection significantly increased the load on adjacent connections resulting in a propagation of failures along the batten.

Different behaviour was observed for lateral loading, where failure of connections occurred in a more ductile manner. Additionally, the mechanisms of transmitting uplift loads are different from transmitting lateral loads. Lateral loads were found to be transmitted through walls and floor and ceiling diaphragms. Upon the failure of one bracing structure, the large in-plane stiffness of the roof diaphragms enabled the redistribution of lateral loads to other components.

Probabilities of failure and resulting progressive failures are then related to reliability theory. The probability of failure of a large section of roof is related to the probability of a defect occurring in a critical location. For a roof structure, the probability of failure is a function of the connections’ resistances, the number of connections, the location of the connections on the roof and the quality of construction.

Non-structural elements such as wall and ceiling linings as well as ceiling cornices were found to attract load and contribute to the overall structural response of the house. This was primarily for lateral loads. Some diaphragms such as the ceiling were found to stiffen with increased load by reducing slack and closing gaps. The effects of these non-structural elements were summarized to have three main effects:

1. They can render actual structural systems redundant due to the unintended load paths.

2. Loads attracted by non-structural elements can cause the premature failure of these elements, reducing the overall strength of the structure.

3. Non-structural elements can give post failure strength by carrying loads that were previously carried by elements that have failed. However this feature is less applicable for uplift loads on the roof surface.

Wolfe and LaBissoniere (1991b) and Wolfe and McCarthy (1989) conducted tests on an assembly of roof trusses for the Forest Products Testing Laboratory (FLP) to study load-sharing behaviour. The process of load sharing was through two-way action and partial composite action and occurred only when a connection deflects relative to its neighbours. Additionally, it was found that when individual trusses are subjected to their design loads along the top cord: 40 to 70% of the load can be distributed to adjacent trusses. A limitation of this study from a wind loading perspective is that uniform gravity loads were applied. Shivarudrappa and Nielson (2012) showed that load transferred to connections for gravity vs. uplift loads could differ by 30 to 40%.

Indicating that influence coefficients would differ significantly depending on the direction of loading.

Paevere (2002) conducted testing for lateral loads on light framed structures. This study was focused on seismic loads and not wind uplift. Additionally, this study, although conducted in Australia, examined a North American type structural system and presents a detailed review of the many types of hysteresis models available and various techniques for nonlinear analysis.

Doudak (2006) conducted field tests on an instrumented building subjected to wind and snow loads. Additionally, point loads were applied and displacements were measured at several locations. The study found that 73% of load applied at the mid-span of a roof joist was redistributed to adjacent joists. This was determined from displacement readings, as actual loads were not measured. Another limitation of this study was that limited conclusions could be drawn for behaviour due to wind loading due to the variability in wind speeds and directions.

Zisis and Stathopoulos (2009) conducted tests on an instrumented gable roof house to determine the attenuation of wind loads as they are transferred from the roof to wall connections due to dynamic effects (energy absorption). Wind pressures on the roof surface were measured and actual reaction loads measured using 2D and 3D load cells. The measured loads were compared to those determined from an idealized structural model with the same applied wind pressures. They found that the reaction forces from the idealized model were 26-46% higher than those measured by load cells, they attributed this to the effect of structural attenuation. However, as commented by Datin (2010), this is likely due to the simplifications of the structural model and not entirely from attenuation effects. Further tests on this setup were performed by Zisis and Stathopoulos (2012), this more recent study included a dynamic analysis using a Finite Element Method (FEM) model.

Morrison et al. (2012b) and Henderson et al. (2013) conducted full scale tests at the University of Western Ontario as part of the ‘Three Little Pigs’ project. These studies have best represented the complex nature of wind loads on roofs and the structural response of connections due to spatial and temporally varying pressures. Pressures scaled from wind tunnel studies are applied to the roof surface using ‘Pressure Load Actuators’ (PLAs). These specialized devices are able to follow a specific pressure trace accurately and apply this to a section of roof surface.

Morrison and Kopp (2009) used 58 PLAs to apply realistic wind pressures to a two-storey Canadian gable roof house and examine the response of toe-nailed roof to wall connections. The study found that roof trusses behaved as rigid members and that simple tributary area methods overestimate reaction forces, indicating load-sharing behaviour. Additionally, the amount of load sharing was found to change throughout the time history due to incremental nail pull out. Hysteresis was also observed during the failure of connections. Ultimate roof failures were observed to occur when multiple connections fail simultaneously

Henderson et al. (2013) later conducted similar experiments on a hip roof. Load cells were installed under the top-plates of selected roof to wall connections. A series of patch loads were applied sequentially to determine influence coefficients at roof to wall connections. Time-history loading was applied to the roof surface and the influence coefficients re-measured at the end of the simulated windstorm. These studies found that influence coefficients for roof to wall connections change only during damaging peak loads, reinforcing previous findings by Morrison and Kopp (2009). Changes to these influence coefficients occur almost continuously during high wind events due to multiple peak loads occurring at different locations throughout time. The change in influence coefficients for roof to wall connection reactions could be used to determine the redistribution of loads and load sharing that occurred in the roof structure