System reliability of timber trusses based on non-linear structural modelling

Structural design is today concerned with single component performance where each limit state is related to a single mode of failure of a single component. Further, in limit state codes the strength variables are related to a deterministic value (usually the 5-percentile). However, in a structure with a number of elements, two different effects (called system effects) can be found: The probabilistic system effect that is based on the reduced probability that weak sections coincide with the most stressed sections. Structural load-sharing that is the ability to redistribute load between members. In this paper, Monte Carlo simulations of a timber W-truss with punched metal plate fasteners (nail plates) are performed. Structural timber displays a significant variability in strength and stiffness both within and between members and is described by a statistic model calibrated against data from Norway spruce (Picea abies). Most studies found in the literature are based on linear-elastic theory and the variability within members is neglected mainly because of lack of data. The FE calculations are performed by TrussLab - a toolbox for MATLAB developed at Aalborg University. TrussLab considers contact between timber members and non-linear behaviour of the joints. The timber members are given linear properties. The system effect is based on reliability analyses. The system effect found depends on the coefficient of variation, the distribution of the random load variable and the reliability level. Depending on the assumptions, the system effect was found to be in the range 8-25%.