The research reported in this thesis focused on the experimental and numerical stochastic assessment of unreinforced masonry (URM) veneer wall systems (that is, timber as a flexible backup system is connected to masonry wall via metal ties), representative of those used in the contemporary constructions in Australia, under out-of-plane loading. Previous investigations have identified a need to develop an improved understanding from the detailed examination, by test and analysis, of the behaviour of wall ties and interaction between single masonry leaf and supporting frame. Although consideration of spatial and temporal variability of the material properties for a single leaf of masonry wall has been directed at developing predictive strength models for masonry, no studies have so far been carried out towards the consideration of spatial variability of all components (mortar joints, wall ties and timber studs) of the URM veneer wall system, model error and their effect on wall system strength. This PhD thesis estimates the veneer wall system's failure load from the Monte-Carlo experimental and numerical technique. In doing so, it is hoped that the stochastic finite element models developed in this study will be used to calculate the structural reliability and fragility of masonry wall systems for new and existing construction during a structure's service life. Probabilistic veneer wall tie characterisation is accomplished to generate a nonlinear tie constitutive law for the representation of the tie behaviours in the nonlinear FEA models. A total of 50 brick-tie-timber subassemblies were tested under compression and tension, and elastic stiffness, peak strength and displacement capacity were recorded for each subassembly. The best-fitted distribution, mean and COV for each point (which define the tie constitutive law) was calculated, and correlations between these points were established. Monte-Carlo experimental investigations of 18 full-scale URM veneer wall systems with theoretically identical geometries and properties under out-of-plane loading was conducted. Ten wall systems were tested under inward loading (ties are in compression), and the remaining eight were tested under outward loading (ties are in tension). For inward loading, the airbag was pushed against the veneer wall directly, while for the outward loading, polystyrene blocks were utilised to transmit the pressure from the airbag to the veneer wall. For each loading type, one specimen was tested for semi-cyclic loading to check whether the monotonic loading can capture the overall behaviour of the cyclic response. For each batch of mortar mixed, bond wrench testing was conducted at the same age as the test for the associated wall constructed using that mix. Batch to batch variabilities were statistically analysed, and probability distributions for flexural tensile strength were established. A lognormal distribution with an aggregated mean of 0.40 MPa and 0.42 MPa for inward and outward loading, respectively, was estimated for flexural tensile strengths. From the wall tests, veneer wall system behaviour was observed and measured until the collapse or 20% post-peak drop of the peak load. Tie force history along with the timber stud deflections were also recorded and analysed to understand the veneer system failure mechanism. Parallel to the wall tests, material characterisation tests for masonry were conducted to develop the material model to define the masonry in the nonlinear FEA model. After the wall tests, all timber studs used to build the veneer wall were tested to evaluate the modulus of elasticity and bending strength. Prior to stochastic finite element analysis, a deterministic model was developed using Diana FEA 10.3 which considers the mean values for masonry, tie and timber material properties obtained from the laboratory material characterisation tests. The boundary conditions and loading arrangements were established in such a way so that it can replicate the laboratory full-scale veneer wall system tests. As expected, deterministic FEA failed to estimate the wall strength (system peak load) for inward and outward loading. A deterministic FEA with characteristic strength values for all veneer wall components was also evaluated to replicate the design (weaker than average materials) behaviour. Sensitivity analysis (one and two standard deviations below and above the mean) for deterministic FEA was conducted for target input variables, the flexural tensile strength of masonry, tie capacity (stiffness and strength), and timber stiffness to understand how these parameters affect the system peak load. The system peak load is comparatively more sensitive to the masonry bond strength and tie stiffness/strength. Moreover, if the veneer cracks earlier (lower masonry strength) and redistributes the forces in those mid-range ties before the top row of ties failed, the veneer system has the potential to resist a higher magnitude peak pressure. Spatial stochastic finite element analysis considered the spatial variability (unit to unit correlation ρ=0 and 0.4) of the wall components (mortar flexural tensile strength) and compared them with non-spatial analysis. The non-spatial analysis underestimates the wall system failure compared to spatial analysis, and the spatial analysis is considered to more realistically represent the variabilities of the URM veneer wall system. Stochastic sensitivity analysis is conducted in turn to check the sensitivity of the veneer system behaviour to variability in the various input parameters, considering one parameter at a time. Moreover, all the variabilities and uncertainties observed in the laboratory wall testing are reported and quantified to make the Monte-Carlo experimental results comparable with SFEA. From the comparison, it is evident that the stochastic finite element model developed in this study can estimate the behaviour and system peak load reasonably and are considered to be from the same population as test results.
|Qualification||Doctor of Philosophy|
|Publication status||Published - May 2021|