Following some incidents in high-rise buildings, such as Ronan Point London 1968, in which collapse of a limited number of structural elements progressed to a failure disproportionate to the initial cause, consideration of robustness was introduced in British Standard. The main method of preventing progressive collapse for providing robustness to steel framed buildings with precast concrete floor slabs focuses on the allowable tying forces that the reinforcement in between the slabs and in hollowcores should carry. However there are uncertainties about the basis of the practical rules associated with this method. This thesis presents the results of numerical and analytical studies of tie connection behaviour between precast concrete floor slabs (PCFS). It is shown that under current design regulations the tie connection is not able to resist the accidental load limit applied on the damaged floor slabs. By establishing the capability of a finite element model to depict and predict the behaviour of concrete members in situations such as arching and catenary action against several experimental tests, an extensive set of parametric studies was conducted in order to identify the effective parameters in enhancing the resistance of the tie connection between PCFSs. These parameters include: tie bar diameter, position, length, yield stress and ultimate strain; the slab’s height, length; and the compressive strength of the grouting concrete in between the slabs that encases the tie bar. Recommendations are made based on the findings of this parametric study in order to increase the resistance of the tie connection. Based on the identified effective parameters in the parametric study a predictive analytical relationship is derived which is capable of determining the maximum vertical displacement and load that the tie connection is able to undergo. This relationship can be used to enable the connection to capture the accidental limit load on a damaged slab. The identified parameters are examined in a three dimensional finite element model to assess their effect when columns of the structure are lost in different locations such as an edge, corner or internal column. Based on the findings of this study methods for improving the connections performance are presented. Also the effect of alternative transverse tying method is evaluated and it is concluded that although this kind of tie increases the load carrying capacity of the connection, its effect on the catenary action is not significant.

Masters Research - Master of Philospohy (MPhil) / It has been shown that masonry material properties, in particular, unit flexural bond strength (ft), vary significantly throughout masonry structures, despite the fact that often only one type of brick and mortar are used. Unit flexural bond strength was previously identified as one of the most important material parameters contributing to the strength of clay brick unreinforced masonry (URM) walls in flexure. It was the objectives of this research, in the context of clay brick URM walls subject to vertical bending, to examine how unit flexural bond strength varied spatially in a clay brick URM wall, determine a best fit probability distribution function which can describe expected variability in unit flexural bond strength and determine how this variability and other factors affect wall behaviour and failure load using 3D non-linear finite element analysis (FEA). It was hoped that modelling a full sized clay brick URM wall subject to vertical bending using a 3D non-linear FEA model would more accurately predict wall failure load (compared to current analytical methods) and allow the examination of crack pattern development as the wall progresses to failure upon being laterally loaded. The first part of the research project was to conduct an experimental program to examine unit-to-unit spatial strength correlation within six full sized clay brick URM walls and to characterise a unit flexural bond strength probability distribution. It was observed that although weak correlation in unit flexural bond strength exists in some courses and between courses, these locations were difficult to predict and didn����t follow any particular pattern relating to for example, mortar batch. Therefore, although somewhat counter-intuitive, the results indicate that statistically significant correlation between adjacent unit flexural bond strengths is not likely to be observed. It was also observed that clay brick wall unit flexural bond strengths obtained for all of the walls tested best fit a truncated Normal probability distribution. Strength of the brick/mortar interface appeared to be governed by factors relating to workmanship (and therefore mortar quality and moisture content), weather (which can affect material characteristics like brick suction rate) and inherent material variability. It would appear that brick suction rate can significantly affect the overall strength of a URM wall. v Stochastic analysis was conducted for walls with and without uncorrelated spatial variability in unit flexural bond strength and associated tensile fracture energy (GfI ). It was found that the TNO DIANA 9.2 FEA package could be used to implement spatial variability of various material parameters and reasonably accurately model failure of clay brick URM walls in vertical bending. From the non-linear FEA model development stage, it was observed that because the brick/mortar bond has significantly more strength capacity in compression, it appears that the lateral load resistance of the wall comes from a combination of the ability of the brick/mortar bond to tensile soften while providing significant compressive resistance at the compressive edge. It was found for a spatial stochastic analysis with spatial variability in bond strength (referred to from now on as a spatial stochastic analysis), with COVs of 0.1, 0.3 and 0.5, that COV of wall failure loads were relatively small, being 0.02, 0.04 and 0.06 respectively. For the non-spatially varying stochastic analysis with fully correlated bond strength (now referred to as non-spatial stochastic analysis), with COVs of 0.1, 0.3 and 0.5, COV of wall failure loads were 0.07, 0.20 and 0.32 respectively. For the spatial stochastic analysis, it was found that with a bond strength COV increase from 0.1 to 0.5 the mean wall failure load dropped from 2.25 kPa to 2.0 kPa (an 11% reduction). Despite the relatively small drop in magnitude of the mean wall failure load with increase in bond strength COV, the mean wall failure loads were statistically different to one another. For the non-spatial stochastic analysis, mean failure load stayed relatively constant at 2.24-2.25 kPa. These results could be explained by examining the 3D wall progression to failure. For walls with spatial variability in bond strength, it is expected that wall failure load COVs would be smaller because those walls would consistently be composed of smaller valued bond strengths which would consistently contribute to weakness in the wall. For the non-spatial wall simulations, this effect would not occur as failure load is determined by one uniform weak or strong bond strength. It was proposed that failure of a clay brick URM wall is not governed by one course only cracking, but rather, instability in the wall is governed by several courses in the vicinity of locations of large bending moment. It was shown that various current stochastic approximations which employ a unit failure hypotheses in combination with a linear/elastic approximation for first cracking load all underestimated wall capacity significantly. The reason for this is suggested as being vi because all hypotheses only assume failure is governed by one course and linear/elastic theory only considers the tensile capacity of a joint and neglects strength capacity available as a result of joint tension softening and the resistance to failure provided by compressive strength on the compression side of the wall. The hypotheses also don’t take into consideration factors which affect overall wall bond strength mean which result from influences such as workmanship, weather and material variability factors, such as (for example), variation in brick suction rate due to weather conditions which can make the overall strength of the wall stronger or weaker. Based upon a comparison in wall failure load COV for the spatial and non-spatial stochastic wall analysis results, a more realistic approach for future modelling attempts of spatial variability in masonry material properties is suggested. This would address the issue of external factors such as workmanship and weather on the overall strength of the wall, as well as the inherent bond strength variability due to material variability. For walls with spatial variability in bond strength, upon examination of numerous wall simulation results, several crack patterns were witnessed and are discussed.