Experimental investigation for the structural performance assessment of square hollow glass columns

In this paper, the results of pilot experiments carried out on glass columns are presented and critically discussed. The aim of this preliminary experimental study consists in the verification and assessment of the actual global behaviour of monolithic glass columns with square hollow cross-section under axial compression, as well as in the estimation of the compressive loading ratios leading to the opening of first cracks and collapse of the same specimens, respectively, together with the corresponding failure mechanisms. In order to properly interpret the obtained test results, the experimental predictions are first compared and assessed towards analytical calculations (e.g. in terms of Euler's buckling load) and Finite-Element (FE) models later (e.g. in the form of load-displacement response). In the first case, classical analytical formulations not able to take into account the effects deriving from a combination of several aspects (e.g. possible initial geometrical imperfections in each glass pane or in the assembled square hollow section, load or boundary eccentricities, as well as the actual bonding effect of the glued joints connecting the glass components) typically result in a marked overestimation of the actual load bearing capacity for the tested columns, hence suggesting more refined investigations. Rather close agreement is found in fact with properly assembled fully 3D solid FE models. In them, a key role is assigned to the geometrically refined description of each specimen component (including the glued joints and the end restraints), as well as to the accurate mechanical calibration of the materials, based also on further experimental tests carried out on small adhesive specimens. Finally, the feasibility and accuracy of simplified 2D FE-models - less time consuming than fully 3D assemblies - is also assessed, so that practical recommendations could be provided. In the latter case, as shown, higher sensitivity of FE results to the main input parameters is found, with consequent underestimation of the actual elastic resistance for the tested columns. Based on the current comparisons and findings, however, it is expected that the current research outcomes could be further extended to various glass columns typologies and configurations (e.g. including laminated glass specimens and several cross-sectional shapes, as well as full-scale experiments), in order to develop appropriate practical design rules.