GNGTS 2018 - 37° Convegno Nazionale

376 GNGTS 2018 S essione 2.2 maximum dimensions, and later on thes panels with severe/visible damage. Preliminary on- site observations carried out during Summer, to qualitatively assess the state-of-the-art of the structure, gave in fact evidence - for some of the roof panels - of (d 1 ) surface abrasion and minor glass cracks; (d 2 ) condensation and debonding phenomena (especially at the LG-to-AN interface); and (d 3 ) dislodgement and pre-stressing loss, for the steel tendons and supports (see Fig.1(c)). All the test measurements were analysed (SMIT), to detect the fundamental mode of the roof (Fig.2(b)). Careful consideration is spent, in this paper, for the 1.45×2.65m G2- panel of the central nave, see Fig.1(d). According to the test setup of Fig.1(d), through the post-processing phase, #4-sensor measurements were disregarded, due to data corruption. The examined glass panel showed a beam-like bending response (Fig.2(b)), with f TEST =14.97Hz the fundamental frequency and ξ TEST =1.20% the estimated damping. FE numerical study. An extended numerical investigation of the glass walkway was then carried out in ABAQUS, to further assess the on-site measurements (see Fig.2(c)). Major uncertainties on the interpretation of the vibration test data derived from the actual supports and restraints of the glass panels, as well as on the effective material properties (especially PVB). The typical FE numerical model herein discussed was implemented to reproduce the G2-panel of Fig.1(d). 2D shell composite elements were used to describe the nominal LG cross-section. Similarly, for the top AN layer, a shear flexible bond was defined on the top surface of the LG sandwich ( E SOFT =1MPa its stiffness), to account for a contact mechanical interaction. 1D beam elements were used for the steel tendons, accounting for their nominal circular section (10mm the diameter). MEMS sensors were then considered via lumped masses (0.15Kg/each, see Fig.2(c)), while a key role was assigned to the lateral restraints (linear supports for the roof panel) and to the intermediate restraints (for the tendons). In the latter case, see Fig.2(c), an unilateral contact interaction was properly defined, being able to react in presence of compressive loads only, hence allowing free relative displacements and rotations at the steel support-to-LG interface. Pre-stressing effects of tendons were fully disregarded, based on on-site observations. A final uncertainty was represented by the input material properties. While nominal elastic features were reasonably considered for glass ( E g =70GPa, ν g =0.23, ρ g = 2500Kg/m 3 ) and steel ( E s =160GPa, ν s =0.3, ρ s =7850Kg/m 3 , with 1600MPa the resistance at rupture, according to technical drawings of the structure), a tentative value was initially used for the shear stiffness of PVB ( G PVB =8MPa and E PVB =24MPa, with ν PVB =0.49, ρ PVB =1100Kg/ m 3 ). Such a stiffness value is in fact suitable for short-term loads (3s) and room temperature (20°) only, while a certain material degradation was expected for the Aquileia walkway (see for example Fig.2(c) and (CNR-DT 210/2013)). Results. The dynamic performance of the glass structure was extensively investigated. Parametric FE estimations, in particular, were compared with experimental measurements and preliminary analytical calculations, where ( k =1 for the first vibration mode): (1) In Eq.(1), m is the linear density of the beam-like element, L the span, E = E g , I the flexural inertia. For the analytical calculations, a 3×12=36mm thick monolithic glass section was roughly considered, in place of the LG+AN sandwich. Within the full set of parametric FE models (Tab.1), the attention was spent for several geometrical and mechanical aspects. The numerical predictions resulted in a fundamental beam-like shape well agreeing with the test observations (Figs.2(b) and 3(a)). Marked variations, otherwise, were observed for vibration frequency, see Fig.3(b). There, the percentage scatter of the calculated frequencies f x is defined as: (2) Worth of interest in Tab. 1 and Fig. 3(b) is that Eq.(1) does not allow to capture the actual vibration of the roof (i.e., poor description of the LG+AN section, lack of tendons, etc., with Δ f