Methodology for the prediction of the strength of naturally aged glass based on surface flaw characterization

Abstract

Flat glass is a material which is used extensively in almost all the buildings, mainly as infill for windows or facade panels. However, neither glass production nor its recycling are sustainable processes, as both incorporate the use of gas furnaces, which produce high amounts of CO2 emissions. Thus, the re-use of glass seems the only way to minimize the environmental impact of this material.

The greatest challenge in the safe reuse of glass elements lies in the assessment of their strength after several years of use. This strength is reduced, compared to the inert strength, due to the damage ("defects") that the environment inevitably introduces into the glass surface. According to the Linear Elastic Fracture Mechanics (LEFM), one of the defects in a stressed brittle material, which in this case is glass, will initiate the failure. Therefore, in order to assess the residual strength of weathered glass, it is essential to quantify it in terms of defects. The aim of this thesis is to develop the understanding on the performance of weathered glass, and to propose a methodology for the prediction of its residual strength, based on its surface defects.

The examined material is a 55-year-old annealed glass, which was used as facade elements in a building in the Hague, in the Netherlands. The X-Ray Fluorescence analysis showed that the composition of this glass meets the common float glass recipe but no tin residues were found on its surface.

This research firstly, focuses on the detection and characterization of the defects on the examined weathered glass. Both sides of glass were examined through non-destructive tests with the mobile optical profilometer Traceit® and the digital microscope Keyence VHX 7000. These tests revealed that the weathering induced mainly dense pits, digs and fine linear scratches on the surface exposed to the outside environment (external surface). The depth of these defects, as measured with Traceit®, ranged from 19 µm to 161 µm, which is aligned with the depths reported in literature, which range from 20 µm to 200 µm (Schula et al., 2013). The internal surface had occasionally some large defects which were probably man-made defects. Overall, Traceit® showed a potential for detecting and measuring the defects on weathered glass, whereas with the used microscope it was not possible to measure the defects nor to scan larger surfaces.

Subsequently, 90 specimens of weathered glass were subjected to Coaxial Double Ring tests, with either the external or the internal surface in tension, to assess the effect of weathering on strength. The specimens were of two different dimensions and they were loaded with two different rings to investigate the "size effect". In addition, 81 similar specimens of new glass were tested, with either the tin or the air side in tension, and they were used as a reference. These tests showed that the strength of the 55-year-old glass ranged from 22,9 MPa to 138,2 MPa, whereas that of new glass ranged from 38,3 MPa to 219,2 MPa. In particular, the average strength of the internal surface of the small weathered specimens was 53% higher than the external, but no major difference was observed in the tests of the large specimens. Furthermore, the average failure stress of the air side of new glass was approximately 45% higher than that of the tin side, while in literature that difference was characterised as marginal. Finally, although the size effect was clearly observed in the tests on new glass and in those on the internal surface of weathered glass, the strength of the external surface was found independent of the size of the loaded area. This implies that the defects on this surface are similar and uniformly distributed, so the probability of encountering a critical flaw is equal regardless the size of the loaded area.

Fracture statistics were used to derive the design strength of the glass of each testing series. Among three probability distributions, the Weibull distribution was found to describe the strength data of glass better. However, for low probabilities of failure, the data did not fit well to the Weibull distribution and thus, the resulted design strength values are probably very conservative.

The information collected during the non-destructive tests, namely the size and the shape of the defects found on the surface of weathered glass, was used for the identification of the critical defect which will initiate failure. At this step, the assumption that the largest defect will be the critical one was made. Then, the theory of LEFM was used to relate the size of the found critical defect to the failure stress, through the critical stress intensity factor KIC and the geometry factor Y. The predicted values for the failure stresses were analysed with fracture statistics and compared to the actual failure stresses of weathered glass, as emerged from the Coaxial Double Ring tests. In this way, the accuracy of the predictions was evaluated.

The effectiveness of the methodology for the detection of the critical flaw was evaluated through post-fracture analysis. This analysis showed that the critical defect was successfully detected in the 23% of the examined specimens. This suggests that the visually largest defect on a glass element could be the critical one. Furthermore, even if most of the specimens did not failed at the measured defect, the average predicted failure stress differ by less than 9% from the average actual failure stress. This outcome suggests that the defects induced by weathering on the glass surface are similar. Thus, a defect which was identified as critical for one specimen but eventually it was not, it was probably very similar to the critical defect of another specimen.

To conclude, this thesis examines and proposes a novel methodology for the prediction of the glass strength, based on LEFM, applied to defects found on the surface of naturally aged glass. If the defects on the glass surface are studied according to this methodology, a good estimation of the average failure stress can be obtained. For lower probabilities of failure the methodology gives conservative estimations, so it has a potential for use in design applications. Overall, the defects on the surface of glass, and especially their size and geometry, appear to have the greatest effect on the strength of weathered glass so far. Further experimental investigation on the parameters which affect the strength of weathered glass should be carried out, as well as further research on the automation of the proposed methodology, which is expected to increase its effectiveness and efficiency.