| Enrico Zacchei, Zhiyuan Wang Clustering-Based Post-Processing of Experimental Data for Double Glass Units |
|---|
| Abstract. Double glass units (DGUs) consist of glass plates separated by a gas-filled cavity sealed by perimeter spacers. The glass plates resist to applied loads, whereas the enclosed gas mainly contributes to thermal and acoustic insula-tion in buildings or other structures. Load transfer between the plates occurs through a load-sharing mechanism that is difficult to be quantified, particu-larly regarding its influence on vertical deflections of the glass. Although this phenomenon has been recently investigated experimentally, a major chal-lenge of laboratory testing is the post-processing of large outputs. To address this issue, this study employs the k-means clustering algorithm as a data-driven tool for experimental data analysis. The objective is to identify opti-mal clusters that provide accurate approximations and enable direct relation-ships between elastic stiffness and deflections. The results show that k-means clustering produces reliable stochastic approximations, allowing the genera-tion of deflection maps for both the upper and lower glass plates. Compared with traditional manual or analytical post-processing methods, the proposed approach offers an efficient and robust alternative for estimating stiffness values and glass deflections. |
| Keywords: k-means; DGU; mechanical deflections; structural analysis |
 Download PDF |
| DOI: https://doi.org/10.54381/itta2026.4.09 |