The Application of Unsupervised Machine Learning Technique for the Evaluation of Physical Properties of Geosynthetics

Geosynthetics are used in a variety of sectors, from transportation to infrastructure to environmental protection. Despite their importance in civil engineering designs, their chemical and physical similarities can lead to several shortcomings in material selection. This study investigates the ability of Principal Component Analysis (PCA) to detect hidden patterns among seven types of geosynthetics: 1) nonwovens; 2) wovens; 3) knitteds; 4) geogrids; 5) textile-based; 6) geomembranes; and 7) plastomerics. The purpose of this study is to investigate how PCA can assist in mapping geosynthetics to their appropriate engineering applications based on their physical properties. This study revealed that extension at maximum load, mass per unit area, and permeability are the most influential physical properties of the studied set of geosynthetics. These properties predominantly affect several geosynthetic material types such as unreinforced plastomerics (G1), heat-bonded nonwovens (A1), and warp-knitted materials (C2). This enables the systematic mapping of geosynthetics to their most suitable uses in engineering applications.