Correlation of Mechanical Parameters in Diatomaceous Soils Using Image Analysis and Digital Processing

In the state-of-the-art, no studies are recognized regarding analyzing or processing images in diatomaceous soils for their subsequent relationship with mechanical parameters. These soils report responses and behaviors that are unusual and distant from conventional soil mechanics. This research included a diatomaceous soil of Colombian origin with cylindrical particles, in which unconfined compression, direct shear, and oedometric compression tests were carried out, and Scanning Electron Microscopy images were obtained. Analysis techniques and filters (Binarization, Voronoi, LUT) were applied to these observations. This research aims to determine if, from the images and the mechanical response results, some correlation can be established that subsequently serves as a predictive tool for the behaviour of this type of soil. The geotechnical response values are based on the study variables (diatomaceous soil content and deterioration cycles). Proportional trends that are almost linear in behaviour are identified between test values. The results of final void ratio, cohesion (kg/cm2), and strain at failure (%) obtained through laboratory tests and those calculated from image analysis are presented. A multivariate analysis was performed, and a quantitative image-based index was proposed to correlate cohesion parameters, friction angle, undrained shear strength, and strain at failure.