Volume 2, Issue 1 — Year 2025 — Article e100029

Amin Danesh¹; Sibel Açık²; Sun Jing³; Mohamadreza Esmailzadeh⁴

1. Department of Civil Engineering, University College of Science and Technology, Urmia 5735133746, Iran
2. Department of Engineering Science and Architecture, Istanbul Arel University, Istanbul 34537, Türkiye
3. Civil Engineering Department, Guangxi Polytechnic of Construction, Nanning 530003, China
4. Department of Civil Engineering, University College of Science and Technology, Urmia 5735133746, Iran

Cone Penetration Tests (CPT) offer valuable insights into the physical properties of soils that cannot be directly obtained from recovered soil samples. This information is crucial for understanding the geotechnical characteristics of soils, including their layering, thickness, stiffness, strength, and consolidation behavior. Practically, CPT is conducted to depths of up to 10m, depending on project requirements. The results are interpreted to differentiate between granular and cohesive soils by analyzing cone resistance and shaft friction measurements. These interpretations are then applied to classify soils using established empirical classification charts or tables, ensuring accurate and reliable geotechnical evaluations. The presented study aimed to develop a machine learning-based model utilizing the Support Vector Machine (SVM) algorithm to classify soil types based on localized CPT data. The approach leverages the high-resolution data provided by CPT, including cone tip resistance, shaft friction, and, where available, pore pressure measurements, to accurately identify and classify soil behavior. By tailoring the model to local soil conditions, the study seeks to enhance the reliability and precision of soil classification, overcoming limitations of traditional empirical methods. This research not only demonstrates the potential of SVM in geotechnical applications but also provides a framework for integrating advanced computational techniques with localized geotechnical datasets for improved decision-making in soil characterization and infrastructure design.

CPT classification, Machine learning, Localized analysis, Geotechnical engineering, Soil behavior