ConeVersion: A Software for Cone-Penetration-Test Inversion Driven by Machine Learning of Calibration-Chamber Experiments

This project presents ConeVersion 1.0, a machine learning-based tool for correcting volume-averaging effects in cone penetration test (CPT) data, particularly in layered soils. These corrections are essential for improving the accuracy of geotechnical applications such as liquefaction hazard assessment and foundation design.

The model is trained on a curated database of 259 physical and virtual calibration chamber experiments simulating CPTs in homogeneous and layered deposits. This comprehensive dataset enables the model to provide more reliable estimates of true stratigraphy from measured CPT measurements.

Reusability: The software can be used to predict true CPT data from measured values for any and all CPTs conducted worldwide.

Uniqueness:
This is the first open-access CPT inversion tool developed using a large, hybrid dataset of both physical and simulated calibration chamber tests, enhancing generalization and reliability.

Audience:
Geotechnical engineers, researchers, hazard analysts, data scientists, and educators focused on subsurface characterization, soil modeling, and machine learning applications in civil engineering.