Application of Machine Learning for Prediction of Cone Penetration Test Data

Abstract

The article discusses the use of machine learning to obtain predictive data of cone penetration tests of soils using the example of studying the geological environment of the Kazan city. The cone penetration test is one of the most popular and widely used types of field tests of soils. As a source material, a database of the physical and mechanical properties of soils and the database on cone penetration tests of soils of various genesis were used. The territory of the city of Kazan was considered separately for each terrace of the Volga river in accordance with the basic principles of engineering-geological zoning. For modeling of the dependence of parameters of cone penetration tests on soil properties was performed using machine learning. Mathematically, the problem was formulated as regressive - need to find functions that closely approximate field test data. As a result, it was found that the dependences obtained as functions of the cone resistance qc have very good convergence. This shows a good prospect of using machine learning to obtain real correlation dependencies, and therefore, to reduce the cost of engineering-geological surveys.