Application of Machine Learning algorithms to drilling data for deep foundations and soil treatments

Modern machinery for executing deep foundations and soil reatments is equipped with sensors that allow the measurement of execution parameters such as torque, thrust, and speed (Measurement While Drilling, MWD). These parameters are elated to soil characteristics and, by using machine learning (ML) algorithms, it is possible to translate them into a penetrometric soil profile. Therefore, each perforation can provide analogous information of that obtained from a penetrometer. However, in order to apply these algorithms in the specific case of pile drilling and soil treatment, several factors must be considered in both ML techniques and specific to geotechnics. These factors include the presence of anomalous values in the data, the distance between the enetrometers and the drilling points or the selection of the most appropriate algorithms, among others. This paper proposes a specific methodology that outlines the various steps equired in order to apply ML algorithms to drilling data for deep foundations and ground treatments.