Using machine learning methods for prediction of drilling rate: case of water drilling operations in gneiss rock

Abstract

An accurate model for ROP prediction in gneissic rock formations was proposed in this study to help reduce drill costs, total drilling time, and allowing companies to be more competitive. The developed ROP prediction model has considered for (04) parameters: percussion pressure, blowing pressure, pressure of compressor, rotation speed. Pressures values range between 5 and 294.3 MPa, while the drilling time and drilling depth ranges from 1.17 h to 44.21 h and from 1.8 m to 4.6 m respectively, for a constant rotation speed of 1 526 tr/min. Multi-Layer Perceptron, Multiple Regression, K-Nearest Neighbors, Ridge Regression and Random Forest have been used for training and validation tests. Comparisons of results based on the R2 shows that the best model is obtained through the Random Forest method, with an R2 of value 0.974 while the second-best method is the K-Nearest Neighbors method, with an acceptable R2 of 0.737. Related RMSE and MAE obtained from both methods are relatively low, with values of 0.0794 m/h and 0.254 m/h for RMSE, and 0.0123 m/h and 0.041 m/h for MAE respectively.