Performance assessment of rotary drilling using non-linear multiple regression analysis and multilayer perceptron neural network

Cost and efficiency estimation for rotary drilling rigs is an essential step in the design of excavation projects. Due to the complexity of influencing factors on rotary drilling, sophisticated modeling methods are required for performance prediction. In this study, rate of penetration (ROP) of a rotary drilling machine using two developed modeling techniques, namely, non-linear multiple regression models (NLMR) and multilayer perceptron-artificial neural networks (MLP-ANN) were assessed. For this purpose, field and experimental data of various case studies were used. Several performance indexes, including determination coefficient (R-2), variance accounted for (VAF), and root mean square error (RMSE), were evaluated to check the prediction capacity of the developed models. Considering multiple inputs in various NLMR models, the most influencing factors on ROP were determined to be brittleness, rock quality designation (RQD) index, water content, and anisotropy index. Multivariate analysis results of developed models showed that the MLP-ANN model indicates higher precision in performance prediction than the NLMR model for both the training and testing datasets. Additionally, sensitivity analysis showed that RQD and water content have significant influence on the ROP. The models proposed in this study can successfully be applied to predict the ROP in rocks with similar characteristics.