Improvement of Drill Bit-Button Performance and Efficiency during Drilling: an application of LSTM Model to Nigeria Southwest Mines

The quarry operators and managers are having a running battle in determining with precision the rate of deterioration of the button of the drill bit as well as its consumption. Therefore, this study is set to find the best-performing model for predicting the drill bit button's wear rate during rock drilling. Also, the rate at which drill bit buttons wear out during rock drilling in Ile-Ife, Osogbo, Osun State, and Ibadan, Oyo State, Southwest, Nigeria was investigated. Artificial Neural Network (ANN), Adaptive Neuro-Fuzzy Inference System (ANFIS), and adaptive moment Estimation-based Long Short-Term Memory (LSTM) machine learning approaches were used to create models for estimating the bit wear rate based on circularity factor, rock grain size, equivalent quartz content, uniaxial compressive strength, porosity, and abrasive properties of the rock. The performance of the models was measured using a new error estimation index and four other convectional performance estimators. The analysis of performance shows that the adaptive moment estimation algorithm-based LSTM model did better and more accurately than the other models. Thus, the LSTM models presented can be used to improve drilling operations in real-life situations.