Real-time rate of penetration prediction for motorized bottom hole assembly using machine learning methods

Drilling rate of penetration (ROP) is one of the most important factors that have their significant effect on the drilling operation economically and efficiently. Motorized bottom hole assembly (BHA) has different applications that are not limited to achieve the required directional work but also it could be used for drilling optimization to enhance the ROP and mitigate the downhole vibration. Previous work has been done to predict ROP for rotary BHA and for rotary steerable system BHA; however, limited studies considered to predict the ROP for motorized BHA. In the present study, two artificial intelligence techniques were applied including artificial neural network and adaptive neurofuzzy inference system for ROP prediction for motorized assembly in the rotary mode based on surface drilling parameters, motor downhole output parameters besides mud parameters. This new robust model was trained and tested to accurately predict the ROP with more than 5800 data set with a 70/30 data ratio for training and testing respectively. The accuracy of developed models was evaluated in terms of average absolute percentage error, root mean square error, and correlation coefficient (R). The obtained results confirmed that both models were capable of predicting the motorized BHA ROP on Real-time. Based on the proposed model, the drilling parameters could be optimized to achieve maximum motorized BHA ROP. Achieving maximum ROP will help to reduce the overall drilling cost and as well minimize the open hole exposure time. The proposed model could be considered as a robust tool for evaluating the motorized BHA performance against the different BHA driving mechanisms which have their well-established models.