Drilling fee of penetration (ROP) is influenced by many elements, each controllable and uncontrollable, tough to tell apart with the bare eye. Thus, machine-learning (ML) fashions corresponding to neural networks have gained momentum within the drilling business. Earlier fashions had been field-based or tool-based, which affected accuracy outdoors of the skilled area. The authors of the entire paper goal to develop one typically relevant world ROP mannequin, decreasing the hassle wanted to redevelop fashions for each utility.
The authors have recognized a necessity for an ROP mannequin that may suggest parameters in actual time, which ideally requires a common ROP mannequin that may be utilized with low prediction errors.
Formation properties can be found in actual time from logging instruments; nonetheless, incorporating logging knowledge into a world ROP mannequin is difficult.