How to model rate of penetration using logging data

This paper presents a systematic approach to building empirical ROP models from conventional and routinely available mud logging data. The method employs simple statistical methods to decrease the contribution of errors inherent in mud logging data, thereby overcoming one of the major impediments to successful ROP modelling of field data. For a given bit run, the approach generates a model relating ROP to applied drilling parameters without presuming the form of the relationship in advance. As a result, the model can be demonstrated to be physically valid under the conditions from which it was derived, a critical consideration given the varying contribution of factors affecting ROP under different drilling conditions. The validity and robustness of the new approach are demonstrated by applying it in a number of case studies, including both rock and PDC bits in a range of formations.