As drilling engineers, the only tool we have to demonstrate that our drilling project can reach high standards of optimization, is to be able to demonstrate that during the drilling of a well we can maintain a drilling rate within the values that allow us to advance quickly without acquiring compromises that hinder the drilling of the well.
That is why the study that can be done concerning the rate of penetration (ROP) is very important to diagnose what may be the high impact that can have the drilling of an oil well in relation to the total cost of drilling.
Being able to constantly monitor the ROP parameter during the drilling of a well goes far beyond optimization, as it can even serve as that parameter that is indicative of whether we are in danger of a lunge and a possible blowout of reservoir fluids. However, it is not necessary that a staff has to be constantly aware of the readings of this drilling parameter such as the ROP, so it is possible to implement proposals in the application of machine learning models that can make such monitoring activity of the ROP by ourselves.
It is important that the reading of the drilling speed (ROP) would have no sense of measurement and evaluation, if it is not compared with the real-time measurement of other drilling parameters, so the goal I set out in this publication is to discover what drilling parameters can be compared in a real measurement with the ROP in a machine learning mode in the drilling of oil and / or natural gas wells.
It is important for a well drilling engineer to understand that the rate of penetration (ROP) cannot be underestimated as just another drilling parameter that only indicates the speed at which it is varying during drilling, since there are many variables that can drastically and suddenly increase or decrease the drilling rate of penetration (ROP), so being able to decipher abrupt and sudden changes can help us to predict a problem ahead while drilling the petroleum well.
It is a problem to maintain a low ROP, since it implies that we are going to take a long time drilling the well, and that implies high drilling costs, however if we try to maintain a high ROP and without conditions of previous studies can also lead to big problems.
My previous experience during 3 years of work as a well drilling operations engineer, lead me to conclude that it is necessary to reduce as much as possible the incidence degrees in the drilling speed (ROP) as much as possible with the manipulation of external agents to the subsurface geology, i.e. drill pipe string designs, drilling fluid designs, drilling fluid circulation pumping pressure, drill pipe string RPM, bottomhole assembly (BHA) design, hydraulics design, design of the bits to be used during drilling, because if all these external agents to the subsurface conditions are well cared and designed, we can decipher the reasons for a sudden change of acceleration or deceleration of the ROP, and most likely we will diagnose that a problem in such sudden changes in the ROP will have to do with reservoir conditions or the geometry of the same well being drilled.
If we manage to control a correct design of the aforementioned parameters, we could develop automatic learning that could track and possible evaluation of the ROP accompanied by variables that only have to do with the well and the reservoir, however it would not be superfluous an automatic learning that continues to monitor the ROP versus drilling parameters that only have to do with the manipulation of ourselves as those already mentioned above, that to see how much the real expected parameters deviate according to design.
In conclusion, I think that the variables to be taken care of so that they can be compared by a machine learning model with the ROP are the pressures encountered in a well drilling, since a general balance of these pressures could save the future of drilling in relation to the integrity of the well itself and life of the people who operate, since decisions can be made in real time and prevents an onslaught of fluids from the formation into the well and thus reduce the risks of a blowout.
References
Prediction of Rate of Penetration of Deep and Tight Formation Using Support Vector Machine
How to Etimate the Maximum Achievable Drilling Rate without Jeopardizing Safety