To date, solutions for drilling automation have focused on assisting functionalities to the drilling process by the mean of fault detection, isolation and recovery (FDIR) functions, safe operating envelopes (SOE), control methods applied to a few parameters, and automated standard procedures. Yet, the decision to execute different functions has been left to a human operator. In this paper, we address the next level in drilling automation: a method to achieve autonomous drilling.
Autonomous drilling requires that the software solution is capable to adapt itself to unexpected situations, to learn both online and from experience, all this in a timely fashion and with incomplete information. Our approach to autonomous drilling is based on the minimization of a "cost" function that incorporates both performance evaluations and risk assessments, in order to reach a certain goal. Here, our objective is to minimize the drilling time to arrive at the section total depth (TD). This time evaluation also includes possible delays that could be incurred in case of drilling incidents.
Since the drilling system is non-holonomic, it is necessary to consider all past states in order to evaluate the current state of the system. Thus, the method relies on the propagation of uncertainties for the evaluation of the current situation and foreseeable alternative developments of the drilling operation to reach the section TD. Therefore, the availability of well distributed measurements in space and time, with acceptable accuracies, allows for driving the drilling operation more aggressively than in the case of sparse information of poor quality. On the other hand, when the drilling downhole conditions start to deteriorate, the autonomous drilling function applies, by itself, remedial procedures to keep the time to TD as low as possible.
The novel autonomous drilling method is built on top of existing drilling automation solutions. The necessary adjustments to these systems are also described. A virtual test rig environment that is a replicate of an actual full-scale test rig, is used to develop and qualify the new autonomous drilling solution, before a final demonstration on an actual full-scale test rig.
The automatic management focusing at the same time on short, medium and long-term horizons when balancing between performance and risk is the key factor to obtain a software solution that adapts itself to unexpected situations. Concomitantly, the use of systematic error propagation enables to learn from past measurements what are normal and abnormal states of the drilling system, therefore fulfilling the main requirements of an autonomous system.