In April 1969, F.S. Young of Humble Oil & Refining Company published a paper, “Computerized Drilling Control,” in which he described an approach to successfully automate a drilling rig for minimum-cost drilling. Almost 40 years later, one might expect to see his ideas flourishing, but drilling automation remains in its infancy.

In his 1969 paper, Young measured the drilling system’s response to several weight-on-bit (WOB) and rotary speed (RPM) combinations and then used these measurements in a

In certain environments, the BHA can encounter unanticipated formations (i.e., hard calcite nodules), which can result in severe or catastrophic drilling vibration, high local doglegs, and reduced ROP. (Images courtesy of Baker Hughes INTEQ)
minimum-cost drilling function to find the optimum values of these parameters. Surface WOB and RPM were then maintained at these values unless some critical parameter changed, and then the drilling system response would be measured again, and so on, as the system was adjusted for the changing drilling environment. Certain items were ignored or assumed constant. These included wellbore trajectory, drilling dynamics, hydraulics, fluid properties, pressure and lithology. However, the development of an automated drilling system is made difficult by the fact that several of these items cannot be considered constant.

To produce a quality well bore, an automated drilling system must adjust its operating parameters to the variables encountered while drilling. These include the variety of geologic formations encountered downhole. These formations can be difficult to characterize and can possess unpredictable properties that can vary widely across disconformities. As a result, lithology can generate a very complex response to drilling.

For example, in certain geologic horizons hard calcite nodules may be encountered, without warning, while drilling. Sudden contact between the drill bit and these nodules can initiate severe drilling vibration, doglegs, and poor rates of penetration (ROP). As a result, the parameters required to minimize the cost per foot function change drastically at the interface between the two lithologies.

Since the boundary between the two lithologies is first crossed by the drill bit, it is critical that we measure the change in response (vibration, bottomhole assembly (BHA) loads or moments, trajectory, etc.) downhole. The surface response to these changes is delayed by the relatively stiff top boundary condition or even masked (e.g., lateral vibration events are severely attenuated and may not propagate to the surface). Therefore, real-time dynamics systems that measure and classify the downhole vibration severity and the forces and moments acting on the BHA are a critical component of an automated system, along with the surface measurements.

One example of a closed-loop system that operates downhole is a rib-steering system. The automatic closed-loop system can maintain a given inclination and azimuth using directional sensors to measure trajectory and automatically adjusting steering ribs to redirect the downhole tool.

An example of a closed-loop drilling automation system located entirely on surface is the “soft-torque” system described by Javanmardi in 1992. In this torque-feedback system, the varying surface torque is used as a signal to control the rotary speed and diminish the
Figure 2. Generic data flow diagram based on a closed-loop drilling system design by Dashevskiy. In this system, surface intervention is possible via rig personnel in the outer control loop while direct (automated) control of the drilling controls is possible via the inner loop.
downhole rotary speed fluctuations characteristic of stick-slip. If the system is modeled properly and the only drilling vibrations are due to torsional oscillations, the soft-torque system may indeed control drillstring stick-slip. However, “gunning” the drill string through a range in rotary speeds can excite drillstring resonance and the resulting lateral vibrations are often more destructive to valuable BHA components than torsional vibrations. This is indicative of the difficulty of applying an automated system as drillstring dynamics modes (axial, torsional and lateral) are fully coupled making it extremely difficult to address any one mode in isolation.

However, a closed-loop system does not have to be located entirely downhole or entirely on surface. The control input (i.e., steering commands) may be located on surface and transmitted downhole using measurement-while-drilling (MWD) telemetry. This two-way communication, absent when Young wrote his paper, allows user intervention and facilitates more complex wellbore profiles.

In addition to input control parameter measurements (WOB and RPM), closed-loop drilling requires system response measurements as well. While ROP is a significant system response measured at surface, other downhole responses (i.e., BHA vibration and loads, borehole trajectory) must also be measured. In many instances, these data must compete for bandwidth with highly detailed logging while drilling (LWD) measurements. As a consequence, closed-loop drilling systems demand high-data rates typified by newer, high-speed, mud pulse systems or wired-pipe systems. Wired-pipe systems have an additional advantage in that the transmission latency (time delay) is minimized and can be measured. Hybrid systems (i.e., mud pulse run in parallel with wired pipe) offer highly reliable MWD/LWD telemetry which is critical in any closed-loop control system that spans the surface-to-downhole interval.

A joint industry effort recently demonstrated the use of a wired-pipe system in a managed-pressure drilling offshore well. Detailed downhole pressure measurements were transmitted to surface and used in real-time to control the flow rate and back-pressure, and hence the downhole pressure, while drilling a shallow gas sand. It is unlikely that this drilling plan would have been attempted without the degree of automation enabled by wired pipe.
In the decades since Young’s paper was published, the industry has developed an MWD industry — permitting us to automate certain downhole certain functions (i.e., maintaining borehole trajectory) while closely monitoring and controlling the drilling responses. The key link in further drilling automation is the step-change enabled in MWD telemetry systems, be it higher mud-pulse telemetry rates or wired-pipe systems.

Recent publications by Dashevskiy and others indicate that the drilling automation envisioned by Young is indeed feasible on a target well using neural networks assuming certain global knowledge about a field (i.e., drilling responses to changing WOB and RPM controls, etc.) can be captured from offset wells. In this instance, an automation system (trained on offset data) was able to “learn” the correct response to potentially damaging vibrations — providing a significant step forward in developing autonomous systems that are not entirely rule-based. The problem with such an automation system is capturing enough data to adequately define the control space. On real-world wells, the drilling program is tightly tailored by drilling knowledge — restricting drilling WOB and RPM ranges to those allowed by the drilling program. As a consequence, the system may be unable to “learn” about the entire control space.

Another potential approach makes direct use of the known low-transmission latency of a wired-pipe system, coupled with a downhole MWD vibration and engineering measurement system. In this case, the closed-loop system adaptively seeks a local optimum in the control space. For example, WOB and RPM are changed adaptively until a cost function (i.e., vibration severity) is reduced to acceptable limits while ROP is maintained above a minimum limit. The optimum condition may not be “global” (in other words, the best combination of WOB, RPM and ROP might not be realized), but the system does avoid detrimental drilling vibrations. It is envisioned that such a system could be operated in the same fashion as the current breed of “automatic drillers.”

The drilling process has proven to be a complex process to automate, largely because a clear understanding of drilling dynamics is key to many drilling optimization tasks, and a comprehensive closed-loop system demands higher data rate mud-pulse telemetry or low-latency wired-pipe systems. Over the years, our understanding of drilling dynamics, both through drilling dynamics models and validation of those models with downhole measurements, has progressed significantly. This drilling dynamics knowledge is proving critical in the development of a fully automated, closed-loop drilling system in which downhole and surface measurements adaptively control surface weight and rotary speed to optimize ROP while minimizing drilling vibration — the minimum cost drilling system envisioned nearly 40 years ago.

Comments