Drilling automation

A Big Question for Digital Experts: What Is the Driller Trying To Do?

“What is the drilling state” has become an important question among data scientists and automation experts. The simplest definition of a complicated concept is that it is what the driller is doing at the time.

Man in command center with monitors.

Digital drilling experts spend a lot of time wondering “what was the driller thinking?”

They are not being sarcastic. The question matters for those doing analysis or writing control algorithms because the significance of readings, such as level of torque at any moment, depends on what the driller is doing at the time.

The interpretation is different if the rig is drilling, where significant torque is required, or reaming, where the resistance is likely minimal.

“During the reaming process, a spike in torque indicates something altogether different from a similar spike while drilling. So, the machine must recognize at least these two states: drilling and reaming,” said Fred Florence, a drilling consultant.

He is among a group of drilling automation advocates who put rig state on a short list of issues that must be addressed to make it possible to create drilling systems where multiple devices can read and react in sync to changing drilling conditions.

In other words, they want to know the information in the driller’s head now, said Moray Laing, director of digital value well construction engineering for Halliburton.

That means an automated device needs to know what the driller is doing and also be aware of concerns that could require quick reactions.

“Unless we can give it the same situational awareness of a human, it will not be able to manage the complexity of the process,” said Darryl Fett, Total’s manager of research drilling and wells in Houston.

Rig state differences also plague those trying to analyze drilling data who need to know what else was going on at the time. They struggle with multiwell data where different methods of calculating the rig state were used.

Drillers leave a record of their work in drilling logs. But this after-the-fact report typically lacks the precise timing sought by digital analysts using high-frequency data to analyze events that can happen suddenly.

“A data scientist working on drilling will tell you that one of their biggest pains is someone will ask them to build a model on lost circulation and here is some data,” Laing said. That analysis is not possible unless someone can offer details about when the fluid losses began, how long they lasted, and other bits of context that might matter, such as the drilling fluid properties at the time.

When asked for a basic explanation of the rig state, Crispin Chatar, drilling subject matter expert for Schlumberger, compared it to bringing a car that has been overheating to the shop. The mechanic will ask what was happening when the trouble began. Did the trouble start while driving fast on a freeway? While stuck in traffic? Did it happen after the radiator fluid warning light went on?

“Every single engineer who works in drilling optimization, drilling analytics, or any one of our remote operation centers uses rig state to quickly and clearly understand what is going on at the wellsite in terms of drilling operations or what the system might be seeing downhole,” Chatar said.

The mechanic also is likely to plug the car into a computer that analyzes data collected by sensors in the car. Rig state service companies also rely on drilling data to diagnosis the state of the system.

Schlumberger has defined 16 rig states—some of which are divided into subparts. Chatar said other companies have systems with 10 to 14 rig states. The company has several ways to calculate the state.

“They are our intellectual property so not open in all cases. Most recently we came up with a very accurate method to do it automatically and with a better accuracy. This latest method is not open,” he said.

State of Confusion

All of this talk seems increasingly distant from the actual work of drillers, who are not known to use the word “state” when describing what they do. “I have yet to find one that is concerned with rig state. I am not kidding,” Laing said.

Among those who see the definition and diagnosis of rig states as a vital concern are members of the interoperability subcommittee within SPE’s Drilling Systems Automation Technical Section (DSATS).

It is one of the barriers to reaching the group’s goal of “open systems from multiple equipment makers and service companies that can be easily integrated into digitally controlled systems,” Florence said.

Initially they highlighted designs that made it hard to physically connect equipment; Fett used the phrase “screw things together.”

From there they moved into the control system where differences in how systems delivered basic commands—the set point and target performance level—made it difficult to interconnect equipment, said Laing.

Frustrations about the lack of progress on these issues led to a symposium held before the 2018 SPE/IADC International Drilling Conference and Exhibition, which ultimately led to the creation of a new industry initiative, Drilling and Wells Interoperability Standards (D-WIS).

Its mission includes creating standards that make it possible to easily combine equipment and control systems from multiple providers to create digital systems. To enable it to do that, it has become part of the Open Group OSDU Forum—an energy industry group created to set open digital standards, beginning with geophysical data. (SPE is not involved in standard setting.)

They are focusing on communication among devices made by different companies. The projects range from sending commands to finding and sharing the data needed to determine what it needs to be doing.

“It is the flow of data into action,” said Fett, program lead for D-WIS, who sees a large upside in being able to easily integrate components made by companies competing on price and performance.

Point of View

The other powerful interest group in this debate are those doing data analysis who struggle to compare data sets using multiple rig state data definitions.

If there are three rigs and different rig state definitions used, “how are you supposed to use that data?” Laing said.

To address the problem, there are companies selling rig state generating software or converting the data into comparable form, but that is like having to hire a translation service to overcome a language difference.

And like language differences, users can make strong arguments for sticking with a different way of doing it.

In a simpler world, the rig state would be determined by using a drilling control screen designed to keep a running record. It might work like a cash register at a fast food place where the clerk pushes a button with a picture of the item ordered on it and that would generate an itemized receipt.

In the real world, Laing said drilling systems on most rigs do not have tags that could make that determination, and the level of detail in tags varies.

The larger problem with a simple definition is that when experts talk about determining the rig state, it is a lot looser concept than a quarter pound burger with mustard, pickle, and onion.

The analysis of the rig state, or the drilling state, varies based on the user’s end goal, the calculation method used, and the available data, Chatar said.

He recently delivered a paper on gathering drilling state data by analyzing video shot on the drilling floor and shaker of an automated rig being tested. It proved useful in helping to fill in gaps in sensor data when the rig state does not include drilling or tripping.

Rig state analysis may not add much when analyzing every drilling performance. “But as we move closer to using data for automation and feeding into digital systems, the granularity becomes critical,” Chatar said.

A key member of the D-WIS team seeking a standard rig state definition is Eric Cayeux, chief scientist for drilling and well modeling for the Norwegian Research Centre (NORCE), whose SPE papers are must reading in the field.

In an email he described rig state in several ways.

  • There is the physical state based on such things as its position or speed, which can be measured in various ways.
  • It can also be described as a dynamic system using differential equations from control theory, expanding the list of possible variables.
  • Or computer science can be applied to define the state based on treating the process in terms of finite state automaton.
  • The level of abstraction can vary, ranging from low (the physical description of a specific mud pit) to high (the rig is drilling a specific section of the well).

There are so many options, each with pros and cons, that the D-WIS committee decided not to start by creating a drilling state standard.

“We realized that it was a huge and complex task that we were not sufficiently prepared to address,” Cayeux said.

Instead they are taking a step in that direction by promoting a system to enable rapid, accurate data exchanges using “semantic” standards. This approach could enable software applications to automatically seek out the data from machines that would announce what they have available (SPE 194110).

This analysis will not determine the state but will allow machines to share information without human intervention, which can create a common view of the current situation.

In a paper written for Energies, Cayeux wrote about the need to get the most out of limited available information to get a sense of the rig state for better decision making.

This semantic approach, which could be proposed as a standard, is designed to minimize the time-consuming role for people in the process by establishing standards for machine-to-machine exchanges that ensure the right data are quickly gathered.

The format ensures that the data grabbed address the concerns of a careful engineer, who would insist on knowing if the data point is a measurement or a device set point, and if it is derived from a measurement, how it was done.

To facilitate quick exchanges, data would be available in predetermined packets based on common uses.

To explain what is being done, the paper turns to cooking analogies. There are sequences of instructions called recipes. The International Society of Automation (ISA-88) offers a recipe for improving the quality of tasks done in batches, which are repeated activities that are not done often enough for mass-production methods.

Cayeux’s semantic system addresses the problem like a restaurant that prepares for the dinner rush with pre-cut ingredients and spice mixes. This allows cooks to quickly combine predictable inputs to keep up with the unpredictable flow of orders by demanding customers.

This is short of defining the current rig state, but it would provide context for analysis and allow control systems to coordinate multiple components while drilling.

“Afterward, we will likely come back to the question of rig states, hopefully better armed after the experience gained from the semantic of drilling real-time signals,” Cayeux said. JPT

For Further Reading

SPE 194110 Toward Seamless Interoperability Between Real-Time Drilling Management and Control Applications by E. Cayeux, B. Daireaux, N. Saadallah, Norwegian Research Centre, et al.

Autonomous Decision-Making While Drilling by E. Cayeux, B. Daireaux, A. Ambrus, Norwegian Research Centre, et al.