Oil and gas operators such as Shell and Oxy are now employing AI together with a vast network of sensors and other machine-learning software to stamp out problems before they happen.
For oil companies whose job is converting barrels to dollars, the enemy is downtime. Downtime can result in lost barrels, and thus lost revenue. Unscheduled downtime is the worst offender. When something goes wrong, and there was no hint that trouble was coming, it can take production offline for longer than otherwise expected. Keeping operations firing on all cylinders is important for both the health of the field and the producer’s bottom line.
Until recently, the best way to keep things humming has been through preventative maintenance. That is, regularly scheduled downtime for examination of equipment for possible wear and tear and potential replacement. The rub on preventative maintenance is that sometimes it is unnecessary. A review of equipment will find everything in order and then you’ve taken your production stream offline for nothing.
Today, the oilfield and other industries are leaning more towards implementation of the predictive maintenance model. Predictive maintenance needs data—a lot of it, to accurately predict when something is going, or will go, wrong. These models can be trained to forecast possible breakdowns before they happen. It does this by employing a network of sensors that takes readings of the equipment. Those readings can then be matched against historical readings to gauge the health of the equipment.
The latest in predictive maintenance employs artificial intelligence (AI) and the Internet of Things (IoT) to gain the most accurate reading possible. This allows for a more proactive approach at inspection, diagnosing any problem, and offering a solution. Predictable downtime becomes more important as the oilfield infrastructure ages and is more prone to service interruptions. A 20-year-old compressor is likely not made to the same design, of the same materials, or employ identical technology as a 2-year-old compressor. All these things must be taken into consideration when employing a predictive maintenance model.
A more recent addition to the predictive maintenance tool kit is the use of a digital twin. Digital twins have become popular over the past half decade or so in drilling complex wells for monitoring and optimization purposes. In predictive maintenance, digital twins can generate data to be combined with sensor data and result in better algorithms for predicting equipment behaviors. By using the digital twin to create equipment fault scenarios, industry can better train programs to look for faults in the actual equipment.
More companies are bringing AI-driven predictive maintenance into the oil and gas space. SparkCognition sees machine learning and AI as the path to a better maintenance paradigm. If a 200,000 B/D offshore platform suffers just 12 hours of unplanned downtime, it can result in up to $8 million in deferred production.
“Predictive maintenance is akin to having a wearable medical device, like a wristband that is constantly scanning a patient’s body, examining every aspect of their health as they go about their day and continuously assessing the results in real time,” according to the company’s website. “This device could then inform the patient that they need to see a doctor for a medical treatment to avert a heart attack they will otherwise have on a specific date. In the same way, the proactive diagnostics of the offshore asset would allow upstream operators to anticipate and mitigate failures before they occur.”
A November 2021 report from IoT Analytics estimated that the then $6.9 billion predictive maintenance market would reach $28.2 billion by 2026. It also estimated that the number of vendors would grow from around 100 to over 500 across that same time.
Shell Focuses on AI-Assisted Safety in Operations
“It’s a form of proactive monitoring … on steroids,” explained Neisha Kydd, operations safety manager for Shell, Gulf of Mexico, to attendees at the recent OTC in May regarding Shell’s deployment of exception-based surveillance. Exception based surveillance has been around for a while, but its marriage with AI and other machine-learning operations has advanced more complex algorithms to predict when something might go wrong. It pulls millions of data points into a single source then applies algorithms to be able to allow users to detect a script or subscript of predesigned anomalies.
“It allows us to have early interventions before we actually have safety incidents,” said Kydd. “It allows us to keep people out of the line of fire and keep them safe. So, for us at Shell, the use of our exception-based surveillance is primarily for the safety of civilians. For us, they to go hand in hand.”
In 2022, Shell conducted a safety review exercise, going back through 5 years of major incidents within the company to isolate common themes or threads that ran across the incidents. One of the findings was that many of the company’s incidents were happening on its auxiliary systems. The company had concentrated the initial uses of AI on equipment it deemed high risk, such as well integrity. However, that same rigor around its air compression systems, instrumentation, flares, scrubbers was not there, and that’s where many of the major incidents occurred.
Another system pattern revealed by the exercise was that the state of equipment and processes in the field was not lining up with expectations back at Shell’s offices. The company had engineers at the office making decisions, writing procedures and sending work to be executed in the field, with a belief that the condition of a piece of equipment was a certain way. What Shell found were gaps between that belief in the office and what was translating to the field.
“So, the question comes, how do we use artificial intelligence to help close that gap so that the real-time data that we get in the field mimics and replicates what the frontline view is actually seeing,” asked Kydd. “That is going to be the next part of our training. That is going to be our next primary focus. We have all these great systems in place, we have all these great minds in place, but if we’re not using them to prevent incidents, then what do we really have?”
Shell’s next step is to centralize its data centers in the Americas. Currently, there are two—The Bridge in New Orleans and The Ark in Trinidad. By uniting the data centers, Shell hopes to achieve standardization of its oilfield processes—key for successful deployment of AI.
Oxy Looks To Push Scale With Help From AI
For Occidental Petroleum, predictive maintenance is about using AI and machine learning to mitigate unplanned events. Generated data are fed into AI models to give predictions based on historical events. This isn’t new to the oil field, but the newer AI models help propel that into this next evolution of predictive analytics and learning, which is the system teaching itself and learning, thus offering better predictions.
“There’s only one problem that we are all trying to solve, and that’s scale,” said Mansoor Nazar, director, enterprise architecture and emerging technologies at Occidental. “How do you scale? How do you scale to a bigger and more innovative approach to be able to do bigger things? To be able to do more things with what you already have. AI is one of those tools that helps you essentially achieve that. Cloud is level one. Of the problems that we’re trying to solve, one of them is unplanned events. That’s the most unfortunate thing that can happen, especially in an offshore facility. So how do you mitigate that? How do you create the predictive analytics?”
Oxy has introduced more robots and drones into its operations to do some of the tasks a human would ordinarily perform. The human can then concentrate on more productive work. There is also a push at the company to employ AI-supported camera technology at facilities to better detect, and predict, corrosion activity—both topside and subsea when it comes to offshore assets.
“What if we could use those cameras to then implement some AI-based intelligence to the system, such as your cameras can look just as good as your eyes,” said Nazar. “Cameras are not smart enough, but if you implement artificial intelligence on top of that so you can detect corrosion through those cameras, or through some other means using AI.”
Beyond scale, the issue of the abundance of data and the guarantee of its quality is another looming issue for the energy industry. Oxy is currently investing in building a data foundation into its cloud environment to keep in front of the massive data growth that modernization and the IoT have brought on.
“I highly recommend that we all should invest in cloud adoption as much as we possibly can, not because we want to migrate from our data centers, but because cloud gives you these capabilities out of the box,” added Nazar. “It gives you a jumpstart on a lot of these opportunities. You don’t have to build any foundational work to be able to do everything that I’m talking about. Cloud gives you that so you can just focus on your use cases.”
The use of AI brings up a lot of questions regarding privacy, ethics compliance, and bias, but there are also technology, IT, and process problems to address.
“It’s not an easy journey,” explained Nazar. “AI is not a product that you can just buy off the shelf and implement. It is something that you must heavily invest in. What we have done at Oxy is form a cross-functional team with subject matter experts from all around the organization. Because it’s not an IT problem. It’s an enterprise problem. You need your legal counsel, you need your supply chain, you need your marketing people, you need your operations people, you need your drilling folks, you need your geoscientists, they all need to come together and come up with whatever use cases you have for artificial intelligence.”