HOW AI IS MAKING NONPRODUCTIVE TIME IN O&G A THING OF THE PAST
Weatherford International has come a long way since its founding in Weatherford, Texas in 1941. Now employing some 30,000 people in over 90 countries, the company has become one of the largest multinational oilfield service companies delivering innovative technologies and services to meet the world’s current and future energy needs. The oil industry likely brings to mind old images of grimy rigs and manual labor. These days, the field requires sophisticated analytics tools far beyond a clipboard or field laptop. Weatherford provides its customers in the O&G industry with services to maximize safety and efficiency at every step of the well lifecycle—from construction to drilling to production. This can involve all types of hardware, a software platform for well monitoring and performance evaluation, or building digital models of drill sites.
One of Weatherford’s leaders in charge of deciding where to go next, for both digital enterprise and market, is Philippe Flichy, Weatherford’s Strategic Business Development Director. Flichy presently develops Weatherford Digital Enterprise offerings, mixing IIoT (Industrial Internet of Things), analytics, big data, physical simulation modeling software, process workflows, and services and hardware to better support Weatherford’s customers.
Flichy immigrated to the U.S. after his family was targeted by communists in his native France (for this reason, he notes how the American flag holds special meaning to him). Prior to Weatherford, he managed the internet feed and intranet systems for the 2002 Winter Olympic Games, which, given the security concerns at the time, meant working with dozens of FBI agents to monitor for cyber attacks.
This experience, combined with his decades of executive leadership in bringing modern technologies to remote oil fields, shaped his perspective on cybersecurity, and the application of advanced technology to solve the O&G industry’s biggest challenges. Flichy made a name for himself in the O&G industry by authoring papers showing how isolation was causing the oil field to lag behind on incorporating modern technology, and documenting how a holistic view of the enterprise could yield powerful results. His findings are supported by credible sources: a 2016 study from MIT Sloan Management Review and Deloitte ranked the O&G industry’s digital maturity among the lowest, behind even agriculture and construction.
Operations in O&G are rough. Though oil will flow freely from a well at first, it soon needs more coaxing to come to the surface, which requires both hardware, software, and complex physical models to understand and optimize the extraction. “I’m not going to go into Petroleum 101,” Flichy says, “but typically what comes out of those holes we’re drilling is pretty nasty. All the equipment is subject to harsh treatments that are way beyond what the worst driver would do to a car.” Hardware failures are a big, expensive mess (a single pump failure can cost $100,000–$300,000 per day in lost production), and can have severe environmental consequences.
To safeguard against nonproductive time, or downtime, much of the O&G industry employs preventative maintenance, a regularly scheduled program which makes it easier for maintenance managers to improve uptime and save money on expensive equipment repairs. These preventative measures ensure machinery stays in working order, a necessity in the O&G industry.
The traditional preventative maintenance approach in the O&G industry has forced maintenance managers to spend most of their time running from one crisis to the next. A regularly scheduled maintenance strategy is often an inefficient maintenance technique. Industry experts claim preventative maintenance programs cost more than a strategy in which machines run to failure. In fact, researchers at The Electric Power Research Institute (EPRI) have calculated maintenance costs on a 1,500-horsepower motor (e.g., in an oilfield pump): a traditional scheduled maintenance strategy costs approximately $36,000 per year, while a reactive maintenance strategy costs $25,500 per year.
The maintenance cost of the individual asset, or the difference between strategies, may seem minimal. However, when multiplied by the number of assets across an entire fleet, the cost skyrockets. For a fleet of 1,500 assets, the expense balloons to $54M per year for a scheduled maintenance plan and $38M for a reactive one. And these costs only account for maintenance—they don’t include the true loss of a machine failure, which is in lost production. For O&G, equipment uptime is directly correlated to the company’s bottom line. When drilling stops at a well site, cash flow associated with the well stops. Therefore, reliability is critical.
Fortunately, predictive maintenance, powered by artificial intelligence (AI), holds the key to better detecting equipment failures long before they become costly catastrophes. Those same EPRI researchers also concluded that a predictive maintenance strategy is the most cost-effective to eliminate the risks of secondary damage from catastrophic failures.
Weatherford is aware of this reality. Given the difficulty of accessing and repairing equipment (most oil rigs are in remote areas), a major area of differentiation for the company is their ability to use a predictive maintenance strategy to determine when their machines will fail. This approach provides Weatherford enough advanced notice to make the necessary preparations to fix impending problems with as little downtime as possible.
One of Weatherford’s many solutions, a software platform, includes an element which relies on data from the edge (at the source of the data, i.e., the sensor). This platform paints a complete picture for operators by using data from daily production, mechanical sensor recordings, and static well properties to determine if maintenance is necessary or if operating conditions can be adjusted to alleviate the problem.
Flichy notes how, in addition to analyzing the data, Weatherford can act as a consultative ally to their customers: “Some companies have the time to do all that. Some don’t. And that’s when they can press the button in our software saying, ‘Weatherford, help! There’s a problem here. I don’t have time. Please look at it.’”
Thanks to these edge analytics (i.e., data collection and analysis at the sensor as opposed to a centralized data storage), Weatherford is rolling out a software platform to check the well from conception to the end of its production. The ability to fully understand your equipment—to have access to data analytics which guarantee asset life expectancy and control maintenance needs—is a huge advantage when repairs require crew to travel long distances and work in dangerous conditions.
Given the volatility of the O&G industry in recent years, compounded by an aging workforce (as many as 60% of current field experts are expected to retire over the next six years), efficiencies to make full use of employees’ time are vital. Flichy sees this as an opportunity to move from predictive to prescriptive maintenance:
If I start understanding where things are going to fail, I can be more organized,” he explains. “I can say, ‘maybe it hasn’t failed yet, but since I have a rig very close by and I have the crew and everything, go ahead and service that one now.’ It allows us to be nimbler and start really planning things.
Reducing maintenance workovers has decreased nonproductive days by as much as 50% (as estimated by Flichy and the Houston Chronicle) which has a direct correlation to ROI. “We are a service company,” Flichy points out, “and the ROI we are looking for is what we can help our clients to achieve.”
So how can Weatherford engineers analyze data to make a decision on maintenance needs? Flichy offers, “Good ol’ Excel,” quickly adding, “The thing that you have to understand is when you reach a certain level with the petroleum engineers, it becomes a mix of high-level science and art. But analytics is not alone: it’s analytics, it’s big data, it’s IoT. That’s the digital enterprise.” Flichy believes the new industry tools are changing the way the industry works, moving from reliance on spreadsheets to decision making based on centralized big data to increase accessibility and trust of the data.
This is also an area where AI can come into play—augmenting the knowledge of a production engineer. Previously, technology took the legacy approach to calculate mean time between failures. More sophisticated AI systems, like SparkCognition’s SparkPredict, can collect the data from the well, monitor daily production and sensor recordings, and apply supervised learning techniques to pinpoint and predict when a site will need maintenance as well as its production output post-workover, making it easy to demonstrate ROI.
However, there are still challenges to data access. The specialized electronics required to work at 400°F temperatures are expensive, and the quality of data obtained in the “noisy” environment of magnetic fields is often called into question. Newer, more sophisticated sensors can self-diagnose and self-correct in real time. Business Intelligence estimated the number of devices at well sites will increase at a 70% compound annual growth rate. Flichy hopes the future of IIoT and edge analytics will help differentiate data by controlling information sent to central repositories versus data saved locally.
Choosing a company for their analytics proved to be a challenge for Weatherford. The field is lacking in experts on O&G, so Weatherford sought out adjacent experience (for example, previous work with wind turbines) and the ability to hear problems out rather than jump to a quick diagnosis. Those firms were given a sample data set to work with, and results were compared in terms of downtime, production, and ROI. SparkCognition led the pack and now works with Weatherford to deploy AI-powered analytics for their customers.
Flichy ends with words of advice to those who are skeptical of cognitive analytics: “I think it’s like keeping a horse carriage when the car is coming. The first cars were not very reliable and not going much faster than the carriage. But pretty quickly, it became absolutely obvious that cars had an advantage. I think we’re exactly in that same intersection point with cognitive analytics.”
“The big vision is that we’re going to have an increase of efficiency of the people we have because of workflows and anticipation. An engineer can look at data and make decisions faster, better, and safer.”
Last modified: November 22, 2017