Every evening after building activity stops and all is quiet at a SAGD module manufacturing facility in Alberta, a robot named Franchesca will soon wheel out of storage and begin automatically scanning the partially built units and the parts and tools scattered around them.
It will then create an exact digital replica of the modules’ physical reality down to the millimetre scale in a format that allows comparison of the physical objects against digital renditions like computer-aided design (CAD) models—to make sure everything lines up as intended.
This “model of truth” will ensure that when the various components are assembled in the field, at remote wellsites in the oilsands region of northern Alberta, companies can avoid the most common cause of capital project cost overruns—rework.
“What we are tackling is the big pain point in capital projects. On a global scale the average capital project overruns its budget by about 30 per cent,” said Steve Fisher, co-founder and chief executive officer of Veerum Inc., developer of Franchesca. “On the global scale, it comes in to about $1.6 trillion in unbudgeted wasted capital on an annual basis. We are going after that opportunity—that’s our focus.”
Veerum is one of nine companies with offices in Zone Startups Calgary (ZSC), an accelerator at GE’s Customer Innovation Centre (CIC) that supports the growth of industrial Internet and energy related start-ups. The start-ups receive support to establish market validation and launch field trials, and have access to mentors and to GE customers around the world that could benefit from their technologies. In more than a year in operation, ZSC has created more than 75 jobs, according to GE, which is working to bring more start-ups into the downtown Calgary hub.
“We are here to have an impact on the provincial ecosystem for innovation,” said Gandeephan Ganeshalingam, CIC leader. “Our claim to fame is speed. From the time a customer comes here, to the time we have a solution in the field can be as short as three months, and that’s because we never try to do anything ourselves. We try to partner up very quickly with a local ecosystem, find the best solutions and get it out to market.”
With access to GE’s global footprint, its customer base, product line teams and market insights, technology developers can build something locally and scale it globally, he said.
“As innovators and as engineers, and as entrepreneurs, you get to validate your commercial and your technical leap of faith assumptions very early on in your spend cycle, so when you actually commercialize a product, it sells, and it’s used by industry.”
Bridging the physical and digital worlds
Veerum is well on its way to commercialization, having separately verified all the technologies that go into Franchesca in previous pilot projects with various oilsands producers, including Husky Energy, Cenovus Energy and Suncor Energy. It is also bringing to market virtual reality technology that takes the digital twin to three dimensions, viewed on goggles that can be hooked up at any internet connected location around the world.
Fisher said that by bridging the physical and digital worlds in project development, tremendous savings are available. “What we find on capital projects is, you have your physical reality, your nuts and bolts and your dirt, and that’s what’s real, and then you have your plans and visions and your schedules, and that is what you are trying to make. The issue has been, those things don’t always line up. The challenge is that you are trying to compare apples and oranges right now—the physical reality is real and the digital plans are something different.”
When components don’t line up properly, and require additional work like a cut and weld in the field, the cost can run up to 10 times higher than fixing at the point of manufacture, he said.
“Once you have that digital twin you now have something that we can compare directly against the actual digital plans, schedules, CAD drawings, all those things. And we have an AI [artificial intelligence] engine that takes those two elements, compares them, looks for those mismatches in the virtual world and then instead of just identifying merely the mismatch, it also starts to suggest corrective action items to the project team before those issues become a big problem on your project. Those problems are what leads to that 30 per cent cost overrun.”
In a project with Husky, Veerum digitized 18 modules under construction, Fisher said. “Every one of them showed up on site and fit perfectly, and they finished the project two months ahead of schedule.”
Veerum is currently working with Cenovus, where it is putting sensor-laden Franchesca to work at a Cenovus modular construction yard. “Previously, they had guys out there every Friday with tape measures and cameras [to assess construction accuracy]. We’ll be doing it by robot. It will come out of its house every night, cruise around their module site and digitize those modules, so then in the morning the project manager can come in and can see, compared to the CAD, exactly where every module is and ensure everything is being built according to the plan.”
Fisher said pilot projects have indicated savings of 20-50 per cent, accomplished by virtually eradicating rework. With its recent work on Franchesca and virtual reality, Veerum is transitioning from a project-by-project focus to creating a product that is scalable, and that GE can help take global through its CIC, he said. Going forward, Veerum is focused on major energy infrastructure projects of any kind, including hydroelectric and renewables projects, as well as the oil and gas market, he added.
Already a commercial technology arising out of the CIC is GE’s Steam IQ, a computer program that helps oilsands companies optimize the steam-to-oil ratio and reduce greenhouse gas emissions. A sort of adaptive intelligence app for steam optimization, Steam IQ leverages machine learning, a branch of artificial intelligence that evolved from pattern recognition and cognitive learning. Algorithms can quickly adapt to new scenarios, learning from data to predict and optimize outcomes, according to GE.“
Customers may have several constraints across the whole field—it may be steam, it may be on the production side, or not having enough capacity to process produced water, or enough shipping capacity,” said Warren Gieck, production optimization leader at the CIC. “In the case of steam [constraint], you may want to know how to best allocate your steam for the most optimal operation.”
In that case, all the inputs into the field operations are used to create a highly accurate model, up to 98 per cent accurate, which can then be manipulated to gain actionable insights.
“The optimizer takes all the individual well models and runs them through a complicated optimization routine, but that only works if you can believe the well models—if the well models are only predicting 50 per cent accuracy, then you have no accuracy to base it on. But if your model is 98 per cent accurate, that means you can start to tweak the inputs to the model and see how that changes the outputs, and do that across all wells and find the right configuration,” said Gieck.
What machine learning can do that people struggle with is to simulate and optimize across an entire field, Gieck said. An engineer uses experience and knowledge to iterate to the best solution on a well pair, and it could take one person days or even weeks to get to the optimized state. “People use rule of thumb or negotiate right now because they know how to operate their wells very accurately, but to operate a whole field and optimize is a whole different thing.”
Machines are much better at iterating and the technology has advanced to the point where massive quantities of structured and unstructured data can be ingested and synthesized, millions of hypotheses can be tested, and insights and outcomes can be derived in seconds.
Steam IQ can answer questions around well completions and well operations, such as which wells are most productive under certain conditions and what allocation of resources offer the best outcomes. It can also address equipment optimization, facilities efficiency and marketing and blending optimization, determining the best mix of production targets, for example.
GE has been able to show that it can increase production on average about a per cent on steady state operations, which can add up to millions of dollars in added value annually. In unsteady state, where for example there is a boiler outage or forest fire disruption, GE has shown improvements closer to eight per cent, Gieck said.