Dirk Uhde isn’t the kind of guy who talks a lot about software. So when a team of data scientists and software developers from GE Digital’s European Foundry met him amid the sparks, grime and welding equipment of the industrial plant he manages in southeastern France, he struggled to follow their technical parlance. When he tried to explain how his machinery worked, they didn’t have an easy time either. Yet it was the most productive meeting he’d had in years, he now says. It brought together two very different worlds: the traditional industrial world and the digital world that’s changing how we make things.
Uhde’s plant in Aix-les-Bains, a thermal spa town in eastern France, builds components for gas-insulated substations, which transform high voltage on the electrical grid. As part of the process, he cuts and assembles aluminium pipes for substations working around the world. These pipes come in two sizes: 9 and 11 meters long, and the plant makes about 20 kilometers’ worth per year for these substations.
Uhde’s challenge was really a math problem. The small team that handles pipe cutting must make on average about 600 cuts to fulfill orders each month, or 150 cuts per week. The problem for years had been that about 10 percent of those cut pipes ended up wasted. That’s not because the workers make mistakes but because of the challenge of calculating exactly how to cut up the standard pipe sizes to leave close to zero remnants. It’s the same problem that comes with figuring out exactly how many floorboards you’d need to put down for a room and how best to cut them to avoid squandering your material. The plant can sell the leftover pipe pieces as scrap, but for less than 30 percent of their wholesale price.
“The process used to be completely manual,” Uhde says. One worker would spend at least an hour making calculations and trying to optimize the sequence of cutting, including how many pipes were available to cut, what their sizes were and what sizes were needed by the end of the day. “We’d done it like this for 30 years and tried to find improvements,” he says.
Then in late 2016, GE’s European Digital Foundry contacted Uhde as part of a pilot project to see whether industrial businesses could use software to improve performance. When Vincent Champain, general manager of the European Foundry, arrived from Paris at the plant along with his team of four other digital mavens in January 2017, he had grand plans to fix things. But the two worlds, manufacturing and software, found it difficult to communicate with each other. “My team was looking at them and saying, ‘This cannot work,’” Uhde recalls.
Understanding emerged when both teams put on helmets and safety goggles and moved down to the factory floor where workers were cutting and welding aluminium pipes. For some of the digital team, this was the first time they had actually been on a manufacturing floor. “It was amazing,” says Igor Dniestrowski, a project manager for app development at the Foundry. “This is the moment where we saw the field reality, the production process as well as the people dedicated to their jobs. It is when the concept becomes a true digital industrial application.”
For example, being able to see how heavy the pipes were and how difficult it would be to shift them from the bottom of a pile to elsewhere in the production hall helped the digital Foundry team to start improving the cutting process.
The most efficient method would be to test all the possible solutions first. But that would take a huge amount of time and processing power for any computer. The team figured out there were more ways of making those 600 pipe cuts each month than there were atoms in the universe — roughly 4×1079 possibilities, Champain says.
Yet calculating all those options was possible thanks to cloud computing and Predix, GE’s digital platform for the Industrial Internet. After their tour of the plant, Champain, Dniestrowski and the others went back to their office in Paris. Within just four weeks — by the end of February — they had built an app that calculated how and where to cut each pipe at the Aix-les-Bains plant for maximum efficiency.
Now Uhde uploads his daily pipe-cutting requirement to the app, and “it’s just a few seconds before Predix spits the results out,” he says.
Crucially, the app has reduced scrap waste from 10 percent to 4 percent, which should save the plant an estimated $200,000 this year alone.
Down the line, the app can be revamped to cut waste in other parts of the plant too. Developers at the European Foundry are now working on an algorithm that can suggest the most efficient way to pack the finished pipes into shipping crates. Dniestrowski says it should take about a month to upgrade the app, which Uhde estimates could double or even triple the annual savings from the pipe-cutting algorithm.
“Running a plant, you try to improve productivity all the time,” he says. “And then comes a tool like this which is a quick hit. The key was to bring an expertise into the process which we didn’t have, which was digital.”