Foxconn Industrial Internet operates more than 44,000 CNC machines across its global operations. At one point, the company’s standard procedure was to replace the cutting blade in its machines every day.
“We didn’t know when the blade was going to go bad and we couldn’t risk that a CNC machine would go out during the middle of production,” said Richard Vincent, chief business officer of Fii. “A blade is $18. If the machine goes down it would cost us hundreds of thousands of dollars in lack of output.”
But changing out thousands of blades every day isn’t exactly an efficient use of time and resources, so Fii set about studying around 10,000 of its machines using deep learning and artificial intelligence.
The study showed blades could actually last around 14 days on average. Foxconn also discovered a sensor put on the machine would detect a certain frequency when a blade was within a day or two of failing.
“Now we replace the blades every 14 days and we have an alert that if that frequency comes up we know that at the next shift change that particular machine has to be changed,” Vincent said. “We saved a lot of money because now we don’t have to have 40 people changing the blades every single night.”
Vincent said Foxconn’s approach to AI focuses on mixing the domain knowledge of operators and engineers with the capabilities and techniques of data science. Most attempts to apply AI in an industrial setting fail, he said, because data scientists lack an understanding of the specific applications they are trying to improve.
Part of Foxconn’s business model in Wisconsin will now rely on the manufacturer’s ability to help other companies leverage technology for the same kind of outcomes as Fii’s CNC project.
“We’ll teach local companies how to do AI in their environment the right way,” Vincent said. “We’re going to charge for it, but I mean that’s how we’re going to make money, but we make money and they learn how to do things and their manufacturing process is better and they can grow and hire more people.”
Vincent said the model for Foxconn in Wisconsin is reliant on the company bringing new technologies to the region, using those technologies to establish new business lines that in turn grow revenue and then create jobs.
“It works in this order,” he said. “It doesn’t work in the other order. It doesn’t work where you bring in a thousand people and you figure out what they’re going to do. That’s a recipe for disaster.”
Foxconn’s original plans for its Wisconsin campus called for a massive 22-million-square-foot manufacturing campus that would produce the largest LCD screens in the world from a Gen 10.5 factory. That vision was enough for the state to offer $3 billion in incentives, plus hundreds of millions more in local investments, in exchange for a $10 billion investment and the creation of 13,000 jobs.
Those plans changed for a number of reasons. New York-based Corning Inc. wanted more incentives to build a needed glass factory, the development of other Gen 10.5 plants created a global oversupply and slowing growth and trade tensions added uncertainty to the picture.
Foxconn now says it is building a nearly 1-million-square-foot Gen 6 fab in Mount Pleasant because it provides additional flexibility.
“We’re going to create the Gen 6 fab to build as many things as we can,” Vincent said. “Large screens, if it makes sense. Small screens, if it makes sense. We have to have the ability to have options to grow that business.”
The Gen 6 fab is just one part of the new vision Foxconn is painting for the Wisconn Valley Science and Technology Park, the 3,000-acre swath of land in Mount Pleasant where the company has begun setting up operations.
Vincent equated the park to a mall, saying Foxconn would be providing the anchor tenants and working to bring in other businesses to fill out the rest of the space.
The Gen 6 fab is the responsibility of SIO International, one of the Foxconn companies included in the original contract with the state.
The other recently announced anchor development is a smart manufacturing center and high-performance computing data center to be operated by Foxconn Industrial Internet, Vincent’s company. Fii was not among the companies on Foxconn’s incentive contract, meaning the deal with the state would need to be amended for Fii’s jobs and investment to qualify for tax credits.
The manufacturing center will be a 260,000-square-foot factory that will primarily build components for server racks, although Vincent said the company will have plenty of options in how to use the facility.
“When we first had the concept for it, the idea was you could take a sheet of metal in one side and out the other side would pop a high performance computing server,” Vincent said. “Now that we’ve been on the ground for a little while we’ve recognized that that is one potential application, but we see lots of other applications.”
Vincent said Fii will need to run the factory differently from its facilities in China, which primarily produce a high volume of products with little variation.
“That model doesn’t work in North America in the electronics space. It just doesn’t make sense from a cost standpoint,” he said. “We’re designing the factory to be a high-mix, low- to medium-volume factory in North America but utilizing all of the high-performance, high-capacity capability we have in China.”
Fii has already announced plans to build self-serve coffee kiosks for Texas-based Briggo in its Wisconsin facilities and another deal to make smart home security products for San Jose-based Qolsys.
The first plans released for the high-performance computing data center called for a spherical structure that evoked images of Epcot or the Mitchell Park Domes in Milwaukee. Foxconn, however, asked Mount Pleasant to put those plans on hold while it considers design options.
Vincent said the data center is an important part of Fii’s strategy of offering technology solutions to customers and partners using AI.
“In order to do a lot of AI you need very fast computers,” he said. “To enable both what we’re trying to do and what we’re trying to do for the community, we need the ability to have high-performance computing.”
Beyond offering AI capabilities that help manufacturers improve their operations, Vincent said there are plenty of other potential applications. He pointed to health care in which AI can dramatically improve the accuracy of mammograms by focusing the attention of radiologists in potential problem areas.
Foxconn’s Wisconsin project is now under new leadership after Louis Woo, a special assistant to Foxconn founder Terry Gou, stepped down earlier this year. Jay Lee, a board member and vice chairman for Foxconn parent company Hon Hai Precision, has been designated to oversee the project. Lee is a University of Cincinnati professor and was previously vice chairman of Fii before stepping down for a similar role at Hon Hai.
Vincent said Lee’s presence has helped unify the various Foxconn businesses operating in Wisconsin behind a single mission.
“Now what we have is a more cohesive group of people where everyone has said, ‘Yes, you are the guy,’” Vincent said, noting it was difficult for Woo to influence the project in some cases because he was not actually part of any of the Foxconn companies operating in Wisconsin.
“We’re standing up a lot of things all at once in a very fierce environment and at a very fast pace,” Vincent said. “Hopefully in the next three or four months you’ll start to see some things happening that will show, OK, they’re not just talking.”