The molten metal, now 3,000°F, is channeled from ladle to caster, cooling to form a burning orange slab. This work has been repeated billions of times over decades. For the last 70 years, it has happened in steel mills around the world; today, it happens in data factories.
Modern industry may look much the same from the outside, but a quiet revolution is underway: Manufacturers are going online. A steep drop in sensor costs over the past decade has allowed businesses to collect data at every stage of production. Fifteen billion machines are currently connected to the internet, and in 2020, the number will pass 50 billion3. By 2025, McKinsey forecasts that “smart factories” will generate as much as $3.7 trillion in value.4
These modern manufacturers are producing unfathomable volumes of data, and artificial intelligence is needed to make sense of it all.
“Modern machine learning can identify patterns from huge, messy data sets,” explains data scientist Alp Kucukelbir. “With human expertise alone, you can't really tease out the insights that you want. Machine learning allows us to unravel those patterns that would be difficult or impossible for people to identify.”
Seeing the opportunity, Kucukelbir cofounded Fero Labs, a company with a platform that pushes sensor data from the factory to the cloud, where it’s processed with machine learning algorithms. Their software offers insights for how to boost industrial output, prevent costly machine breakdowns, and reduce waste--all contributing to higher product quality and lower costs.
See how Fero Labs and Intel artificial intelligence are helping factories save money and simplify the data they're collecting.
“Machine learning allows us to unravel those patterns that would be difficult or impossible for people to identify.”
Putting sensors to work can save steel companies millions every year by reducing the use of ferroalloys, an expensive material, and preventing “mill scaling,” the unwanted oxidation of steel. Fero Labs is able to predict mill scaling with an accuracy of 78-100 percent, according to Kucukelbir, reducing it by 15 percent.5
Most manufacturers already have the sensors in place, but few get the most out of them. One Fero Labs client had 12,000 sensors installed in a steel mill, but only actively used five. When setting up their AI platform, Fero Labs increased sensor data usage 40 times over simply by feeding previously unused information to their algorithms—offering comprehensive insights into factory activity without installing any new equipment.6
But running these giant workloads in real time requires serious compute power. Fero Labs uses newest-generation Intel® Xeon® processors to accelerate the speed of their algorithms. This helps them refine machine learning models even before they go live with a customer.
Looking beyond steel, artificial intelligence will have even bigger payoffs in other fields.
“I believe automotive, aerospace, and oil and gas will be the main industries that can benefit immensely from AI technologies,” said Nandini Natarajan, senior research analyst for industrial automation at Frost & Sullivan, a consulting firm.
These manufacturers all have complex supply chains involving thousands of diverse components and specialty tools. Any delays, breakdowns, or mistakes can shut down a production “cell,” the individual assembly points in a lean-manufacturing system. Only AI can predict the complex interactions between each production unit, automating requests for parts, labor, tools, and repairs to maximize efficiency.
Stephen Ezell, an expert in global innovation policy at the Information Technology and Innovation Foundation, said that manufacturers who don’t adopt data-driven strategies will be left behind: “If you're hidebound, if you're stuck to the old way and don't have the capacity to digitalize manufacturing processes, your costs are probably going to rise, your products are going to be late to market, and your ability to provide distinctive value-add to customers will decline.”
“If you're hidebound, if you're stuck to the old way and don't have the capacity to digitalize manufacturing processes, your costs are probably going to rise, your products are going to be late to market, and your ability to provide distinctive value-add to customers will decline.”