By Jennifer Smith
Consumer-goods giant Unilever PLC is building virtual versions of its factories, using data streaming from sensor-equipped machines to create digital models that can track physical conditions and enable testing of operational changes.
The "digital twin" strategy uses machine learning and artificial intelligence to analyze torrents of information from connected devices, and is aimed at making production more efficient and flexible.
"We've got it in [plants that make] mayonnaise, soap, shampoos and conditioners, laundry detergents," said Dave Penrith, Unilever's chief engineer.
The approach is part of the growing use of the enormous streams of data that flow from come from the Internet of Things, devices embedded in objects including factory equipment that send out information on how the equipment is operating. Such technology is gaining traction in industrial operations from high-tech and pharmaceutical manufacturing to oil fields and refineries as companies look to improve operations by using tools such as predictive maintenance to get to machine parts before they wear out.
Unilever is working with Microsoft Corp. to create virtual versions of dozens of its roughly 300 global plants over the next year or so. The technology lets the Anglo-Dutch company make real-time changes to optimize output, use materials more precisely and help limit waste from product that doesn't meet quality standards.
The devices send real-time information on temperature, motor speed and other production variables into the cloud. Algorithms take in the data and use advanced analytics to map out the best operational conditions. Workers on site track product quality with handheld devices, modeling solutions to problems and sharing data with colleagues in other locations.
Unilever and Microsoft set up a pilot last year at a facility in Valinhos, Brazil, that makes products including Dove soap and ice cream. It took three or four weeks to create a full digital twin, which the company used to set parameters for standards such as the temperature at which soap is pushed out before being cut into bars.
"If the temperature is too high you can use the machine to cool the soap, " Mr. Penrith said.
The project has saved Unilever about $2.8 million at that site, the company said, by cutting down on energy use and driving a 1% to 3% increase in productivity.
Unilever now has eight such digital twins of plants in North America, South America, Europe and Asia. There have been some challenges along the way, including developing ways to stream data from older machines that aren't equipped with the latest technology.
The strategy has helped increase production yields for products such as shampoo and conditioner. Instead of stopping production to check quality, operators track whether the process is hitting the right parameters as set out by the algorithm and test the quality offline, only interceding if production is getting off track.
"We don't remove the quality check but it allows us to move into the next phase of the product with a much higher range of confidence," Mr. Penrith said. "That's freed up capacity in our factory without the need to install extra equipment."
The company is streaming data from 15 factories, with plans to connect 70 sites by the end of the year and another 100 or so in 2020.
Unilever wants "to have a very clear digital representation of their supply chain" by bringing in as much data as possible and using advanced analytics to inform design and management decisions, said Gartner Inc. vice president and analyst Noha Tohamy. Rivals Procter & Gamble Co. and Colgate-Palmolive Co. are also leveraging artificial intelligence and smart factory technologies in their supply chains, she said.
Many companies are still trying to figure out how to harness the huge volumes of that stream from connected machines and factories without "drowning in data," Ms. Tohamy said. They also need to get employees on board with the technology. "There can be some skepticism on the part of the front-line users on how they're going to use all this," she said.
Write to Jennifer Smith at email@example.com