Although artificial intelligence has transformed many business functions, applying it to manufacturing remains much more difficult because factory data is fundamentally different from the structured information used in sales, finance, and product design.
According to Jon Sobel, co-founder and CEO of Sight Machine, manufacturing data is often collected over decades from equipment provided by different vendors, leaving companies with inconsistent labels, disconnected systems, and little context about what individual data points actually represent.
“If you take a lot of data from a lot of sensors and throw it all into a data lake, no one knows what it means,” Sobel says. “The labels on the data, the way the data is named on the shop floor, the data is added every year by different OT integrators…no one even knows what the data is, they don’t care how it all fits together.”
In addition to inconsistent labeling, manufacturers also have to reconcile data collected at different speeds and times. Machine sensor data may be available in real time, but quality results, inspection data, or defect reports may take hours, days, or even months to arrive.
“Do you actually want to understand whether I’m making something good or something bad?” Sobel says. “But quality data isn’t available until maybe an hour, a day, or a week later. Integrating all of this over time is a very difficult problem.”
Sobel says that before AI can generate reliable insights, manufacturers must first organize, map, and understand their production data. Once these challenges are resolved, AI will be able to identify relationships that would otherwise remain hidden.
“We see data and AI as another next-generation tool that will be put into the hands of experts, allowing them to see deeper and farther than their senses can,” Sobel says.
Looking for quick answers on assembly and manufacturing topics?
Try Ask ASM, our new smart AI search tool.
Contact ASM
