Jeffrey Nygaard, Seagate's executive vice president of operations and technology, pointed out that through smart manufacturing, Seagate saved over US$145 million in the past two fiscal years through its smart manufacturing initiative. For the next 3-5 years, it will focus on integrating AI and machine learning into more processes.
Seagate received the 2022 Manufacturing Leadership Award from the Manufacturing Leadership Council (MLC) of the National Association of Manufacturers (NAM), as a result of the mass introduction of AI analysis into its factories. The winning project OPICA, Optical Inspection with Centralized Analysis, supports factories with edge analysis to identify defects during the production process, further preventing them from reaching downstream. This project has saved 95% of resources and improved accuracy by 20% and the overall return on investment (ROI) was about 300%.
Nygaard pointed out that manufacturing is really about three things: delivery, quality, and cost. These three are also the focuses of smart manufacturing optimization. The first is delivery, ensuring that factories are shipping the correct number of parts on time. Then, the products' quality needs to be good. Last but not least, the cost needs to be more competitive.
He stated that there are two central elements to smart manufacturing, namely data and security. On the other hand, automation is the backbone of smart manufacturing. Its infrastructure includes platforms, sensors, and a communication system to deliver data. It not only connects factory to factory but also connects factory to clients. In between, big data analytics is used to perform machine learning to constantly optimize the process.
Currently, Seagate has seven factories around the world, including wafer fabs in the US and Europe. The entire production process for a hard drive takes roughly a year. It goes through thousands of steps to achieve that end product. Right now, about 96 different categories of AI solutions have been introduced into smart manufacturing. Each category has several models, meaning that there are hundreds of AI models in operation.
The first step to producing hard drives is wafer manufacturing. The data produced by Seagate's wafer fabs, including images and numerical data, will be captured, analyzed, and forwarded to the hard drive factory in Thailand for machine learning. Nygaard pointed out that this data allowed them to individually optimize each part, rather than a generalized process. This model can reduce variables and increase the yield rate.
20 years ago, Seagate was a very different manufacturing company, focusing more on manual manufacturing. Now, it's the machines that are doing the learning. By making decisions and automating actions with these inference engines, the company can better utilize its resources, including people, equipment, and space. For example, through OPICA, 7 automated machines could do the work of 150 operators using manual microscopes to inspect millions of parts daily. That means those operators could be upskilled and redeployed to other areas. In total, OPICA allowed 95% of resources to be moved elsewhere, with a 20% increase in identification accuracy.
