Jeff Winter has nearly 20 years of experience in the manufacturing industry, with a focus on automation, safety, controls, and OT and IT systems. Jeff became a thought leader in the fields of Industry 4.0 and digital transformation, and has actively participated in industry associations, academic groups, advisory boards, and industry research teams.
Jeff has teamed up with Scott Achelpohl, managing editor of Smart Industry, to create (R)Evolutionizing Manufacturing, a monthly series of chats about how industrials of all sizes and budgets can embrace technology. The two experts plan to cover a range of topics, including digital twins, predictive maintenance, cybersecurity, IT and OT convergence, automation, and much more. This episode takes a deep dive into data and its importance to manufacturing and digital transformation.
Below is an excerpt from the podcast:
SI: Jeff, our audience is, as you know, continually invited to give us questions to answer here, as we will do every month. They can be posted to our social channels or sent to [email protected]. We've had a few more for this month, so let's go about answering them. For that, I'm going to ask Jeff for a big assist with the answers. So, here's the first question from Stephanie who asks, “What are some unconventional data sources that manufacturers can tap into for better decision making?”
JW: That's tough, because what's unconventional to one might be normal to another. But let's start by breaking out the types of data, and there are a couple of different ways to think about this. I was actually part of an article that got released last year with MESA called Manufacturing Data Capture and Exchange, and in that, the group of us that helped write it, we came up with nine types of data that are typically found in manufacturing environments. You have your production data, you have your process data, you have your product data, your equipment data, your quality data, your financial data, your facility and environmental data, supplier data, and audit and compliance data. But we also talked about how there are three different main levels of structure and how we store the data that the manufacturing systems creator used. You have your structured data, your semi-structured data, and your unstructured data. And then on top of that, the data can be described by its frequency, whether it's discrete, continuous, or batch, or its relationship to time, whether it's transactional or real-time data.
So, unique data will be hard to answer in terms of finding that. The only study that I'm aware of is one that's done by LXT, but it was specifically focusing on AI. In it, LXT evaluates 13 different data types and said which was most used today and what will be used most in the future, at least for the purpose of AI, and they did across all industries. And since computer vision was the most deployed AI solution across all industries, it may not come as a surprise that images came back as the top data type used in manufacturing at 39%, with time series second at 37%, and then sensor data coming in at 36%. Now future manufacturers picked product or skew information first and time series second. The one I found most interesting in there, at least in terms of the data types, is because of other industries too, is they had handwriting, they had voice, they had gesture, and they even had user behavior data, all of which I would find pretty unique to manufacturing. So you're aware, user data was dead last in manufacturing, but actually first in healthcare, so I'm sure that there's something we can use from that industry.