Issue link: https://iconnect007.uberflip.com/i/1189040
82 SMT007 MAGAZINE I DECEMBER 2019 Suppose you're a contract manufacturer, and you're getting a lot of pressure from your cus- tomers to share more data from the lines. So far, that only happened when it was custom-devel- oped, which takes a lot of effort and energy. They're only going to do it for some customers and special occasions, but now that we have this infrastructure in place, they can more eas- ily share this data securely to their customers. They also want that data for themselves, their operations, and some of the machine vendors might want certain data to do predictive main- tenance on their machines. Johnson: And there are plenty of original OEMs who are looking to get actionable data from the manufacturers making the products so that they can monitor what's going on. Kantelal: And the truth is, as a manufacturer, you might not want to work with every single vendor and every single method out there to be able to send that data to them. Johnson: Good point. Kantelal: You want something that is like an infrastructure layer that allows all these things to happen, and the broker provides that. Johnson: Going back to my hypothetical facil- ity manager, the tendency for someone in that role is to get bogged down in the hardware, sensors, PLCs, retrofitting, cables, and connec- tors, but that's not where the solution is. Kantelal: That's just a means to an end. Johnson: Let's talk about the software part. Kantelal: I was just on a panel earlier today, and we talked about the skill gaps in manufac- turing. There are a lot of conversations about data scientists and software engineers, but we see one of the biggest key things that can enable the future of Industry 4.0 is data engi- neering and understanding how data works. You have all these silos, so how do you com- bine all this data to then be able to do what you want to do with it? That's what we do— the engineering. We're combining all this data from all these different sources and allowing there to be what traditionally has been called a data lake. Now, once you have a data lake with all this clean, structured data—"clean" is a keyword here because there have always been data lakes—then you can analyze how you want. It just opens up all these opportunities that you probably wouldn't have thought of before, or you thought of it before but didn't know how. The cost and the time requirements were too high before. Now, you can start some queries, put in whatever visualizations you want, and start trying to understand your problems in a different way, combining it in different ways, and analyzing in a different way and come to a solution faster. Johnson: I'm getting the sense that what my facility manager needs to know is that the atten- tion needs to be on the software. The tempta- tion will be to concentrate on all the hardware and hookups, but the solution is on what kind of information you're going to get out of your system at the end of the day. Kantelal: I wouldn't say just software; it's about machine data. How you get the machine data is less important as long as the quality is not being affected. We have had cases where the machine vendor had their own way of getting data out, and then we get the data out and compare it. Our data has fewer gaps because we're going straight; we don't have all these different aspects, which is where the data engineering comes into play. It's difficult for Figure 2: Sensor pods.