Issue link: https://iconnect007.uberflip.com/i/1447212
28 SMT007 MAGAZINE I FEBRUARY 2022 Stepinski: Yes. Most PCB shops have an ERP system, but you need to go a step further. You need to get an MES system. e cost is typically $500–$2000 per user for an annual license. It will do all your correlations for you. You don't need to do too much, just add sensors and put in measurement plans. ere's not much to it. But how many PCB shops in America have MES systems? It's not something they do. e more advanced facilities overseas already have this. You take it a step further and you get a product lifecycle management (PLM) system to manage your NPI process for you, so you can group products into families and have a predetermined engineering process for how you do things, so you're not just doing trial and error all the time on new products. You can have more predictive results that you'd be able to extrapolate, "I'm taking this order. It's got these conditions." You have an MES system, a PLM system, it automatically tells you it's not going to make it by the date you committed. It's going to be four days late, for example, or it's going to be four days early. But you must use a system like this to know this. ere are a lot of complexities and no one person knows all the constraints in the factory at any one time, unless it truly is a mom-and- pop shop. at's the approach, but we are really late to the game. ese tools have been on the market for years now. Other very complex industries have adopted it, but PCB is just behind in this area for whatever reason. Johnson: How do you make that leap from cap- turing the data with the sensors, looking at your error situation, and then moving into a more predictive system? You started making the point that by using your systems, over time you could start to predict things like the expected weight of a new part number as it goes through the process. How do you get from data collec- tion to using that to predict and then check against those predictions? Stepinski: e CAM soware systems on the market have the tools built into them to tease out the key input variables that you need to correlate to the data. You take this information and build your correlations, your regressions— whether they're linear or nonlinear, it's imma- terial—and then you have equations based on simple correlations that you can use to predict your recipes. It's very straightforward. Johnson: Alex, phase one is setting up sensors and collecting the data. And if phase two, for example, is starting to be able to do predictive work using that data, is there a phase three? Stepinski: Phase three is now you've made your efficiencies such that you can do more. Now, what does "more" mean? Once you learn how to do your own process engineering instead of outsourcing it to the suppliers, which is what I would say is being done right now and it's encouraged even by ISO to do things like that, use your suppliers for everything. I think you need to have native capabili- ties to do this kind of work, and then you can take on bigger Industry 4.0 projects. Do that by partnering with suppliers now that you're on a stronger footing with them so you're not doing zero-sum discussion so there's value cre- ation happening. You can also just take on big- ger R&D projects, add a couple people to your team, and do bigger things yourself. There are a lot of complexities and no one person knows all the constraints in the factory at any one time, unless it truly is a mom-and-pop shop.