Issue link: https://iconnect007.uberflip.com/i/1372612
MAY 2021 I PCB007 MAGAZINE 33 sured parameters, and you must establish your quality factor at the end of this thing to root out the manual judgment of your quality guy. In this case, let's say you have some growing metric measurements of your copper etch rate or something like that, then you compare it against your set points and acceptable thresh- olds. en your network can get data that show whether the result is valid. You already have it digitized and at the same time you are stacking this information with a set of parame- ters that the product was running with. By vir- tue of having those two in a big enough popu- lation, you can provide the teaching sequence for your neural network. It's pretty much the same, but we tried to make it a little bit more automated. Matties: Does a company have to go all in, or can they piecemeal with your system? Zajac: Partial involvement is possible, but aer the initial install there's going to be an appetite for more. Matties: But you have to give them the first bite to build that appetite, and they have to be able to see that it's right for them. Johnson: Robert, I'm curious what your rec- ommendation is to a fabricator who is looking to take steps in this direction. Is there a stan- dard place in their manufacturing line where you would suggest they start? Zajac: I would start with the part of their pro- duction line that is of the biggest interest to them and has the biggest throughput. I think this is the best course of action. It's going to bring the biggest benefit by virtue of having the highest volume we'll put through it. Johnson: So, you're suggesting working on the part of their process where they have the most throughput, rather than their biggest bottle- neck? Zajac: at's correct. Matties: Is this soware ready for deployment in factories for customers? Stepinski: e soware is being developed for our factory clients and they are receiving it. It is in the scope of these projects, the timeline— and of course there will be upgrades later—but the timescale of these factory projects really facilitates great strategic improvements in in- dustrial capability. If you are a regular equip- ment manufacturer building one-off machines, you don't have the runway to go ahead and do what Robert is doing right now. You can't just sit back and say, "I have two years to contem- plate, put a staff in place, figure out exactly what I want to do, then test, qualify, and im- plement it." at's not a luxury anybody has. Johnson: Alex, you were just touching on some- thing that is so interesting about this whole pro- cess. When we were talking with Schweitzer (SEL), we asked if they had a pool of talent in their area for PCB fabrication; they said there wasn't. en, they immediately referred to your training program in which you have some Schweitzer people that you are training. You have them over at a facility where you're help- ing them gain that knowledge to run their fac- tory. A key part of your whole transition plan is to not only put the facility in place but get that human knowledge up and running quickly. You were just saying that when you're talking to OEMs, they don't have any history in PCB fab—even the OEMs who were doing fab in the U.S. before. Aer 20 years, you've lost that tal- ent in your company, nobody remembers, and everybody is new to this again. What's your vi- sion on getting that on-the-floor expertise up and running quickly for a client? Stepinski: is year, we are really focusing on developing a training program with courses, education tests, and a whole curriculum on getting trained in a PCB shop. SEL is actually