Innovative use of "golden batch" optimization in discrete manufacturing at Weber Metals

Guests: David Mitchell, Automation Engineer, Weber Metals. 
Sean Upson, Systems Engineer, AVEVA. 


David Mitchell is an automation engineer at Weber Metals, a major supplier of aluminum and titanium forgings to the aerospace industry. Among the thousands of things Weber manages with sensor-based data is a 60,000 ton forging press that is the world’s largest private investment in aerospace metal forging. When Weber installed the PI System it was to consolidate various silos of sensor-based data into a single source to improve situational awareness. Over time Mitchell found more valuable uses for the data. For example by comparing each forging to other forgings in a production campaign operators now have a "golden batch" profile that helps them improve the quality of each forging. Integration with their MES (manufacturing execution system) further improves their ability to meet production targets by having all pertinent information in a common data infrastructure. Improved visualization tools allow Weber to identify more easily anomalies and thus detect and troubleshoot more effectively quality, compliance, and maintenance issues.

Subscribe to our biweekly newsletter. Visit our website. Follow us on LinkedIn.

Transcript

0:09

Well hello everybody welcome to Radio PI. I'm Nick D'Orazio from OSIsoft, which is now part of AVEVA. Today we're going to be talking about the use of sensor based data in discrete manufacturing. Weber Metals is a major supplier of aluminum, and titanium forgings to the aerospace industry. They're located in Southern California and among the 1000s of things that they need to manage with sensor base data is a 60,000 ton forging press that is the world's largest private investment in aerospace metal forging. Joining us from Weber is the man who's responsible for creating a centralized data infrastructure to optimize production. That's automation engineer David Mitchell. Hello, David. Hello. And also joining us today is co host, Shawn Upson. He's the system engineer who consulted with David as he planned the project out. So he's going to be able to tell us about the project, and about the just how he worked with David to get answers about how to put together the design. Hello, Shawn. Hey, happy to be here. Happy to talk about PI with you. All right. David, can you go and start by just telling us a little bit about what what you do at Weber? Um, well, I do a number of things. Right now. I started out working in with the SAP and MES team. Because I, I'm familiar with databases and familiar with both traditional and SQL databases. And I've also developed business applications. So when I started out at Weber, I was on the more of the it data side. And now I'm in the maintenance equipment engineering side of things. So I'm really trying to be the bridge between, you know, the shop floor and all the data from the shop floor, and our centralized databases and centralized enterprise networks and such to bring all the data on the plant floor. Up to the people. Okay, yeah, yeah, that we're running show. So one of the things that I'm fascinated about this topic is, you know, you tell them the standard kind of getting rid of silos of information, stories that we hear from so many people, can you describe, to me some of the issues that you had before you started putting this project together? Um, well being from the, the, the IT side, and previous to really do in the more of the IT business application stuff, I was controls engineer, she come from electrical engineering background, I got into controls, and I got into software. So I had seen a lot of plant floors, and seeing what good was and seeing what was not so good as far as sensor based data. And so I come to the company knowing about the challenges of sensor based data. And when I got there, I looked in and saw there's at least eight silos of data, you know, and I'm like, hey, and then they're trying to add another one and use abs. But I was like, No, this is not the right way. Because, you know, column data for for time series data, just really, you're fitting, you're trying to fit one thing into another thing. They don't really doesn't really work that well. I know, we don't all the time. But it's, it's great for transactional processing, because you know, yeah, credit card, you want to back out, it's great for that. But time series, yeah. So watch, optimize. Now, it's not so good. It's not optimized, it's clunky. With the time series stuff, you have the swinging door algorithm, you can save much space on the on the hard drive, which people might not care so much about, but what they do care about. It gets up to your to you, you know, so, okay, so it sounds like it sounds like the story of situational awareness. Yeah, stuffing in in a variety of places. Why was it difficult to pull all that stuff together? Maybe if I correctly, you had a real grab bag? Yeah, well, most people don't understand this stuff. Most people don't get the difference between a relational database and a time series database. I took a number of talks with people to, you know, to get them to understand the benefits and how we can do things better than than we're doing and how these data silos are limiting us and well, let's, let's get everything in one one spot would be way better. Okay. So now with that the situation and you're looking for situational awareness in a single in a single structure. Where did you start? What's what how did you get started and was Shawn involved at that point? Well, we started with just I started explaining people what Hey, what would Doing we could do better. And then really what happened is we come up with a new need, hey, we need to record data on this, we need data on this. And proposals were put up for some more data silos and I was able to say no, we got there's a better way, and was able to get it put in the other project is 60k. area that is a new stuff. That had been we have been doing some data recording with the SCADA system, the Siemens SCADA system, the historian and that was still row column data. What was hard was for us to get that data out in the format that that was required to get, we're working on getting a power rebate. So the power, we need a certain format, and we couldn't get to be in that format. With PI. It wasn't that hard. Is that up to get in? to pretty much any format? I don't know format that we need to get that we can't get it in as far as you slice it, however you want it. Shawn, can you? Can you describe what's it? Like? What's the process you go through when you're working with a customer like, like, David, and trying to define what their system should look like? Yeah, sure. I mean, the first thing really, is to understand what David's going through, you know, I want to listen to him, I want to hear everything that he just said, I want to hear what all of his problems are, I want to hear where he's at right now. And I really want to understand where he wants to go with it. You know, so when I was, you know, initially talking to and listening to David and getting really excited about, you know, just as goals of breaking down all these different silos, and we, once we hear all those things, then I can start asking some questions to the details, what are the different systems, what protocols? Do they get those systems speak? You know, where are they in the network, and we can really start to get in there and understand more of what's going on and, and what is going to be required for the system at the end of the day to accomplish that. But the other thing I was also thinking of when we were initially talking about this, and hearing David's goals, I was very, very excited about that. He knew that we could do that. But I also felt like there's a bit more that we could pull out of it within that scope. At the same time. And I think, David, as we started to work, your perception of what PI was changed from the beginning towards the end, could you could you describe that process? And how that that work? Yeah, I have, because so I had, I had previously worked with PI, because we needed it in the facility that I was in, and it's very successful, then I got doings other stuff, I kind of got this story and world for a bit more into business application stuff. And when I came back, I was like, well, PI is what we need here, guys. And so when I started, I already I knew his story, and I knew what I could do as far as that. And so there was either I didn't know what the older system could do, or we there were many new improvements in the last 10 or 15 years since I've been introduced to it. And yeah, there's a lot of things we can you can do. I've noticed with the Asset Framework, with connections to, to relational databases, and mixing both time series data and relational data having Event Frames and notifications. So you can you know, it's not just a historian, all the best, the main thing, that big thing that we need, we must do historian stuff, but it can do a lot more, it's a lot more flexible, a lot more powerful than the majority effect any of the other systems that we have on the shop floor for this type of stuff. Okay, and we'll get we'll get into some of those other things. So David, which specific silos so far have you been able to combine within that a 60k Press area? Well, the first one I've duplicated everything that's on the Siemens SCADA device baghouses as well. Yeah, so the baghouses were a new system. So that was t 04. data recording wasn't really replacing something. But what we did do was prevent another silo. Right now I'm working on duplicating everything that is on the we have an ignition system. The thing with that is it it works. It's in row column data, so it's not really efficient. And while it meets requirements for many of the things that we do, it's no nobody knows out it is currently at the company how to program it. So we'd like to get rid of that and hopefully backfill, hopefully someday not too far in the future, we can backfill all the data from that database into the PI System is what I'd like to do to fully decommission some stuff and still retain the historical data. And I think that's possible, too. Okay, great. Hey, so David, can you describe how having better visibility into, you know, just being able to trend and see things in more detail has been able to improve things? Yeah, well, so part of the one of the things that we did was during commissioning of the baghouses, I was able to see how it was running, and see how the pod systems were working. So everything was working and notice a few things that weren't working, right. It helped the commissioning of stuff, just to start out. Okay. So anything with troubleshooting or maintenance? Oh, yeah. So I use it for for maintenance all the time. mean, you can put whatever in there that you're trying to do. If you're trying to tune a pod, you can put you know, you're not only the result, you can actually record maybe the integral wind up inside the pod that could potentially help you, you know, tune up ID. So, we can, you can do a lot of these things for a startup with improving systems, maintaining systems. And then, of course, there's a quality aspect, which I typically serve up data to others, so that they can see that. You know, I don't I help other I help the quality of people by giving them the data. You know, when I help myself with the maintenance stuff, you're the the human server. Yeah, yeah. Okay. So now you're in discrete manufacturing. And, yeah, but you've actually taken a page out of the folks who are doing batches by by dividing your discrete manufacturing into events, comparing those events. Can you tell me what's what's going on with that? Yeah. So I create what well, Shawn actually showed me a part of what when, when we were building this thing, he says, you know, what, why don't I just show you like, kind of the art of the possible and there is a system that he had access to, you know, the corporate system is for testing and display, and showing stuff and training, I believe. So there's all sorts of different stuff in that he was showing me a lot of it just was possible. I had some ideas, and I made a system that can use Event Frames to know when the press has gone into auto mode. And I'm also recording the part number, and shop order and other information. So we've integrated that a bit with the system so that when somebody at the the press says, Hey, I'm gonna clock this job in that job, the data from the job will go to the PLC, and I will read it with PI System. So now I not only have the process data, I have the data about the process data, like part number, shop, order, alloy, some other things like that. So what you can do now is say, show me all of the automatic modes, and I programmed a bit to go through during the time when automotive stop started and stopped. I programmed an event to go true or false. And I was able to use that to compare runs for for different type of products and line everything up with each other so that you can see the press curves overlaid on each other for different product lines. And you can see the variation that does or doesn't happen a product line to to try to discern. You know, better. Golden batch. Yeah, I think the golden Berta exactly what our batch customers are doing. So what's what's it like? I mean, with that golden batch, does that help you? better quality better, better end product? Yes. So some things are really take need to be really precise that we do. Like when we're doing titanium engine parts. There are a lot of parameters that have to be just right. Or you're not going to make that part, right. We're just gonna throw it away. We just like go back to you know, all that whatever you did didn't work. Hey, keep on doing it. You get the right part. So, okay, and use this lengthy processes, right? This is like five or six seconds or 10 seconds or something like that. Yeah. Well, that the So, yes and no. So Yes, the very critical thing happens in five or six seconds, but you have a very hot part coming in, that has to come in from an oven and there's a manipulator that grabs apart and puts it in the oven. So there's minutes in between this five seconds event, right? You have to get an old part out, perhaps it dies, but a new part in that set a certain temperature, get it seated. Right. And then you press so it's a big, there's a big going on to do this stuff. So look at the whole data. Yeah, yeah, yeah, if you look at the whole data, it's, you know, you'll, you'll look at an hour's worth of data. But really, what you really care about is mostly happening. There's about a minute of that. That's the goal, right of the hour day that because, yeah, the pressing event for these things are usually pretty quick. You know, you have a machine that's jogging down to get close to it, but now is really what you care about is when the die hits the metal until it gets to the end. So that's, that's the curve that you want of the hole, maybe five minutes to get one part is five seconds. Of that you're really looking at, David, I remember the first time when we brought that screen up that had all the different events of you know, 60,000 ton press, you know, pressing down on your your parts, you know, just the 10 seconds and you know, some values going over it. And yeah, when you're looking at that your eyes were just lighting up a bit. Yeah. Yeah. You don't know this is great. Like, are this cool. Or metallurgist need to see this right now? Yeah. Yeah. One small victory. Oh, that's Yeah. That you're like, Can we get Can I get this data more often? Can I make this better? Yeah, yeah. So I'm still looking at, you know, what? What's the next thing? You know? So, we hear a lot of customers talking about energy management. If you're doing anything with energy management now, yes. What kind of the project I'm actually was working on today, I had a small victory. We're working on getting a rebate from Southern California Edison. With the new press, we put a good amount of money in it to make it extra energy efficient with very modern vfds. And algorithms that spin motors down and try to reduce the power consumption of giant power hungry press. So this, the modern technology should be much more energy efficient than the older stuff. So if we can prove that that we've done it, then we get, we get some rebates. It's always fascinating to talk to people who've had to do some of the some of the proselytizing or some of the popularizing of a system across an organization. And if I understand correctly, you I mean, you you had some really good milestones with that. What was it like working with this, showing it to different people getting gaining acceptance? What was that like for you? Yeah, well, it takes some work. Because you have to go around and talk to people, you're not just sitting at your desk, a lot of it was going around just showing people that say this, this is what we're doing. Now, this is the better way to do that stuff and stuff like that, or just explain to people most people don't even really want to hear it the difference between a time series database and relational databases. They just don't they don't care. What kind of response Did you get when you showed, you know, what the capabilities? Oh, yeah, everyone liked it. So everyone's liked it. Particularly the the people deal with the ovens and stuff. It's, it's much easier system to use than the other systems we have, as for an end user perspective, just for the the tools that that are available for reporting. Okay. One question we always ask is, has this help with any larger corporate goals? I mean, a lot of folks have big into sustainability or they're looking for, like I said, to move into condition based maintenance or something like that. How about Yeah, was this was this did this align well to anything that you guys had been planning? On? ending with? Yeah, and we basically the reasons we initially did it, it's working for all of those. And I would like to take out some of the these older systems and centralize all the data into one spot, instead of many different spots, be maintaining less stuff, have less licenses for different products, have people you know, don't need to know so many things, less things to maintain. Kind of where I'm where I want to head. Also the the ability to be able to relate data from one foot from one thing to another. If it's all in one spot, you can do it much easier. Okay? So, like a lot of engineers, do you have a piece of broken gear, or like a fried motherboard or something that you keep on your desk as a momento? Well, I've got an old XP laptop on my desk right now, it's not a memento. I want to get it off because I'm programming a 20 year old HDMI today. So it only runs on XP, and this machine's about to die. So that's what I have on my desk right now. Gracious, so you're working with an XP. So you can Yeah, I was forced to said, Well, speaking of programming, so what's the coolest either the coolest code you've ever written, or the coolest calculation you've ever done? Ah, that's thinking that that, you know, I'll tell you what it was in my previous job. I programmed the control system. To do a fast start for an extrusion process that saved us a lot of time. The other thing I did is during for a stretching process, I was able to program the machine to automatically stretch if you really have to get into physics and stuff, how this works, but I was able to do very fast data calculations and understand when different stresses met a certain point. And it wasn't just distance. So that really revolutionized a stretching area, which was very important for us. So that was, that was accomplished doing that I felt smart. I like feeling smart. I felt really smart. I no one else can figure this out. I'm so awesome. So I haven't seen an operating area yet that didn't have some weird animal that's always clump getting into into the works. Well, what's what's going to animals, torment your folks up. Well, we've got a bunch of cats. Yeah, there's a bunch of cats that Weber metals that we just, we've just let be there. Well, I really appreciate your time. Thank you so much for joining us. We've been talking to David Mitchell from Weber metals. Thanks again, David. Thank you. Thanks, Dave. We've also had as a co host, have shown ups in of a movie but thanks again, Sean. Yeah, absolutely. Okay. I thank you again, everybody. We'll see you in another two weeks and take care. Bye bye.