• More News

  • Tim Burke, CTO, Arch Systems

    tim-burke

    Arch Systems was launched in 2015 by two Stanford graduates: Andrew Scheuermann (CEO) and Tim Burke (CTO). The company has had a global impact on Flex’s systems and their success has helped attract additional tier-one customers such as Jabil, Plexus, and Celestica.

    The React tool has been added to monitor and guide actions on the factory floor, contributing to improvements in attrition, quality, and performance. The partnership with Aegis and Factory Logix integrates data insights and actions back into the MES for operators’ benefit.

    Trevor Galbraith, Editor and Publisher of Global SMT & Packaging Magazine, spoke with Tim Burke at the recent IPC APEX EXPO.

    Arch Systems has been quite revolutionary in this industry over the last five years. You gained much respect by assuming control of the data for Flex, essentially worldwide. And with that, you managed all their systems globally and introduced significant optimization and savings. You describe it as top-floor and line-level changes. Talk us through what you did and how you’ve improved their capital equipment management?

    When you think about what we do with our customers, we typically work with large enterprises. So Flex, as you mentioned, was one of our most important customers, our starting customer, actually. We take customers from data to insights to actions to improvements. And when we think about those improvements, we find two categories. We call them top floor and shop floor.

    Top floor is if you are an executive in a large EMS company and you have trusted global visibility in what is actually happening on your shop floors, as well as your real OEE, utilization, quality, and First Pass Yield (FPY), but not after it has passed through many manual processes, but rather directly from the machines, trusted and synthesized, allowing you to make better decisions faster, to respond to market needs. For example, rather than buying 20 more lines, maybe you move lines around because you know you have excess capacity.

    So when you get a new customer, now you’ve saved a lot of CAPEX, or you understand that you actually need to purchase many more of this type of equipment because you can solve a process somewhere else. So these top-floor actions are what we focused on for the first couple of years when we were with Flex. Then, as we did more of that, we realized that there are many other problems that need to be solved moment by moment on the shop floor, and they result in top-line improvements like OEE and quality. But there is no one thing; instead, it’s like “death by a thousand cuts.” You need to have a system that is constantly monitoring the factory, guiding action to process engineers to quality engineers and test engineers, prompting operators. And so Arch has been adding more of those shop-floor features in a tool called React, which is about reacting to data. This includes monitoring your factory, knowing exactly how you’re doing, if you’re on schedule, how your job is going, how the work order is going, and what your top problem is. Then Action Manager, which is about monitoring all of the information coming out of your equipment, synthesizes it through Process Insights with these expert modules that focus on attrition, quality, test, repeated failures, pin failures on ICT, etc. Then, it prompts and guides actions to your process engineers and technicians. It doesn’t just tell you that you have a quality problem, but that you have, for example, a problem on this particular head on this nozzle and this feeder, and it’s causing you 5% quality loss at AOI.

    So the top-floor CAPEX has obviously resulted in massive savings for Flex, which has reflected in Arch picking up other accounts like Jabil and Plexus. What about on the shop-floor level? Have you contributed to the FPY and quality improvements on the actual processes themselves?

    Yes, it happens daily. We have some customers that are using our tools in all their factories worldwide, every day on the shop floor. For example, one factory has 1000 actions they’ve tracked through our tool in the last month, and every single one closed, resulting in shop-floor improvements that contributed to decreased attrition, increased quality, and increased performance. It’s about a way of working using data to then ensure that every shift goes a little bit better than the one before. The line leaders are using data to have meetings, the operators are using data to understand where to focus, and the process engineers are using data to know when/where they’re needed. All of those little actions ultimately result in increased OEE, yield, and throughput from the same lines.

    You also recently partnered with FactoryLogix with Aegis. How does that integrate into the work that you’re doing?
    Yes, we’re proud of this partnership. We’re excited to work with Aegis and their FactoryLogix MES tool, and we think about what we’re doing: providing data to insights, to actions, to improvements. If you have an MES like FactoryLogix, you already have a tool for your operators. You have a screen that’s in front of what they’re using every day, and there’s a lot of data in your system that is captured expertly from the equipment. We’re taking data out of FactoryLogix and putting it through our Insights engine, which we connect directly to equipment, and then turn it into insights and actions, and put it back into the MES. The information is shown directly to the operators in the screens they’re already familiar with. It’s not about adding a new screen; it’s about using these expert data modules we have in our system on top of data already being collected in an MES like FactoryLogix to have this extra supercharging add-on that then provides guidance and actions back into the MES, making it more effective.

    So it interprets that data and brings it back into actionable insights?

    Yes. We call it intelligent actions for manufacturing.

    That’s great. The other thing that you started a couple of years ago is the Insights modules for smaller companies that want to take part in this, correct? If they choose to, they can also share the data and get the benefit of that pool of data to improve their processes. Has this been successful?

    Yes, so much so that it’s now a core part of our product. I have about 40 of these, and they’re basically process-specific modules relevant from pick-and-place to AOI, ICT, and functional test. Think of the level of expertise you have in an expert process engineer. That level of understanding baked into a model. What we realized over the last couple of years since we deployed these in the field and to our customers is that there was a need for one layer on top of it, which is knowing which one is relevant for a given problem. Nobody has time to look at all the dashboards and figure out what’s going on. So we added the Action Manager layer, which automatically looks at the dashboards and prompts the user about potential issues, down to showing that there is a problem on a specific line of a specific machine. This is one area in which we’re actually seeing true AI come into play. These large-language models can synthesize information for existing dashboards and tools and then present that there’s not just a problem, but what it is and what might be done about it. While that entire synthesized bit of information is presented directly to a process engineer or technician, it also is translated into the language of the local operator vs. whatever language the machine is programmed in.

    Wow. So that’s one example of where you’re using AI. Are there other examples?

    We’ve actually found many interesting areas, including alerting or any type of monitoring that ensures there aren’t too many false positives. Everyone is busy; no one wants to be bothered if there isn’t an actual problem. We’ve found that large-language models can be quite effective at being a type of pre-screener/pre-filter. They analyze the data and information from the machines as well as the context, and reason. If it’s likely to be an issue, they summarize and present the findings to the user with more context. If it’s not, they just close it as a false call. This is something that you’re seeing with inspection machines, but it’s applied more broadly to the entire factory in our tool.

    This industry is changing rapidly, quite frankly. Technically it’s growing by application, which is bringing in all sorts of new processes, and there also seems to be a merge between SMT and back-end assembly packaging. What type of challenges does that bring you?

    This definitely is challenging for our customers. It used to be that there were electronics and semiconductor fabs, and both were very different. But now the lines are blurred. You have advanced packaging, you might have a die attach line next to an SMT machine. The level of data and analytics needed to run a semiconductor fab now is becoming important to similar levels of data for electronics manufacturers. There’s a need for much higher levels of connectivity, expertise, and insights. For Arch, this is an opportunity to not only help customers transition but to also navigate this transition. Companies need a partner and a tool that simplifies and synthesizes this additional data into actions because it’s becoming more complicated to run a good factory these days.

    Right. I know that some of your Flex factories have clean rooms in them, and they run alongside the SMT lines. Do you see that as a growing trend or is it just something that tier ones are doing?

    No, I think it is a growing trend. You see the Chips Act and all of the reshoring and repositioning of global manufacturing as well as the co-location of advanced packaging next to the SMT lines. I see this as an evolution of these processes. It didn’t need to be distinct before, but now there’s no reason to have a semiconductor fab here, an OSAT here, an SMT here. Let’s put them together when it makes sense. But this is leading to challenges that need more widely varying expertise in the same factory to run all of these complex processes.

    Absolutely. Well, it’s fascinating to catch up with you and to hear about all the different things that are happening both at Arch and the wider world. Congratulations on the tremendous success you’ve had recently and thank you for talking to us today.

    Thank you very much for talking to us as well. It’s always a pleasure.
    –TREVOR GALBRAITH

  • Subscribe today to the industry leading magazine and newsletter

    Keep up-to-date with information about worldwide trends straight to your inbox or through our printed magazine.