IPC Course Overview: AI Applications of Machine Data in the EMS Industry

Estimated reading time: 7 minutes

Tim Burke, CTO and co-founder of Arch Systems, is presenting a new course through IPC, AI Applications of Machine Data in the EMS Industry. These topics—AI, machine learning, and Connected Factory Exchange (CFX)—are top of mind in the industry, and in this interview, Tim shares insights into how his company helps customers leverage machine data analytics to enhance manufacturing efficiencies and the tremendously important role of AI, which is making more meaningful data aggregation and its usage significantly easier for everyone.

He provides a sneak peek into the curriculum, emphasizing the importance of CFX and its pivotal role in streamlining factory operations through standardization and intelligent data utilization.

Marcy LaRont: Tim, there are no more exciting topics right now than AI, machine learning, and CFX. What is your connection with these topics?

Tim Burke: We are a company that provides machine data analytics to manufacturing companies broadly, but also specifically in electronics. We provide an end-to-end product that goes from data collection—using CFX when it's available—through AI-powered shop floor tools to help factories actually use their data to make guided, intelligent actions that improve the operational efficiency of the shop floor.

I've been involved with IPC for a while. I am a co-chair of the CFX committee, helping drive forward standardization around data collection to improve factory efficiency.

LaRont: Can you help us understand the concept of CFX in the simplest of terms? 

Burke: CFX is a way for pieces of equipment and machines in your factory to send the right information about what they're doing to one another so that other applications and people can consume it and make better decisions. CFX directs what data to send, when to send it, and how to send it in a way that's plug-and-play and interoperable. Unlike some other standards, it also identifies and encapsulates the most important information about a process or a piece of equipment, and it provides that guidance. So, it is more than just a shared language between machines. 

IPC’s CFX standard defines all these things and outlines how you should communicate, the protocols, and other related matters. The standard also includes a free software development kit that helps machine vendors and software providers implement CFX in their own applications.

LaRont: Tim, your course syllabus covers everything from collecting and analyzing machine data to explaining AI co-pilots, agents, and applications of AI technologies. What is most daunting or challenging for companies considering implementing this into their facilities?

Burke: Let me start with two challenges because they encapsulate why we structured the course the way we did. First, companies often struggle because they don't know how to start collecting data. Imagine walking into a factory with all these expensive machines. There's no cable you plug in. It's all software and understanding. How do I talk to each piece of equipment, and how do I get the data? That's where CFX helps. It provides that common language: Here's how you get started, the kind of data you can get, how you get it from the machine, and how it gets sent. Now, you can collect it. 

The second daunting challenge we address is how you use the data you’ve collected to make decisions or to automate something to make your factory better. When I encounter these two core challenges, I need a use case or an understanding of what I am specifically trying to improve or how the particular data you are collecting can help you to achieve a desired end in your factory. How do we get and use the data? It’s important to be able to answer these questions.

I’ve already taught this course, but this time will add the focus on AI because I wanted to give very practical examples of how AI helps that second problem of knowing what to do with data. Often there's a worry or a belief on the part manufacturers that they don't have enough data expertise. They know how to do manufacturing processes and factory operations, but they’re freaked out about knowing what to do with data. Luckily, AI agents and some of the newer technologies can guide you along the journey.

Previously, you needed a person with a lot of expertise to analyze data. Now, once you have compiled some data, you ask AI, “What's the root cause of this problem?” AI sifts through the data and provides a readable summary of what it finds. On the application side, a new level of guidance and support can help you get over that first step of how you can take this end-to-end and solve some problems in your factory. You can simply give these AI systems the data with the prompt: “Summarize it for me. How does this help me solve X problem?” and it will provide reasonable answers. 

LaRont: It's interesting that AI itself has been very daunting to people, and for something as big as CFX, AI will make it more accessible and attainable. 

Burke: Exactly, and we’ve seen it in our work. The mechanism was unclear for the practice of collecting data to help improve one’s factor. You might have needed to hire a data scientist or even a team of data scientists with different skills or expertise. Now, these AI agents can go all the way from analyzing data to Lean manufacturing or Six Sigma-style changes such as, “It’s time to change this feeder in your pick-and-place machine. This is your top cause of downtime.” Those are classical answers and solutions that factories understand deeply. It is a whole different situation vs. data scientists who say, “Here's a bunch of data in Excel and, with the right skills, you can take that next step and figure out what to do with it, but we can't help you get there.” AI fills that key missing gap. 

LaRont: That’s powerful, for sure. Do you have a business case study you could share with us?

Burke: I’ll share two examples. The first is just about data. If you have thousands of machines across many different factories and you know which machines are idle, then you know where to put the next customer order. One particular factory line had excess capacity so work could be moved around to where they had that extra capacity within their fleet of machines. It's a fairly simple data problem, but it provided a ton of value in terms of productivity and profitability.

We have another customer that saved $14 million by using their data and AI agent to ask, “How many new factory lines do I actually need? Where can I reuse or reoptimize?” It provided them with data and precision to plan better for the next year. Those are potentially big capital savings in not over-purchasing, for instance. 

The shop floor is always a bit of a fire drill. Everyone wants to get things done, meet production orders, and improve quality. Any downtime means you're losing money. One role of the floor supervisor is to record the information about why there’s downtime. Every week, they look at that list for commonalities and ask, “What can I do better next time, or how can I continuously improve?” We found that AI can do that categorization process for you so you can focus on solving the problem and get back up and running. Then, you ask AI, “What was my last cause of downtime? What was the top cause of downtime last week?” It might respond, “You lost six hours due to this cause,” and you can ask for reasons. 

Another feature is voice capture in 100 languages so you can just say what you saw on the shop floor or even just take a picture. You don’t need to wait for expert analysis, or typing a long essay at your computer. It takes just 10 to 15 seconds, and everyone can do it—shift by shift, day by day. At the end of the week, you ask AI, “What was the top cause of my downtime?” and you find out that “this X setting is consistently set wrong across these 15 different assemblies, and that resulted in three hours of downtime.” This is a critical insight, and not only do you get the information, but AI helps you synthesize it into a picture. Previously, it was very challenging and time-consuming to capture that sort of tacit knowledge.

LaRont: Your ability to run lean is accelerated. Your case examples saw significant benefits straight to the bottom line. Is this particular course for anyone in the electronics industry? 

Burke: Yes, at the end of the day, the applicability of AI is the same. We see it broadly in all discrete manufacturing: PCB, assembly, and even machine shops, and those kinds of activities. What's different is the exact data. The nature of the process is slightly different. Therefore, the questions you ask might be slightly different, but the tools and the benefits are the same. The techniques are the same. Actually, IPC-CFX can be used broadly in all the different scenarios. Even though it started around SMT, it has extended broadly to talking about discrete manufacturing. Anything where you're doing cycles and building a product is a good application for CFX.

LaRont: What do you hope your participants come away with after attending your class? 

Burke: I hope they come away with an appreciation that AI doesn't have to be scary, and that it can add real value in the factory. Just three or four years ago, when we talked about AI, we used the term “machine learning,” and we said it was something in the domain of experts, that we must hire a machine learning person. I would give that person a bunch of data, toss it over the fence to them, and hopefully they came back with a machine learning model. I had no idea how that machine learning model worked; I just trusted that it would tell me what to do. It’s different now. It's more about talking to a “person” that’s always available, has information at its fingertips, and acts as an assistant that wants to help me.

It's a lower barrier to entry, the tools are much easier to work with, and it’s much easier to understand. In the class, I’ll show you what's possible and that you can play around with it.

LaRont: It sounds like a really good course. Thank you for your work with IPC. 

Burke: Thank you, Marcy.