In Stay Sharp Episode 112, The Data Models Dilemma in Digital Engineering, co-hosts Jonathan Scott and Juliann Grant discuss a subject of growing importance in the manufacturing world: data models. While other podcast episodes have touched on the topic, Jonathan and Juliann spend some time investigating the role of data models in digital engineering and how they impact PLM, digital twins, and the digital thread—in short, why data models matter.
Why Data Models Are Important Now
Jonathan explains a data model is an “informative representation” of a product, process, or other aspect of digital engineering. Put more simply, it’s the data you capture that represents the thing you care about. Part of the point—and the challenge—of data models is that each software application and manufacturing organization represents their data in a particular way. Juliann notes, “From a software perspective, nobody’s using the exact same data model. And then every manufacturer has a unique data model. There isn’t a lot of standardization.”
The reason data models matter so much now is the prominence and prevalence of digital twins. To create digital twins you need a digital thread, and a digital thread is built of lots of different types of data sitting in multiple, siloed systems—each of which has its own data model that’s appropriate to that system’s purpose and context. “Let’s take a BOM component as an example,” Jonathan says. “The way a BOM component looks in PLM is not the same way that a BOM component looks in MES, is not the same way a BOM component looks in ERP. In fact, they don’t even call them the same thing.” He adds, “How do you put it together if it doesn’t match?”
Because a manufacturer’s goal is to get a specific question answered, the data models you have must be specific to the information required to make business decisions, create or support products, and more. The data models and digital threads will reflect your business purpose. Jonathan explains, “That’s why data models matter, you need them to be as comprehensive as they can be to give you those answers you want. But they need to be flexible because the questions you ask will change tomorrow.”
Why This Isn’t Solved Yet
The variation in how each application structures data models is part of what makes data models a complex and thorny problem. Another is the fact that every manufacturer’s business is different. “If no two digital threads are the same,” Jonathan asks, “how could you have a standard data model?” In addition, compute power has finally caught up with the sheer volume of data points that need to be collected and analyzed.
There’s also no consensus yet on the “best” way to handle digital threads and digital twins, specifically how to collect the data they require. One school of thought is creating data lakes or data warehouses in which all data is stored. Another is linking all the systems together, but keeping data in its original location, because each system fulfills its own specific purpose in ways a single system couldn’t accomplish. “There’s this important concept in engineering about separation of concerns. That you do one thing over here and you do it well, you do another thing over there and you do it well. Because if you try to do everything all at once, sometimes that fails,” Jonathan says. “The problem is how do you put it together and do you have the right data? That’s the one that I think is a big challenge. Did you even collect the right information in the first place?”
How You Should Get Started
The way to start is to understand what outcome you’re looking for, at least for the short term—because your desired outcome will adjust over time and so will your data models. With the outcome in mind, work backward to the right data model to achieve it, and back that up by ensuring you’ve got good data going into the model. Then simply start on something. Jonathan says, “Think about what you need, in what order over time. Are there dependencies to it? Like, well, I can’t solve that until I’ve solved this. Those are important things to have in mind, but you’ve got to break it down and solve those problems a piece at a time.”
Learn More About the Importance of Data Models
The full podcast contains more discussion and details on a variety of topics, including the risks you might be taking in breaking the contextual element of the data model, why you can’t forget about carrying forward historical data, how your data model is also going to feed your AI engine and help move business forward, and more.
Be sure to check out Stay Sharp Episode 112: The Data Models Dilemma in Digital Engineering and join us each week for a new podcast.
Follow the Razorleaf Podcast, Stay Sharp in Digital Engineering on:
