Mike Boudreaux, director of performance and reliability monitoring for Emerson Process Management, has published an important article in Plant Services magazine discussing some limitations of the Purdue Model incorporating the Industrial Internet of Things. There are many more applications (safety, environmental, energy, reliability) that can be solved outside the control system. They just are not described within the current model.
Interestingly, about the same time I saw a blog post at Emerson Process Experts quoting Emerson Process Chief Strategic Officer Peter Zornio discussing the same topic.
I’ve been thinking about this for years. Mike’s article (which I recommend you read–now) brought the thoughts into focus.
Purdue Enterprise Reference Architecture Model
The Purdue Enterprise Reference Architecture Model has guided manufacturing enterprises and their suppliers for 25 years. The model is usually represented by a pyramid shape. I’ve used a diagram from Wikipedia that just uses circles and arrows.
This model describes various “levels” of applications and controls in a manufacturing enterprise. It describes components from the physical levels of the plant (Level 0) through control equipment and strategies (Level 2).
Level 3 describes the manufacturing control level. These are applications that “control” operations. This level once was labelled Manufacturing Execution Systems (MES). The trade association for this level–MESA International–now labels this “Manufacturing Enterprise Solutions” to maintain the MES part but describe an increased role for applications at this level. The ISA95 Standard for Enterprise Control labels this level as Manufacturing Operations Management. It is quite common now to hear the phrase Operations Management referring to the various applications that inhabit this level. This is also the domain of Manufacturing IT professionals.
Level 4 is the domain of Enterprise Business Planning, or Enterprise Resource Planning (ERP) systems. It’s the domain of corporate IT.
Hierarchical Data Flow
The Purdue Model also describes a data flow model. That may or may not have been the idea, but it does. The assumption of the model that sensors and other data-serving field devices are connected to the control system. The control system serves the dual purpose of controlling processes or machines as well as serving massaged data to the operations management level of applications. In turn, level 3 applications feed information to the enterprise business system level.
Alternative Data Flow
What Mike is describing, and I’ve tried sketching at various times, is a parallel diagram that shows data flow outside the control system. He rightly observes that the Industrial Internet of Things greatly expands the Purdue Model.
So I went to the white board. Here’s a sketch of some things I’ve been thinking about. What do you think? Steal it if you want. Or incorporate it into your own ideas. I’m not an analyst that gets six-figure contracts to think up this stuff. If you want to hire me to help you expand your business around the ideas, well that would be good.
I have some basic assumptions at this time:
- Data is not hierarchical
- Data has many sources and many clients
- Eventually we can expect smart systems automatically moving data and initiating applications
Perhaps 25 years ago we could consider a hierarchical data structure. Today we have moved to a federated data structure. There are data repositories all over the enterprise. We just need a standardized method of publish/subscribe so that the app that needs data can find it–and trust it.
Now some have written that technology means the end of Level 3. Of course it doesn’t. Enterprises still need all that work done. What it does mean the end of is silos of data behind unbreachable walls. It also means that there are many opportunities for new apps and connections. Once we blow away the static nature of the model, the way to innovation is cleared.
OpenO&M Model
Perhaps the future will get closer to a model that I’m writing a series of white papers to describe. Growing from the OpenO&M Initiative, the Open Industrial Interoperability Ecosystem model looks interesting. I’ve just about finished an executive summary white paper that I’ll link to my Webpage. The longer description white paper is in process. More on that later. And look for an article in Uptime magazine.
Data is not hierarchical.
Data has many sources and many clients.
This is probably the most concise way to say these very obvious facts and should provide absolute direction to manufacturing enterprise IT strategies. Specific information from anywhere to anywhere delivered in a role based form should guide us quickly away from any layered approach.
Hierarchies are having a tougher times in the 21st century with everything & everyone but a hyperlink or social network connection away. Now we’re rapidly evolving sensors, cloud storage & access, analytics and remote expertise into this non-hierarchical mix.
Gary, I appreciate the callout to the blog post, too!
Hi Jim, thanks for commenting. I still read your blog. It’s only you, Carl Henning (Profibus), and me still at the blogging thing. Technology is coming together quickly. It’s going to be interesting. And listening to Mike and Aaron from Emerson, looks like you guys are right there.
Good point Gary, “Data is not hierarchical and has many sources and clients”. The models we talk about were all developed considering several constraints when they were developed – storage, transmission, presentation, number crunching, correlation etc. As some of the constraints get relaxed, now we have no difficulty getting a real time data directly into Level 4 if the need be (and present it in a way that can be understood at that level). I use to think about how a Cloud expert will explain about it to a Mainframe expert who has come from that age and not aware of anything after that. Same way, explaining data analytics to a dynamic process model builder using plant time sequence data. Many a times it leads to a basic concept which was realized earlier subject to constraints at that time.
Love this discussion, Gary! Looking forward to the whitepaper and article in Uptime.
I love concise, @Michael, so the IIoT bumper sticker can just say:
Data is not hierarchical.
Data has many sources and many clients.
We see the absence of hierarchy now as well. Complex devices like robots, drives, and some process instruments connect as I/O using PROFINET, but use OPC UA to reach HMI, historians, and analytics. The future is not a pyramid.