We’ve heard a lot about the evolution of automation architecture, particularly as companies embark on digital transformation plans. But what are the drivers of this march toward what Emerson terms “boundless automation?" To find out, Control talked with Peter Zornio, Chief Technology Officer at Emerson.
Q: What is driving this inflection point - the evolution of automation architecture - in the industry and why now?
A: First, the automation architecture we work with today has been fundamentally the same since the mid-1990s. The last big technology switch was moving to open systems; Windows, PCs, Ethernet networks and utilizing what was at the time state-of-the-art IT technology for automation. Since then, there's been huge advances in IT computing with the cloud playing a large role. So, it's time for the automation world to think about how we're going to take advantage of these technologies that the IT world pioneered, because historically we adopt those after they are proven. And as customers implement digital transformation, they run into limitations because of current systems and the very siloed data repositories and architecture they've had up until now. Plus, the big push around analytics, artificial intelligence (Al), and new technology we can use to analyze data to drive better results and optimization means there's also a new drive for integrated data that's contextualized in a way that Al and analytics can work with it.
Q: Is boundless automation only for new construction, or can it be adopted for existing operations?
A: For greenfield facilities, it's a no brainer. We have a customer who was building a new facility and said, “Transforming sounds hard. I want to get it right day one. I want to be born digital.” “Born Digital” as a term stuck, and we've been using it ever since. Plants that are born digital have many of the core elements of the boundless automation vision.
“Born Digital” new construction means putting all the infrastructure in place from the beginning, including a comprehensive data infrastructure. It also means building a plant with a digital twin from the beginning and using the digital twin to train operators as well as figure out new ways that you can change operations. It includes designing for reliability-centered maintenance, adding wireless instrumentation for equipment condition data, and being able to diagnose equipment's health using both first principles and Al-based models. Starting up with state-based control and APC from the beginning. And it means a data management strategy that is secure but more open, making data accessible from anywhere, by anyone who needs it. All these things we talk about in digital transformation need this infrastructure and more.
For existing facilities, for the last six, seven years, we've been trying to put in many of these elements, such as a comprehensive data infrastructure, into those facilities. It can be done a piece at a time, but every piece must be cost-justified, and have a return on investment (ROI).
Q: What are the technical differences with this kind of architecture?
A: Some of the technical differences and the technologies include new field connectivity technology such as Advanced Physical Layer (APL), which is going to provide a bigger data pipeline for the field. The fact that it's going to run already well-established standardized Ethernet protocols should make integration easier between field devices.
We're also pioneering a new family of wireless devices. We see the low power sensor bands that are built into cellular as a great new opportunity for more internet of things (loT) wireless devices. Narrowband loT (NB-loT) and LTE-M are special radio bands that are rolled out across most cellular infrastructures in most countries and are specifically designed for low-bandwidth, long-range sensor data. You can build a sensor that will run a long time on a battery like we do with WirelessHART in a plant but use the existing cellular infrastructure. We see a whole generation of mobile sensors coming out of that, not just for inside the plant, but anywhere. For instance, a product we're coming out with soon goes on chemical tote tanks, a radar level sensor that goes on the top and tells you where that tank is anywhere in the world and what's the level measurement.
As we move to the more modern Linux based edge technology, we believe we can make that software execution environment, running on off-the-shelf hardware, reliable and low-latency enough to run control. The last piece of proprietary purpose-built hardware in the automation system is the controller, and we can see it eventually being replaced, just like the rest of the automation architecture has, with off-the-shelf hardware and off-the-shelf operating systems, all moving to the edge. Then the automation system is truly, purely “software defined” - an integrated suite of software that runs on open technology.
Q: What is unified data and how can it be achieved?
A: As people embarked on their digital transformation programs, they found often the
hard part is figuring out how to get the data together in a consistent context where they could apply the analytics or AI. A core part of boundless automation, at least for the suite of Emerson and AspenTech software applications, is a common, consistent data infrastructure and data model. This unified data infrastructure spans what we think of today as level two, level three and level four, all together into a unified data “fabric.” We also see a new way to do cybersecurity. ln the IT world, they talk a lot about what's called zero-trust security, especially for authentication. Today, for OT, we rely on layers of networks, the Purdue model and what we call layers of defense with firewalls in between. That's not going to work well when we need to have a hyper-connected and hyper-data-sharing architecture. For OT, we envision zero trust evolving to where you don't depend on where you are in a network or segmentation of networks, it's more about the applications and users having to authenticate and authorize themselves with each other application regardless of where they are executing.
Q: How are standards organizations evolving to enable this?
A: The place where there's the most progress being made on coming up with some standard data models is in the field. For example - the German chemical organization NAMUR together with FieldComm and other field standards organizations are advancing a standard called Process Automation Device Information Model (PA-DIM). They're defining a standard set of data that every device of a certain type would have, that would be consistent regardless of manufacturer. There are also standards efforts to try to standardize a subset of data for pumps, compressors, more elaborate pieces of equipment. Not as much progress has been made on having those standards, although PA-DIM will be moving into that realm. That's an area where end users figured out that having some standard definitions for each equipment type will greatly improve and simplify data availability.
Q: What is the relationship of Al to boundless automation?
A: I often like to say, in the automation world, we were Al before Al was cool. We've been using Al since the mid-1980s with multivariable predictive control. DeltaV came out in mid-1990s with neural networks as a standard function. Machine learning has been diagnosing equipment health since the mid-1990s. Most of AspenTech's products have embedded Al, whether deep learning, neural networks or machine learning.
The new generation of technology, large multipurpose general models that are multimodal like GPT, contain good, large language models that open the possibility of new applications like operator advisory, very advanced product support chatbots, and auto configuration of systems with language as the interface. And for operations, all of this AI needs an infrastructure that makes data easily available in a consistent contextualized manner. So that's where, again, it comes back to the unified data fabric being an important part of the strategy.
Ultimately, the boundless automation vision is about making data more accessible and easier to convert into actionable insights by pushing computing power closest to where it can add value. AI is one very exciting use case for that vision, but there are countless other ways that a boundless automation strategy will help organizations operate more effectively.
