Leave your master's degree in engineering or business management aside; when it comes to understanding automation architectures, it's all very elementary. These complex control systems that have become embedded in your production facilities through the years may best be described by a term we all learned in fourth grade: ecosystem.
Simply put, an ecosystem is an assemblage of organisms living together interdependently and functioning as a loose unit. Replace the word "organism" with the term "automated equipment" -- something that's not alive but lively nonetheless -- and the parallels jump out at you, clear as the lunch bell.
Why the analogy? Because it's important for manufacturers to rethink what automation architectures are. A decade ago they were closed, self-sufficient systems that didn't need anything but some bits and bytes to survive. They did the job -- which was to control a single function or process -- and they were a reliable, important element in keeping production lines going. But getting information out of those cloistered systems was nearly impossible.
As the world changes, demands on manufacturers intensify. Industrial companies need to extract data from many different pieces of equipment, assemble it and manage it from a single system, and deliver it to the enterprise. They need interoperability -- but they won't give up reliability, security, and the granularity of being able to control even the tiniest sensor tucked away in a pipeline. And they won't -- because they can't afford to -- rip and replace systems.
Recognizing these requirements and constraints, automation vendors such as ABB Inc. (Zurich, Switzerland), Emerson Process Management (St. Louis, MO), GE Fanuc (Charlottesville, VA), Honeywell (Phoenix), Invensys Process Systems (Foxboro, MA), Rockwell Automation (Milwaukee, WI), Schneider Electric Co. (Palatine, IL), and Siemens Energy & Automation Inc. (Alpharetta, GA) have revamped their technology platforms over the past several years. Collectively, they've spent millions of dollars redesigning their systems in order to expand their reach both horizontally (controlling lots of things, rather than just one) and vertically (communicating with business systems).
The results of all this hard work have trickled out in various forms over the last few years, but most vendors finally have a full arsenal of multi-disciplined, standards-based platforms that support sophisticated applications such as asset management, analytics, and the ability to lower total cost of ownership through cross-functional programs. And, in fact, the structures are radically different from earlier generations of control architectures. "You can't even think about them as architectures anymore," says Dave Woll, vice president of consulting at ARC Advisory Group (Dedham, MA). "You have to think of them in terms of features."
According to Woll, today's control systems -- which he defines as a bunch of applications collapsed onto a single backbone -- are a result of information technology's influence on manufacturing, particularly Ethernet IEEE 802.3, which is based on TCP/IP. It's a very different architecture from the token-passing 802.4 that was "proprietary in nature, had supervisory applications at one layer, controllers at another layer, and field devices at the end of a proprietary network... Ethernet forced them to come up with new architectures because they had to tear out the heart of the old ones," Woll says.
The shift was rather drastic, but also a necessary step in getting the factory floor to operate as part of a larger corporate infrastructure, using Ethernet as the physical pipe, TCP/IP as the common communication protocol, and applications as information generators. Today it's a much more interdependent ecosystem that connects production and business systems -- kind of like a river feeding into the ocean.
They All Look Alike
Let's say you are lucky enough to start from scratch with a greenfield factory project. You need a control system, so you invite in the usual suspects to tell you about their respective technology solutions. They all explain what they offer: support for all the major fieldbus networks, scalability to hundreds of thousands of I/Os, redundant controllers, and compliance with ISA-95, OPC, and various safety standards, to name a few functional requirements. These are all must-haves for end users such as Stefan Ljungkrantz, automation manager at Tetra Pak Ltda. (Monte Mor, Brazil), a food-packaging systems provider.
"We focus on following ISA standards S95 and S88 to enable present and future integration of legacy-to-business systems and also to be able to re-use engineering solutions," Ljungkrantz said in an e-mail interview. "It's important that we offer long-term stable production solutions with reliable suppliers and best total cost of ownership."
All the vendors will offer a recipe that complements Tetra Pak's and also meets manufacturers' needs. But manufacturers will soon realize that all of the recipes sound the same. Indeed, a look at the available technologies shows many similarities among the vendor platforms.
Whereas at one time a distinct line could be drawn between discrete programmable logic control environments and process-oriented distributed control systems (DCS), there's now a sizable gray area. PLCs have evolved into hybrid control systems, with pre-integrated suites of capabilities characteristic of a DCS. Meanwhile, the control systems in some cases have downsized to be more versatile and cost-effective for smaller environments.
"The problem is you hear a lot of the same words coming out of all of us," explains John Nesi, vice president of commercial marketing at Rockwell Automation. "The devil is in the details and the details are fairly subtle." In Rockwell's case, for instance, it gets down to things like how a system handles instructions and even reusability of code, he says.
Siemens, on the other hand, highlights its change management system. Emerson touts asset management of intelligent field devices and sophisticated alarm management, while Invensys talks about contextualizing data. However it is described, it always comes back to digging out data and turning it into actionable information. This is a big deal, given that most manufacturers, if not all, are not lucky enough to be starting with a greenfield. That means most are working backward; whatever is already installed is not going away.
Backward compatibility and a clear migration strategy are critical considerations for manufacturers and the vendors rolling out new systems. "The question is, who is most open to dealing with legacy," says Craig Resnick, director of research, manufacturing advisory services at ARC. There is no rip and replace going on these days, he says, and the proof is on eBay. "They are selling 150 PLCs a day because companies are still supporting legacy."
And, if a vendor does not have a clearly defined strategy for extracting information from legacy systems and managing it as part of a plant-wide information system -- a more modern way of referring to manufacturing execution systems (MES) -- there is no way to reach that critical next level, which is integrating plant systems with enterprise systems. "You can talk ERP integration until you are blue in the face, but if you haven't pulled the plant floor together, it doesn't matter," says Peter Martin, vice president and general manager of performance management at Invensys.
To that end, Honeywell built its Experion PKS (Process Knowledge Solutions) with the mindset that the DCS must be married to people, processes, asset management, and business requirements. "We put more in the envelope," says Peter Zornio, Honeywell's director of marketing. "Things that were previously thought of as being in the MES world... like production planning, yield accounting, sophisticated batch applications... we expanded to include in the DCS circle."
Pre-integrated applications mean less integration work, fewer mistakes, and more security -- three of the top items manufacturers expect from next-generation architectures. "For the Honeywell installed base, Experion is 100% investment protection because it layers over what they already have," Zornio says.
The Legacy Lives On
Last year was a good year financially for many of the automation vendors, including Rockwell Automation, which posted revenues exceeding $5 billion for the first time since separating from Rockwell Collins five years ago. Emerson announced another strong year of growth, while others like ABB, Invensys, and Siemens, which are still in reorganization mode following changes in leadership within the last 12 to 18 months, were stabilized primarily by their industrial automation businesses. Industry observers say this resurgence is due to stronger spending for technology and capital equipment. And there's a definite need: ARC reported in 2002 that there was about $65 billion worth of decades-old control systems that were outdated and required replacement.
Ripping and replacing is rarely an option, however. That limitation was top of mind for GE Fanuc engineers when they designed the new PACSystem family of programmable automation controllers. "We made it a primary focus to connect better than we ever have before into legacy," says Connie Chick, the company's PACSystem platform product manager.
The GE Fanuc approach focuses on not only programming, but the electrical hassles that can accompany a change in control. Manufacturers often decide to retain a legacy system to avoid high rewiring costs. To deal with this issue, GE Fanuc enables its customers to simply power down an I/O rack, unhinge, for example, a legacy Series 90-30 CPU, and install a new PACSystem CPU and power supply. I/Os and programs stay intact, but the PACSystem adds object programming and more connectivity to data in the systems that can be accessed by the company's Proficy plant-wide data repository.
ABB, on the other hand, developed a somewhat different method of migration using its Aspect Objects, an object-oriented technology that "allows elegant connectivity to prior generation systems as well as a conduit to ERP," according to Bob Hausler, ABB's vice president of systems marketing. Aspect Objects acts as a framework of display-object libraries for all seven of ABB's early-generation control systems. It's the difference between connectivity, -- which anyone can achieve using OPC standards -- and true integration, Hausler says. Aspect Objects transforms the operator environment to include the graphical data to which users have become accustomed. This, he says, lowers the cost associated with retraining and eases the upgrade process overall.
The common aim across all the companies' strategies is to create context. "What we need moving forward is common, actionable context," says ARC's Woll. "When you see data you need to know all of the relationships that it has," he says. For example, product data needs to have a relationship with different stages of production as well as the quality associated with it.
ABB does a very good job of this, Woll says, and others are following suit. Like ABB, Emerson, which is a strong player when it comes to the control, diagnostics, and asset management of dispersed intelligent devices, has its own object database and a data historian to record and configure changes within the system as well as all operator actions and data from field devices.
Rockwell often touts the data tags embedded in its Logix control architecture that use common language so that a limit switch is identified as a limit switch, rather than as a rack number, for example. The named tag is available to the HMI and can be stored in a common database, making it available to Rockwell's FactoryTalk Information Suite -- the glue that seamlessly transfers information between controllers, historians, and event-handling applications. "It starts with the tags but the backplane network is using the same protocol, and fundamentally the network is the control," Rockwell's Nesi says.
Siemens' migration strategy uses its Data Base Automation (DBA) product, which can read third-party tag information. Siemens does not touch the controller or the algorithms. Instead, it puts its own HMI next to the legacy system and, using a data bridge, transfers information from the legacy controller into its SIMATIC PCS 7 database. The primary driver behind this set up is preparation for unplugging the old system. "They can [run] the old and new systems to get the operators familiar with the new technology until some point in time they decide to take the old one out," says Andreas Aufenanger, director of marketing and product management for Siemens Process Automation Systems.
It's safe to say that all of the big automation vendors have a clear strategy for migrating end users from their own legacy systems to next-generation control systems. But what about leveraging connections to other vendors' legacy systems? Invensys officials say they have the solution. The company's Wonderware unit has built a very large library of connectivity drivers to a variety of non-Invensys systems. In addition, Wonderware is the first Invensys division to leverage the company's ArchestrA integration framework. But it's an architecture that will soon cut across all of Invensys' systems.
The new initiative, to be announced next month, will also answer the enterprise integration question -- can the automation system connect to higher-level business systems -- that is popping up in boardrooms, Invensys says.
Closing the Loop
Next month, Invensys will announce a new kind of automation platform that it refers to as an "enterprise control architecture."
"One of the failure points of our industry is that we've been trying to fix the automation architecture to work with the business architecture," says Invensys' Martin. "We believe that approach is destined to failure. The correct thing to do is to look at all domains as a single system."
The ArchestrA integration framework will be converged with Invensys' Foxboro I/A Series platform as well as other Invensys systems, including those from Triconex and SimSci-Esscor. Careful not to disclose too much, Martin describes the initiative as "a new class of system that is not just an automation system, it's an enterprise-level system that will go across domains that are traditional in terms of security, safety, and environmental... yet also [open] the whole thing up."
Currently an oil company in Pakistan is testing the ArchestrA-enabled Invensys solution to gain better visibility into a refining unit that has multiple control systems from both Foxboro and Honeywell. Using a SimSci modeling tool married to ArchestrA, the company was able to model the oil fields and gain a better understanding of production output. "Layered on top of Foxboro and Honeywell, it is now one unified operation," explains Mark Davidson, vice president of ArchestrA marketing.
The enterprise control system that Invensys describes and that all automation vendors will aspire to offer should help solve some of the biggest issues related to plant-to-business integration, including real-time execution at an operations management level and common systems management, according to Woll. As the underlying infrastructures of the automation system and the enterprise system begin to fuse, end users can pick and choose the applications that best meet their needs, rather than force-fit themselves into what the vendor can support.
"I'm looking for companies that provide what I call best of suite, rather than best of breed, which, in my mind, is dead," says a solutions architect who wished to remain anonymous. "You need to be able to choose a suite of applications that works best in your environment."
And, ironically, Woll says, even though it seems so much more complicated, all of this work will result in automation simplicity. "Discrete and process are essentially identical in structure and we need to eliminate the barrier [of] thinking of them separately."
Moreover, we need to start thinking of automation as part of a big-picture ecosystem. "We are in the middle of a transition in this industry," says Invensys' Martin. "These production facilities have to run as part of the business. It will happen."