Centralized vs. Distributed: Is Bigger Better?

To distribute or not to distribute, that is the question. Is it nobler to use one, large, centralized control system for the entire plant or process? Or is it better to use a number of small, field-based distributed controllers, each dedicated to a particular machine, process, or control loop? Those are the questions, and the answers are changing rapidly due to advances in communications, ruggedness, and software standards.

Process control systems used to be highly distributed because large, centralized, computer-based control systems were simply not available. Local processes were controlled by a combination of relays, timers, and pneumatic or electronic loop controllers. These hard-wired systems were interconnected by contacts and 4-20 mA signals. Operators and technicians roamed the plant in droves to monitor operations and maintain system components.

Ironically, the misnamed distributed control system (DCS) was perhaps the leading agent for control centralization and the chief cause of the demise of distributed controllers. These systems were installed in central control rooms and connected to field devices through miles of dedicated wiring. They served the industry well for decades and did much to improve the operation of large process plants.

In the past few years, a new class of field-based controllers has emerged and is making distributed control a more viable option. These controllers are small, inexpensive, rugged, and equipped with high-end communication capabilities. Cost-effective control of local processes is now possible, and fast, reliable networking of these islands of control is easily accomplished.

Controllers that can be field-based in a distributed control system include micro-PLCs, smart I/O, smart motor starters, smart pushbuttons, smart valve controllers, field-mounted solid-state controllers (Figure 1), and single and multi-loop controllers. Smart products can perform local logic or process control functions and can communicate with central controllers over digital communication networks.

As recently as five years ago, field-based distributed control was impractical. In 1998, a 100 I/O PLC with an open communications protocol such as Ethernet cost about $5,000. At those prices it just didnt make sense to distribute control to the field. Today, a 100 I/O PLC with Ethernet can be purchased for less than $500.

Comparable performance increases can be found in field-mounted solid-state controllers and in single and multi-loop controllers. Smart components with open communications protocols such as I/O, motor starters, pushbuttons, and valve controllers are recent additions to the field-based distributed control team. These products add local logic and process control along with digital communications into footprints not much larger than those required for similar "dumb" components.

These incredible leaps in performance make distributed control a much more attractive option for many processes.

Technology has made distributed control viable, but process needs are driving its adoption. "Field-based controllers are faster and can be dedicated to a specific control application," says Matt Bothe, senior automation engineer, CRB Consulting Engineers, Plymouth Meeting, Pa. "Distribution of control functions can reduce wiring, facilitate maintenance, and increase the speed of logic and process control."

Field-based controllers also offer other advantages over centralized control systems (Table I). North Fork Electric, Crumpler, N.C., is a systems integrator for AutomationDirect and other vendors. President Andy Feimster says, "In critical control applications, distributed controllers may offer a level of safety and dependability that centralized control systems (whether they be PLC, PC, or DCS) cannot."

Field-based controllers reduce wiring, but so do centralized control systems with remote I/O. The problem is these remote I/O systems can cause heavy traffic on communications networks. Moving intelligence to field devices and to smart I/O allows control to be accomplished locally, thus reducing network traffic.

Status information concerning the local process is still transmitted to other controllers, but time-critical data used for real-time control does not need to be sent over the network. "Less information needs to be handled by the communication path, and I can choose exactly what information to send to the main control system," observes Paul Parker, cogeneration plant engineer with KSL at Los Alamos National Labs, Los Alamos, N.M.

Some processes are just too fast-acting for a centralized controller. "A single controller running a lengthy program can be slow on event-triggered interrupts," says Jeff Birch, president, Industrial Control Services, Centralia, Mo. "We use several dedicated controllers to isolate critical programming and maintain reaction times."