Thin-client control architecture in which programs and data are stored on central servers off the production floor instead of on PCs throughout the shop is the most significant progression between 1991 when Pepperidge Farm built a plant in Denver, Pa., and 2003, when production began at a $72 million facility in Bloomfield, Conn., reflects Dave Watson, director of engineering. Instead of backing up 75 PCs throughout the plant and worrying about the consequences of a motherboard crash, servers in remote locations maintain the system. Relatively inexpensive HMI terminals on the floor have become throwaway items that are easily replaced when malfunctions occur.
Beyond that, the biggest change Watson sees is the comfort level with the technology among vendors responsible for deploying it. "Functional specifications and PLC architecture was a bigger deal 12 years ago," he says. "Today, startups on the control side have become much more seamless, and there's a lot more flexibility to extend it."
The evolutionary nature of control technology is reflected in the sophisticated devices and higher levels of integration in plant systems devised by Thinkage, a business unit of BOC Gases. The company is working with a number of commercial bakeries, meat processors and others, including Gold Kist (see "Real Time Revolution," Food Engineering, June 2003) and Odom's Tennessee Pride.
The sensors and other hardware managed by Thinkage's on-site process control technicians aren't new. What has changed is how those devices interact with each other to modify equipment performance to match the nature of the product. Thinkage President Mark Grace characterizes the architecture as "peer to peer distributed control," a term fraught with opportunity for misinterpretation, he frets. Typical PLC architecture resembles a master/slave relationship, with a hierarchy of devices that require increasingly complicated controls as new functions are added. In a peer-to-peer configuration, devices maintain two-way communication, simplifying expansion as additional control technology is brought on line.
To illustrate, Grace cites the sequencing that occurs in a typical poultry plant's kill line. After birds are shackled and bled, the carcasses move through a scald tank and then a feather-picking machine. Controls on the scald tank react to temperature changes that occur as the mass of birds expands or contracts, but temperature set-point changes are a reaction to what was, not what will be. Result: if the number of birds coming through the line drops, the carcasses will be subjected to excessive temperatures, resulting in a loss in yield.
Similarly, controls on the picker adjust spacing to changes in carcass side, but at an interval after the change has occurred. In the meantime, excessive force can mean broken appendages and other product degradation.
"Bird sizes and weights change on a farm to farm level, and a control system can be designed to incorporate live weights into statistical calculations that sound an alarm when a machinery needs to be adjusted because a 10 percent change in weights has been observed at the front of the line," Grace explains. Picker spacing, shackle speed, flock history and bird weight and size are all variables that can be correlated and analyzed to make real-time changes in a single machine, provided that peer-to-peer architecture is used, he says." You can't manage the results without managing the symptoms. It's all about product control, which leads to process control."
Grace rejects the statistical process control (SPC) label, insisting SPC packages remain quality assurance tools relying on data points captured at a specific time to make later process adjustments to achieve a middle-of-the-road results. Adjusting the process as changes are occurring in the product itself requires real-time variable adjustments to multiple units of machinery, and that requires automated data capture and process adjustment.
SPC with visionThe same control technology that accepts data keyed in by QA personnel can accept data captured automatically. Devices capable of generating quality data are beginning to be incorporated in control architecture. An example is a vision-based inspection system at Gold Kist's Carrollton, Ga., plant.
Gainesville, Ga.-based Gainco Inc. recently licensed a kill line screening system from the Georgia Institute of Technology after two years of testing at the Gold Kist plant. A digital camera that can scan up to 200 birds a minute is screening for discoloration that indicates septicemia toxemia and visible signs of overscalding and bruising. Deviations from a body-shape template are recognized as broken wings, missing parts, mis-hung birds and empty shackles. The system effectively eliminates human grading and is a natural complement to Gainco's online electronic weighing, sizing and distribution systems, according to Gene Parets, Gainco vice president. Processors can purchase and operate the vision system themselves or apply it as part of Thinkage's contract support service.
The service-bureau model for process control also is being offered by FMC FoodTech. Last year, the firm introduced LINK, an integrated control system that bundles advanced process control with database functions such as recipe management, data collection and trending information (see "The Integration Imperative," Food Engineering, October 2003). This year, FMC has extended LINK to include Process Analysis, a module that adds analytical and relational tools to the mix to help improve yields and throughput. Belt speeds, fryer-oil temperatures and other process data are combined with product data such as food temperatures and discharge weights to identify the causes of inconsistency and to provide finished-goods tracking.
Process Analysis can be installed on any existing line, and Lundqvist expects poultry processors to be the system's early adopters, followed by other value-added processors who are being challenged by their customers to raise food-safety standards. He expects foodservice customers such as McDonald's to pressure their suppliers to install process control systems like Process Analysis. "QSRs are pushing for quick rollouts of new products, but their suppliers have difficulty meeting the new specs," Lundqvist says. "If you have Process Analysis, you will know immediately if you were in spec."
OPC Data Collector is the server for FMC's data encapsulation and relay in the system. Unfortunately, OPC is an open standard for motion-control that stirs trepidation in some manufacturing circles these days. Shorthand for object-linking and embedding for process control, OPC has been at the center of patent litigation in recent years. The Square D division of Schneider Electric patented a method of moving data and commands between a PLC and an Excel spreadsheet using OPC 13 years ago. Three years ago, the patent was transferred to a licensing firm that successfully litigated against manufacturers who used PLC-to-spreadsheet products using the OPC protocol.
A red flag was raised in the manufacturing community, Lundqvist acknowledges, but he believes new control programs have evolved so far beyond the Square D patent as to render it moot. Nonetheless, the issue persists, and FMC indemnifies Process Analysis users against any infringement claims involving Patent # 5,038,318.
While an extreme case, the OPC controversy highlights the relativity of openness in data communication standards. Absence of true openness is a windfall for systems integrators and a significant speed bump on the road to advanced control. "We support OPC as well as all the other communication protocols and standards," notes Dan Throne, industry manager for food & packaging at Bosch Rexroth, "but just because devices adhere to the standard doesn't necessarily mean they're interoperable. You can have three OPC-compliant machines that may not work when linked together.
"Controllers have been open for years, depending on whose device you're talking about," he adds. "You can have open standards for data transfer, but when you try to commoditize controllers, you've taken away the intellectual property. You will never be able to manipulate what is inside a controller."
Two schools of thoughtRexroth bundles motion and logic control and pushes it down to the drive level. The distributed architecture lends itself to high speed and multi-axes control, which is why many packaging machinery OEMs use Rexroth drives. "We can control up to 40 axes, and half a dozen OEMs are pushing us for more," says Throne. "Each time you add an axis, you don't have to go out and buy the latest and greatest 3 gigahertz processor."
Rockwell Automation applies centralized architecture in which all decisions are made for devices at the PLC level. To satisfy escalating demands for control system data, Rockwell has responded with next-generation PLCs which are being dubbed programmable automation controllers (PAC). A hybrid between the PC and PLC, PAC combines sequential, motion and process control with expanded information-handling capabilities.
"The lines between PLCs and PCs are vanishing," notes Craig Resnick, an analyst with ARC Advisory Group, Dedham, Mass. "They all contain microprocessors, they all contain software, and if they combine logic control and motion control, there's a lot more functionality in the box." GE Fanuc, Emerson Process Management's Delta V and Logix from Rockwell are examples of the migration toward PAC, Resnick says.
"People are looking to cut automation costs, and that makes a multi-purpose platform attractive" because multiple control functions are rolled into one device, according to Keith Hogan, product marketing manager for Rockwell's Logix business group. Maintenance is simplified, and when logic, motion and process controllers share a common language and single database, set-up is simplified. Programming costs to export data to MES and ERP systems also are reduced dramatically.
Rockwell's fifth and final generation of PLCs had 200 kilobytes of memory, Hogan recalls. When Logix debuted in 1998, it boasted 8 megabytes, a quantum leap which prompted Rockwell to drop the PLC name. Gen-5 PLCs had limited process control capability; Logix executes code significantly faster and eliminates the need for separate motion controls, he points out.
In IT parlance, systems that are more "robust" make it possible to perform multiple functions with a single device, simplifying the application of control technology. "In the plant, change is evolutionary, not revolutionary," ARC's Resnick observes, "and processors will extend the lifecycle of process-control equipment as long as they can." When the time to upgrade finally arrives, they're likely to find that the technology's accessibility rather than its functionality has changed, much like the punch card-driven mainframe mutated into the desktop PC.
For more information:
Daniel Throne, Bosch Rexroth Corp.
Thomas Lundqvist, FMC FoodTech
Keith Hogan, Rockwell Automation,
Mark Grace, Thinkage