01 February 2005
Software in Action
By Pat Dollard and Megan Ray
Bulk flow tricky to measure.
A bulk laxative plant in Arizona had digestion problems of its own. Predictive adaptive control got everything on track again.
Procter & Gamble's (P&G) facility processes all Metamucil sold around the world. P&G decided to revamp its control system to improve performance of the existing batch control technology, which was more than 20 years old.
The control system was obsolete, and spare parts were no longer readily available. In addition, the system tended to drift out of tolerance. Typically, operators had to manually adjust the system four to five times per day. This took valuable time away from workers and hindered production.
From a manufacturing availability standpoint, the goal of material transfer is to feed the exact amount of material into the mix tank in the shortest amount of time.
In most bulk material operations, dynamic scale readings differ from the actual weight fed at any moment during the feed, and in particular, at the moment of cutoff.
This discrepancy occurred in this case because the existing control system could not account for material in suspension, valve let-through, scale/filter lag, and deceleration force.
The plant needed a system that could both predict and adapt to process variations to achieve the correct material charge. In turn, this would reduce the need for operators to reset the system multiple times a day as well as reduce variation in the final product.
The ability to validate the process was also critical to ensuring compliance with U.S. Food and Drug Administration (FDA) regulations.
Predictive adaptive control (PAC), a breakthrough technology for material transfer control, works with scales, flowmeters, and weigh-belt controllers.
PAC is an on-off control algorithm that predicts the exact feed cutoff point at full feed rate. P&G developed the algorithms that support this technology and uses them throughout its operations to reduce batch cycle time and increase material transfer control accuracy.
These algorithms include an adaptive algorithm for spill and predictive algorithms for spill.
The system compensates for normal process variants that cause feed errors. This approach to material transfer generates material cost savings, improves product quality and performance, reduces operating costs, and lowers capital investment.
P&G chose to upgrade controls at the Phoenix plant with Mettler Toledo's Quantum Impact material transfer controller—matroller—and Honeywell's PlantScape C200 controller system.
The matroller embeds the PAC algorithms within. Purchasing this technology off-the-shelf ultimately reduced engineering costs and the time required for software development at the Phoenix plant.
The matroller hardware includes a JagXtreme Chassis, up to one control network I/F card (ControlNet or Profibus), and up to three input cards with a maximum of two analog load cell input cards and maximum of three flowmeter pulse input cards.
This matroller provides feed control for flow meters and load cells—loss-in-weight and gain-in-weight— while providing fast cutoff at the lowest point of control—20 millisecond response.
The maximum matroller architecture comprises 20 matrollers, 200 instruments with simultaneous feeds, and 1,000 material paths in a cluster for automated material feeds.
The Arizona plant application consisted of six matrollers clustered on Ethernet, six vibratory feeders on load cells, and about 30 material paths.
This load-and-go integration operates on standard communication protocols and system I/O. Process data, information, status, and diagnostics contained in the matroller and terminals are available to Honeywell's control system.
The function blocks provide a wealth of information to operators about material transfer, in addition to the scale readings and flowmeter values.
The availability of information through the function blocks expands the roles of the operator, enabling the operator to troubleshoot the programming and correct operational issues as needed, instead of immediately calling the engineering and instrumentation department.
Operators can also develop custom screens to track batch sequences or chart information that will help them monitor the process and make adjustments as necessary.
A common bulk transfer problem is slight raw material variation. The material transfer controller can learn this variation and adapt its control to execute accurate material transfers. This compensation occurs without the operator having to be involved.
Here is how the system works:
PlantScape initiates the sequence to begin material transfer.
The I/O will enable on/off valves to open, and the system will issue a start command to the material controller terminal over ControlNet.
The matroller terminal monitors all material transfer paths and determines when material cut-off should occur. At the optimum moment, the matroller terminal will close the on/off valves and issue a feed stop command to the system controller along with other process data including feeder weight, total amount dispensed, and any errors if they occurred.
The system I/O will command process valves to close. PlantScape retains data ownership at the controller and server level.
Accuracy sans adjustments
During the first year of use, this integrated control system has produced a high return on investment. Most notably, the number of manual adjustments required went from four or five times per day to as seldom as once a week.
The plant has also gained the ability to archive critical process data that is helpful for troubleshooting problems and which ultimately reduces system downtime.
This process data recorded in the electronic batch record is tamper proof, which means it can serve to validate the system configuration and the delivery of accurate feeds.
These records of validated information would be critical in the event of an FDA audit. IC
BEHIND THE BYLINE
Pat Dollard is section head of Control and Information Systems in Proctor & Gamble's Global Health Care unit in Cincinnati. Megan Ray is a health care engineer with Proctor & Gamble in Phoenix and is control systems manager for the company's Health Care unit.
Mass producing custom work
By Pat Barry
HMI, PLC help manufacturer hike production
When TrimJoist Corp. of Columbus, Miss., switched from manufacturing custom-built floor trusses to mass-producing wooden trusses, the company's sales skyrocketed. During the first year, TrimJoist manufactured about 3,000 feet of floor trusses per day. Now, the company produces 20,000+ feet per day.
With the growing demand for the product came the problem of where and how to test it during design and manufacture it to comply with industry regulations. To test their products during development, TrimJoist faced the costly option of transporting material and personnel to North Carolina or Wisconsin, where the nearest floor truss testing facilities were located, or the less reliable option of sending their material an hour away to a facility that wasn't equipped to test trusses by conventional methods. As volume increased, so did the need for more frequent adjustments to manufacturing processes. That's when Craig Snyder, production engineer at TrimJoist, thought it was time to invest in an in-house testing system.
The patented TrimJoist system combines an open web floor truss with a trimmable wooden-I joist. TrimJoist allows builders to custom-fit the truss on site while enabling TrimJoist to mass-produce the product.
"Small changes that we make to the product and the way it is manufactured can have a huge impact on cost and quality," Snyder said. "It's extremely difficult to evaluate these changes without on-site testing equipment."
Snyder designed the new TrimJoist test station around HMI software and a PLC that communicates via Ethernet. When a TrimJoist sample goes into the test station, pressure ramp-up and product sample failures come together based on analog and discrete inputs the PLC receives from two load cells, a linear variable differential transformer (LVDT), and a pressure transmitter. The results display on a full-color graphical interface. As the testing continues, slide bars change colors as the load increases, and the Load vs. Deflection values graph in an autoscaling XY plot shows the progression of the test in real time. Simulated gages allow operators to monitor the load cell, LVDT, and pressure values and display the pounds per linear foot (PLF) and load over deflection (L/d) ratios as well as the time duration of the test.
Snyder was able to write the control program and edit the screens armed only with knowledge gained from a seminar and software demo from Memphis, Tenn.-based distributor Bluff City Electronics.
Better quality tests
In the TrimJoist test stand application, the electronic controls "make our test equipment more accurate, more reliable and easier to use," Snyder said. The typical hydraulic test stand incorporates manually operated controls and applies the load on two-foot spacings. TrimJoist's electronically controlled machine applies pressure every foot, allowing the company to more accurately simulate the mathematical model of its trusses.
The electronic controls are also more responsive and provide more accurate control. "Machines with hydraulics can only control pressure to within perhaps a 10 PSIG range," said Snyder. "The electronic controls allow us to control pressure accurately to within 1/100th of a PSIG." In addition, the electronic controls provide a real-time snapshot of the trusses during testing. "When a truss fails, it moves," Snyder said. "However, the hydraulic cylinders can't respond quickly, and this results in a time lag and yields imprecise data. The pneumatics used on our system along with the electronic controls move the cylinders almost instantly so that we can obtain data in real time."
It all adds up to better information faster. "In the past, we depended on empirical data when designing our product," Snyder said. "Now that we have real data, we have a greater level of confidence that the changes we are making are the right ones, and we can test ideas that may be good but that we may have discarded in the past simply because we couldn't prove them."
Saving time, money
Snyder plans to use the automated test system to precisely size—and potentially reduce the amount of metal used—the plates that are a large part of the expense of building plated trusses. Currently, TrimJoist oversizes the plates to be sure they carry the design load—a reliable, but expensive, quality measure.
Future plans include performing static load tests on trusses for several days duration to determine the amount of creep. And, with the mathematical capabilities of the PLC, TrimJoist will be able to perform cyclic loading tests. The test frame can also determine E (modulus of elasticity) values of the lumber that goes into the trusses. IC
Behind the byline
Pat Barry is a Six Sigma Black Belt for GE Fanuc Automation. Barry joined GE Fanuc through GE's Technical Leadership Program. Barry has served as a territory manager and channel manager for GE Fanuc in the Southeast U.S., primarily serving the automotive and aerospace industries. His e-mail is email@example.com.
Adaptive control: Continuously adjusting the gain—proportioning action—of the control loop from a signal external to that loop. Sometimes other parameters are also modified, particularly integral—reset action. When referring to advanced control techniques, the term has come to have the broader connotation of a system of advanced process control that is capable of automatically adjusting—adapting—itself to meet a desired output despite shifting control objectives and process conditions or unmodeled uncertainties in the process dynamics. Such control is often performed through neural networks and or fuzzy logic coupled with traditional PID-type algorithms.
Predictive control: A type of automatic control in which the current state of a process compares to a model of the process and controller actions change so as to anticipate and avoid undesired excursions.