As shown on the pyramid on page 90, regular regulatory control (RRC) using PID (proportional, integral, derivative) loops is typically sufficient for relatively simple processes. PID control has trouble, however, dealing with more complicated processes which may be nonlinear (fluctuating); involve long time delays; are subject to frequent dynamic changes due to recipe, batch or load changes; involve complex relationships between process variables; or require control of multivariables which cannot be handled by single-loop PID controllers. PID loops also require repeated manual tuning.

Advanced regulatory control (ARC) handles these complex loops, and without manual tuning. Advanced process control (APC) generates and controls supervisory setpoints to optimize production and product quality.

New adaptive and predictive control systems are providing advanced regulatory and process control for complex processes. Most are model-based; one is model-free.

MPC at Anchor Products

As reported earlier by Food Engineering (November ’97, December ’98), Anchor in 1997 formed an alliance with Pavilion Technologies (Austin, TX) to apply Pavilion’s Process Perfecter and Process Insights multivariable model predictive control (MPC) technologies to evaporation and drying operations at Anchor’s new Waitoa milk powder plant.

Process Perfecter is a neural-network model-based system which extracts historical data from process controls and solves relationships between complex process variables. This new knowledge is then continuously applied to the model in real time to optimize the process. The model is non-linear — i.e., the learning process is continuous, not intermittent — for steady-state optimization of the process. Process Insights builds “soft sensors” known as Virtual OnLine Analyzers (VOAs), which create neural-network non-linear steady-state models relating inputs and outputs.

As of October ’98, these systems had increased capacity by 7.4 to 8.2 percent, reduced variability in concentrate density by up to 70 percent, and greatly exceeded ROI targets on three evaporators at Waitoa. “It’s only a matter of time before MPC will be as common in the control engineer’s toolkit as PID is today,” said Ian D. Steele, engineering manager at Anchor Products.

On August 9 of this year, Anchor and Pavilion announced a three-year extension of their alliance to apply Pavilion’s Production Chain Optimization suite to new and existing Anchor plants until 2003, boosting Anchor’s cumulative investment in Pavilion systems to more than $10 million. The suite includes several Pavilion packages which enable the user to learn, predict, control and optimize continuous processes within a unified software environment.

Paul Sabiston, advanced process control manager at Anchor, updated Anchor applications September 15 at “Insights ’99,” the Pavilion Technologies Users Conference in Austin. Since commissioning the first Process Perfecters at Waitoa in May 1997, Anchor has implemented or is currently implementing 27 additional Perfecter and VOA projects applied to evaporator, dryer, boiler and cheesemaking operations in various plants. Company-wide rollout was approved last March and Anchor plans 150 more such projects. As at Waitoa, the Pavilion software will integrate with Anchor’s existing process control systems based on Allen-Bradley PLC, Realflex SCADA, DCS, relational database, and Ethernet TCP/IP systems. Anchor now has nine engineers trained in advanced process control to manage these projects. All have chemical engineering backgrounds and are graduates of the company’s Dairy Technology Graduate Training Program as well as Pavilion courses.

As a project case study, Sabiston presented the implementation of Process Perfecters and VOAs on Drier No. 1 at the company’s Te Rapa plant. Drier No. 1 consists of a spray dryer and two vibrating fluid-bed (VFB) dryers, with moisture content of the final milk powder measured by near infra-red (NIR) instrumentation. Project phases included plant testing, model building, process simulation, system tuning, and commissioning. A Process Perfecter measures disturbance variables and manipulates outlet air temperature on the first fluid-bed dryer to compensate for the variables and optimize moisture content.

Within three days of installation, Sabiston reported, the system achieved 50 percent reduction in moisture variability and — at this writing — the project team continues tuning the system to sustain performance benefits. To further sustain this improvement, the plant will integrate Pavilion’s web-based Economic Insights software, which measures and reports system performance and compares performance against the project’s targeted Internal Rate of Return (IRR), a measure of ROI.

MFA at Del Monte

During the three-month tomato-harvest season, the plant converts 350,000 tons of tomatoes into canned and aseptically bagged tomato paste, as well as canned tomatoes. From early July through early October, the plant operates 24 hours per day as a continuous caravan of gondola trucks unloads tomatoes into flumes feeding the hot-break lines. Continuous throughput is critical to cost efficiency during the short processing season.

Product flow, however, is irregular between truckloads, causing temperature variations in the rotary-coil hot-break systems. Hot-break process variables include tomato flow, steam pressure, condensate pressure and temperature of the incoming product, but the biggest variable is flow rate, says Plant Manager Michael Maulhardt. “Flow rate can change from zero to 50 tons per hour in minutes,” he adds. The optimum hot-break temperature (to inactivate pectinase enzymes) is 205?F, but the intermittent flow rate formerly caused temperature to vary as much as 15 to 20 degrees. Variations in hot-break temperature caused variations in viscosity (density) of the sauce pumped to the evaporation system, which in turn caused variations in product quality and energy consumption. The first stage in the evaporation process is mechanical vapor recompression (MVR), which is sensitive to input temperature changes, Operations Manager Rick Fenaroli points out.

So maintaining optimum temperature in the hot-break process was the major problem, says Fenaroli. The PID loop controlling the steam valve which regulated hot-break temperature was incapable of optimally adjusting temperature to compensate for the intermittent flow rate.

In June of this year, prior to the start of the 1999 campaign, the plant installed a CyboCon model-free adaptive (MFA) controller with nine loops to control temperature on the hot-break lines. The system is supplied by CyboSoft, a unit of General Cybernation Group, Inc. (Rancho Cordova, CA). Unlike model-based control systems, CyboCon MFA requires no process identification, precise knowledge of the process or controller design. Unlike PID, it requires no complicated manual tuning. Described as a “plug-and-play” package for single-variable and multivariable control, CyboCon integrates with major PLCs, DCS and HMIs.

At the Del Monte Woodland plant, CyboCon integrates with an Intellution FIX SCADA system communicating with Allen-Bradley PLCs. CyboCon was installed in just a few hours by system integrator Easi, Inc. (Tracy, CA). The PID loops were retained, offering the operator a choice of control, “but since installation the operators have used CyboCon 100 percent of the time,” said Fenaroli.

Product temperature now varies no more than plus 2 to minus 5?F. As the 1999 campaign drew to a close early in October, there had been no failures in any of the nine CyboCon loops. Maulhardt and Fenaroli are now evaluating further applications “where you have multiple inputs and a single output,” such as in boiler control and reducing solids variability in evaporator control.

At ISA Intech/1999 October 5 in Philadelphia, CyboSoft introduced CyboCon CE, which combines CyboCon MFA control technology with Microsoft’s Windows CE real-time operating system. According to Microsoft, Windows CE is a compact, scalable, modular operating system designed for products which use embedded systems, including industrial controllers, data terminals and “new PC companions” such as handheld PCs, palm-size PCs and voice-activated PCs. CyboCon CE is embedded in GE/Total Control’s FactoryClient 2000 Industrial CE Platform for adaptively controlling complex processes such as extruders, dryers, evaporators and distillation columns.

Model-based UHT control

Developed by Universal Dynamics Technologies (Vancouver, BC), Brainwave models a process, including dead time, and uses these models to predict the effect of control changes and disturbances to reach setpoint smoothly and quickly. It applies model-based feedforward control to observe and learn the effects of disturbances, then models and includes their effects on the control action. Brainwave is also adaptive; it tracks changes in process and feedforward models and continuously adapts the models, eliminating the need to tune loops. It can build and apply up to 10 process models to provide precise control of the same loop. Last year, at ISA Expo ’98, Universal Dynamics received the ISA President’s Award for significant innovation.

At Kan-Pak, Plant Engineer James Steinbacher re-engineered an aged helical-coil tubular heat exchanger and equipped it with new instrumentation and controls. “We put our own PLC on it, wrote our own software and incorporated Brainwave in it,” says Steinbacher. Brainwave reads several inputs to build a model relating steam pressure to product temperature and controls the temperature. Products include Cappuccino and fruit-flavored beverages, some dairy based, aseptically bagged on two DuPont DA-2000 machines and a Rapak Intasept aseptic bag-in-box filler. The aseptic bags fit beverage dispensers in convenience stores and foodservice establishments.

Brainwave is reportedly applied also by Ocean Spray Cranberries, Inc. to control temperature and flow rates in the juice pasteurization process at the company’s Henderson, NV plant.

Snacks, cheese and citrus

QuickStudy generates process models from either historical or real-time operating data without “bump testing,” uses the models to predict the direction the process is taking, and takes corrective action before the process can deviate from setpoints. This eliminates the need for loop tuning inherent in PID systems. Models can be continually updated to meet changing process dynamics. It predicts the effects of control changes and feedforward events to let the process reach setpoints quickly and without oscillation, even with long time delays or variable dynamics.

At ISA Tech ’99, Adaptive Resources introduced Q-MAST, a modeling and simulation tool that develops process models off-line from a few days of historical data, and predicts process behavior in the simulation mode. Q-MAST can be used with a manufacturer’s existing control system or with QuickStudy. Q-MAST received the ISA President’s Award for innovation at ISA Tech ’99.

QuickStudy systems are currently tightening variability in frying temperatures, drying temperatures, oil and flavor application for a major multi-plant manufacturer of salty snack foods. For a major cheesemaker, QuickStudy controls evaporation temperature to minimize variability in whey solids. At a Florida citrus-juice plant, QuickStudy is embedded in the OEM’s process-control system to control the variability of Brix in orange-juice concentrate.