Model-based controllers have been used for applications as diverse as steering supertankers through narrow channels and maintaining comfort levels in commercial office space. It should come as no surprise, to find a model-based controller in the Molson Breweries of Vancouver, British Columbia, Canada.

This particular model-based controller is BrainWave from Universal Dynamics Technologies (also Vancouver, BC). The controller keeps two kettles of liquid "wort" (the sugary end-product that results from mashing and infusing malt) boiling as vigorously as possible.

A vigorous boil is characterized by a high evaporation rate and a thick layer of foam above the surface of the liquid wort. The foam thickness or "boil height" is easier to monitor in real time, so it is used as the feedback variable in the boil control strategy. It can be raised or lowered by opening or closing the valve that feeds steam into the kettle jackets and directly into the wort.

The control objective is to maximize the boil height without causing the kettle to overflow. Several external factors affect the boil height. Fluctuations in ambient air pressure (resulting from climatic changes or the starting and stopping of the venting fans) can raise and lower the boiling point of the liquid wort. Addition of hops into a kettle releases oils that dampen the boil and reduce the boil height.

At the other extreme, safety overrides prevent the wort from boiling through the open ports at the top of the kettles. Before BrainWave was installed, such "boil-overs" were prevented by limiting the steam valve to a maximum opening of 83% and by limiting the fill level to just 725 hectoliters (hl, or 19,140 gal) in the 940 hl kettles.

A programmable logic controller was originally used for boil height control. The PLC first measured each kettle's boil height using an 18 in. probe with three discrete levels. It then opened or closed the kettle's steam valve based on how many of the three levels were covered and for how long. This control strategy worked well enough and managed to keep the evaporation rate just above the 5% per hour target value. However, it proved to be unduly conservative. The steam valves were moved too slowly to use the steam effectively, and the limited fill level kept batch sizes smaller than necessary. Improving the control loop's performance with traditional proportional-integral-derivative loops proved difficult since the loop behavior becomes nonlinear at the point where the malt starts to evaporate.

Molson needed a control strategy that could handle boil height more efficiently. It chose BrainWave for a three month trial period during which the PLC remained in control of the first kettle while BrainWave, running on a Microsoft Windows NT PC, took over the second. The two kettles were operated in parallel, producing the same products during the test.

The two controllers were not playing by exactly the same rules, however. The steam valve under BrainWave's control was allowed to open 100% in order to realize the maximum heating effect from the steam. Although this strategy risked a boil-over, none occurred during the test (and none has a occurred since BrainWave was put in charge of both kettles).

BrainWave's predictive abilities made the difference. Using its dynamic model of the boil process, BrainWave was able to forecast the boil response so that the boil height objective could be reached as rapidly as possible without overshoot. BrainWave reduced manipulation of the steam valve since it knew in advance where the valve needed to go.

The tricky part was generating a mathematical model suitable for making accurate projections of the boiling behavior; that's where BrainWave's adaptive capabilities came in handy. By observing the effects of its recent control actions, BrainWave automatically determined an empirical input/output relationship for the boil process. No operator intervention was required to create the initial model or to update it as the process dynamics changed.

BrainWave managed to increase the evaporation rate from 5.17% per hour to 5.59% per hour. The difference resulted in an 8.1% improvement in the volume production rate. Brainwave was also reduced peak loading and overall steam usage by using exactly as much steam as necessary.

The improved boil efficiency in turn reduced the amount of hops used, saving over $40,000 per year. BrainWave paid for itself in less than a year.

Future benefits are expected as well. By suppressing boil height overshoot, BrainWave should allow for larger batches to be brewed in fuller kettles. Rather than a 3.5 ft gap between the nonboiling wort level and the rim of the kettle, Molson expects to run closer to a 2.5 or 3 ft "head space." Batch volumes should increase by 3 to 6%.

For more information, visit www.controleng.com/info .