TABLE OF CONTENTS Imagine that you are the master brewer for your neighborhood microbrewery! For an upcoming event, you are to make a large batch of beer. The typical fermentation cycle takes a few weeks, so you have to make a judgment call as to when the perfect brew is ready (and avoid the wrath of your patrons). Biochemistry teaches us that one of the critical indicators of brew quality is the diacetyl content. However, you need to take the brew sample to a chemical analyst to measure that content, as no viable physical sensors exist.
Now imagine that you can get a new software program that will provide a real-time measurement of the diacetyl content throughout the fermentation cycle. You can monitor the quality, make sure your beer is not getting "malflavors," and stop the batch as soon as it reaches the desired value. Such a software sensor is now possible thanks to neural network (N N) technology [7-20].
When neural networks came on the industrial scene in the 1980s, they were viewed as a panacea for many of the difficult or unsolvable control and measurement problems that plagued the industrial control community. However, when neural networks were applied to these problems, few successes occurred. Initially, this gave them a bad name, but today there are many...
