Predicting the speed of beer fermentation in laboratory and industrial scale

Abstract

Characteristic of the beer production process is the uncertainty caused by the complex biological raw materials and the yeast, a living organism. This uncertainty is exemplified by the fact that predicting the speed of the beer fermentation process is a non-trivial task.

We employ neural network and decision tree learning to predict the speed of the beer fermentation process. We use two data sets: one that comes from laboratory-scale experiments and another that has been collected from an industrial scale brewing process. In the laboratory-scale experiments a neural network that employs characteristics of the ingredients and the condition of the yeast, could predict the fermentation speed within 2% of the true value. Decision trees for classifying whether the speed of fermentation will be slow or fast were constructed from the same data. Astonishing simple decision trees were able to predict the classes with 95%–98% accuracy. In contrast to the neural net experiment, even the highest accuracy could be reached by utilizing only standard brewery analyses.

We then set out to check the utility of these methods in a real brewery environment. The setting in the brewery is more complex and unpredictable than the laboratory in several ways. Regardless, reasonably good results were obtained: the neural network could, on average, predict the duration of the fermentation process within a day of the true value; an accuracy that is sufficient for today's brewery logistics. The accuracy of the decision tree in detecting slow fermentation was around 70%, which is also a useful result.