Predição da cor da cerveja filtrada: software de automação para predição

Abstract

Beer is commonly produced with four inputs: water, malt, hops, yeast and, in some cases, adjuncts. One of consumers' first impressions, even before drinking beer, is the perception of the color of the liquid, which makes this characteristic extremely relevant to beer drinkers. In this way, breweries control the process steps so that the beer produced meets the standards specified for each style. However, variations in raw materials and fluctuations in industrial process parameters cause deviations in the results obtained. In the brewery that was the subject of this study, deviations reached 3.2% of batches outside the specification, which causes product blockage and the need for correction. In this work, the amount of roasted malt to be dosed in the manufacturing process could be prescribed in approximately 8 hours after the start of the production stages based on the properties of the raw materials, parameters of the industrial process and mathematical modeling. For complete and assertive modeling, the influences of process variables in the brewing, fermentation and filtration stages were also correlated. The study was able to leverage the brewing process in terms of technology and data science, identifying which manufacturing steps influence the final color of the product. The main gains were the assertiveness of the color parameters of the brewed beer, reduction of operational saturation, reduction of rework for color correction, increased performance of quality indicator results, reduction of roasted malt consumption and leveraging of knowledge in artificial intelligence. The results achieved were the reduction of fermenter color deviations from 3.2% to 0.5% and generated financial savings of R$ 2 million/year due to the optimization of the use of roasted malt.