Desenvolvimento de softwares para classificação do café cru e torrado através de indicadores químicos e físico-químicos de qualidade

• Autor(a) principal: Lima, Priscilla Magalhães de
• Data de Publicação: 2017
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Repositório Institucional da UFLA
• Texto Completo: http://repositorio.ufla.br/jspui/handle/1/12598
• topic: Redes neurais; Café - Qualidade; Café - Composição química; Café - Classificação; Neural networks; Coffee - Quality; Coffee - Chemical composition; Coffee - Classification; Ciência de Alimentos

The coffee quality evaluation is performed by physical analysis of the grain and through sensorial analysis, by the "cup-proof" technique. However, this is a subjective classification which may vary from individual to individual. Thus, several researches have been done to relate the beverage sensorial characteristics with the chemical and physicochemical analyzes of the raw and roasted grains, helping on the quality assessment. In this way, this work aimed to create softwares capable to classify raw and roasted coffee according to their beverages class based on their chemical and physicochemical parameters. The raw grains were submitted to the physicochemical analysis: potash leaching, electrical conductivity, acidity, pH, soluble solids, enzymatic activity of polyphenoloxidase and total sugars. The results of these analyzes were used as training data and validation of the neural network of the Raw Grain Classification Software (Classcafe 1.0). Then, the coffee samples were sent to the cup test by trained providers to confirm the classification obtained in the cooperatives. After grading, the roasted beans were submitted to analysis of total sugars, reducing sugars and non-reducing sugars, pH, soluble solids, acidity and ethereal extract. The results of these analyzes were used as training data and validation of the neural network of the Roasted Grain Classification Software (Classtorr 1.0). The neural model used in the developed system correctly classified 100% of the samples tested.The neural network was able to correctly classify the raw and roasted coffee according to its sensory class using the chemical composition data of the coffee beans. The created systems are friendly and easy to use and they can be applied by coffee growers, cooperatives and by regulatory agencies, helping on the coffee qualification process.