Browsing by Subject "laadunvarmistus"

The literary part of the thesis dealt with coffee composition and the factors that influence it. It was found that coffee variety, climate, post-harvest processing and roasting affected the composition. Coffee quality assurance concentrates on checking the moisture and determining the sensory quality of coffee beans. Sensory quality is determined by trained professionals and is the slowest unit process in coffee processing and should be speeded up. NIRS (Near Infrared Spectroscopy) is a non-direct method that uses the near infrared region of the electromagnetic spectrum (800-2500 nm). In the last few years it has been used to analysing the composition of coffee, screening for defects, determining roasting degree and modelling sensory characteristics of coffee. The main objective of the experimental part was to develop efficient prediction models for green and roasted coffee moisture. Analysing other chemical (green coffee protein, sucrose and caffeine) and physical (roasted coffee colour) parameters were tested as well. The roastery aims at using the technique in coffee quality control but also possibly in production management in the future. Green and roasted coffee were analysed by both NIR and reference methods. Moisture was measured in the roastery laboratory with an air-oven, crude protein was determined by Dumas method in the University of Helsinki. Sucrose and caffeine were measured in an external laboratory (HPLC method). Coffee colour (L*) was measured by the roastery quality controllers as a part of their daily routines. Reliable prediction models were achieved for both green and roasted coffee moisture, and coffee colour (R2 0,98–0,99, SEP(C) 0,06–0,18 (0,95–1,45 %)). These models worked well although there were variations in the samples’ origins, post-harvest processing and roasting degrees. Also, the model developed for green coffee crude protein seemed promising. On the other hand, green coffee sucrose and caffeine contents proved to be hard to measure and these methods need further development in the future.