Premium
Monitoring coffee roasting cracks and predicting with in situ near‐infrared spectroscopy
Author(s) -
Yergenson Nathan,
Aston David Eric
Publication year - 2020
Publication title -
journal of food process engineering
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.507
H-Index - 45
eISSN - 1745-4530
pISSN - 0145-8876
DOI - 10.1111/jfpe.13305
Subject(s) - roasting , partial least squares regression , degree (music) , cracking , near infrared spectroscopy , event (particle physics) , in situ , spectroscopy , biological system , environmental science , computer science , process engineering , materials science , chemistry , optics , metallurgy , engineering , acoustics , machine learning , composite material , physics , organic chemistry , quantum mechanics , biology
The prediction of start and end times of the first and second crack events in roasting coffee is feasible with in situ near‐infrared (NIR) spectroscopy. Roasting samples analyzed herein consist of eight varieties of arabica coffee with different origins and processing methods, roasted with four temperature–time profiles. Real‐time analysis of recorded audio from coffee bean popping sounds provides a basis for determining the start and end times of each major event. A custom in situ diffuse reflectance probe improved NIR output, and partial least squares (PLS) regression generated a separate model for each crack event. The resulting PLS models show strong potential for process control implementation. A newly developed roast degree scale based on the progression through crack events is arguably more meaningful than common color cues to connect and correlate the complex chemistries and consumer qualities in roasted coffee. Practical Applications The first and second cracks are two sets of popping sounds that occur during coffee roasting. These events are commonly used as indicators of roast degree for determining the endpoint of the roast, but this judgment can be subjective. This work explores the use of NIR spectroscopy to predict the start and end of the two cracking events to provide a more robust method for controlling the roast based on the cracks. The predictions are used to generate a numerical roast degree scale, which provides a consistent method of comparing the roast degree of coffees from different origins or roasters.