Abstract
There is a need in the roasting industry for better real time feedback during roasting to consistently achieve the desired qualities in roasted coffee, and NIR spectroscopy has shown some promise for this purpose.
This project generated a new roast degree scale based on in situ NIR prediction of the cracking sounds which are commonly used as a real time indication of roast degree. This required measuring the start and end times of the two crack events by analyzing recorded audio from a large set of roasts including eight varieties of coffee with four roast temperature profiles each. An NIR spectrometer with a custom diffuse reflectance probe attached to the roaster recorded in situ spectra during each roast for predicting the crack event times and determining the roast degree.
Measurements of coffee acidity provided a basis for comparing the newly developed crack scale to two other scales: a simple scale generated based on NIR absorbance and the common CIELAB scale based on visible color. Both NIR based scales performed better than the visible scale in this comparison, and the simple NIR scale performed best. This scale was then used as a basis for determining the effects of different roast temperatures on acidity, to explore the ability to control acidity independent of roast degree. A new set of constant temperature roasts provided the dataset for this analysis, resulting in simple linear relationships between roast degree, roast time, and acidity.