An artificial neural network approach for non-invasive estimation of essential oil content and composition through considering drying processing factors: A case study in Mentha aquatica

Abstract

The potential of artificial neural networks (ANNs) for assessing key herb quality aspects (color features, moisture ratio, essential oil content and major constituents) by considering two drying processing factors (temperature and time) was evaluated. Water mint was employed as model species owing to global popularity and economic importance. Variation in herb quality was induced by employing different periods (0–450 min) and temperatures (50–70 °C) of drying processing. Samples were occasionally imaged (400–700 nm), and then quality features were invasively determined. Green color attenuation was triggered by drying processing duration and temperature. Essential oil content peaked at a shorter time period and at a lower value as drying processing temperature increased. In all drying processing temperature levels, three main components (linalool, linalyl acetate, 1,8-cineole) accounted for more than 77 % of the total essential oil, while five components (α-terpineol, geraniol acetate, hedycaryol, myrcene, neryl acetate) for approximately 15 %. These eight constituents were thus considered in the model. An ANN model was obtained, including an input layer (drying processing temperature and time), 10 hidden layers, and an output layer (12 quality features). This structure corresponds to a 2−10-12 topology. The presented methodology yielded precise estimations of the quality features under study (correlation coefficients in the range of 0.79 and 0.99). Therefore, the technique under study was proven to be very promising for non-invasive in situ estimations of several critical herb quality features.