A neuro-fuzzy computational approach for multicriteriaoptimisation of the quality of espresso coffee by pod based on theextraction time, temperature and blend

We demonstrate how soft computing methods can be exploited to solve multicriteria quality optimisation problems in food science and technology. In particular, we link neuro-fuzzy modelling techniques with simulated annealing to optimise/design the quality of espresso coffee by pod. The design variables are the extraction time (ranging from 10 to 30 s), temperature (80110 degrees C) and blends (100% Arabica, 100% Robusta and Arabica Robusta: A20R80, A80R20 and A40R60); they are not the only variables that affect the sensory profile of a cup of espresso coffee, but have a strong impact on the sensory quality of the beverage. Based on the framework, we show that the particular problem is a nonlinear one. Hence, an espresso coffee characterised by a specific sensory profile can be extracted using different sets of parameter values. For example, the same sensory profile can be obtained using either pure Robusta extracted at 22 s and 94 degrees C or 90% Arabica and 10% Robusta extracted at 25 s and 99 degrees C. Yet, the global optimum with respect to the distance to the optimum sensorial values is obtained using 70% Arabica and 30% Robusta extracted at 15 s around 93 degrees C