Arquitectura ANFIS del Proceso de Fermentación de Aceitunas Negras Naturales

entrenamiento a escala piloto y validación a escala industrial.

Abstract

In the fermentation of natural black olives, process conditions affect the quality of the product and they differ at different scales. The aim of this work is create mathematical models to characterize the properties of natural black olives obtained at large-scale process, from pilot scale one. ANFIS architectures with physicochemical parameters as inputs, and color and texture as outputs were created. Such structures were trained with data obtained at pilot scale and validated with those from industry. Networks showed better performance have two membership functions for input incorporated in the form of generalized bell, and a linear membership function as output. The structures has demonstrated ability to predict the evolution of the process at industrial scale in the first period of fermentation. To increase their ability, it would be useful to try other models, such as artificial neural networks.