Artificial Neural Networks: accurate and convenient tools to define and control process conditions during food freezing and thawing

GOÑI S.; CAMPAÑONE, L.A.; SALVADORI V.O.; MASCHERONI R.H

Congreso; 11th World Congress of Chemical Engineering; 2023

The modeling and control of food freezing processes is complex due to the nonlinearity of the heat transfer mechanisminvolved, the variable thermal properties that depend on temperature, composition and structure of foods, and the phasechange that depends on food composition. The actual state of the art has allowed the development of precise numerical,approximate/simplified and neural network prediction methods of food freezing and thawing times, having equivalentaccuracy in their results [1]. While numerical methods are not practical for rapid definition of process conditions ormonitoring, and approximate methods require complex shape factors for irregular shapes, Artificial Neural Networks –developed and validated based on a large database of food freezing and thawing times – have proven to be accurateand efficient tools for both the design and control of freezing and thawing processes. They are equally accurate for foodsof different size, shape, composition, and structure, under a wide range of operating conditions [2].This work describes in detail the characteristics and possibilities of use of Artificial Neural Networks in design and controlof food freezing and thawing processes. Three typical processes of food freezing and thawing were simulated fordifferent food types, shapes, composition and sizes:1- Freezing of meat burgers in continuous belt freezers: the study involves burgers of 6, 8, 10, 12.5 and 15 mm ofthickness, frozen in different freezers (different values of heat transfer coefficient) with air at -25, -30, -35 and -40°C.Not only freezing time can be predicted, but any change in food thickness, air temperature or speed can be immediatelymanaged.2- Freezing of spherical products in fluidized bed continuous freezers: peas of different diameters (6, 8 and 10 mm),blueberries (8 and 10 mm), small meat or melon balls (10, 12.5 and 15 mm). In all cases, different air temperatures andheat transfer coefficients were considered.3- Thawing of meat or fish blocks (in whole pieces or minced) of different thicknesses at diverse ambient temperaturesand with air or water, within a usual temperature working range (5 to 25°C).In all cases ANNs predicted the influence of food characteristics (type, composition, shape and size) and processconditions (temperature, heat transfer coefficient) on process time. This allows to determine – for each food – workingconditions that comply with certain quality specifications (like maximum freezing time or minimum thawing time) orspecified production rates.