Modelling of heat and mass transfer phenomena and quality changes during continuous biscuit baking using both deductive and inductive (neural network) modelling principles

The aim of the present study is to specify the chosen methodology to obtain a mathematical model that simulates the functioning of a continuous industrial-scale biscuit oven. This model includes the prediction of typical baking curves (kinetics of product temperature gain and water loss, kinetics of radiant, convective and contact flux variations affecting the product over the entire length of the oven...) but also integrates the prediction of two essential end-product characteristics: dimensions, and surface colour. This combines a classical heat and mass transfer model (deductive modelling approach) developed in a previous study and two artificial neural network models (inductive modelling approach) in order to model the kinetics of biscuit colour and thickness variations during the baking. As the underlying physico-chemical principles governing the system to be modelled (biscuit surface colour and size changes during baking) are not clearly elucidated and understood, the ANN modelling approach is of considerable interest since a functional form for the relationship between the process data and product quality data can be estimated without a priori putting forward hypotheses on the underlying physico-chemical mechanisms involved.