USE OF HYBRID MODELS FOR PREDICTING AND CONTROLLING FOOD PROPERTIES
In the modern food industry there is a growing need for models that can predict complicated phenomena such as taste development or the effect of complex food processing events on product properties. Such models that can predict product properties with respect to changes in the processing conditions can directly be used for process optimisation or control. Very often problems are encountered which involve sensory panel data or which require detailed information about the hydrodynamic flow to be able to predict product properties. To be able to model these kinds of problems several modelling techniques can be combined in a hybrid model. White box models based on physico-chemical relations can be combined with computational fluid dynamics (CFD) and black box approaches (e.g. fuzzy logic, neural networks). Hybrid modelling makes it possible to predict how changes in the processing conditions affect sensory evaluations or product properties which are the result of complex food processing events. Moreover, hybrid modelling allows for an optimal use of the available data and knowledge. In this paper we briefly discuss two examples of hybrid modelling: taste development in cheese during ripening with respect to changes in the processing conditions and secondly, the modelling of agglomeration in spray dryers and the effect on the product properties. By using white box sub-models in the hybrid global model the predictive capacities of the model increase.
Verschueren, M., Verdurmen, R. and de Jong, P. (2001). USE OF HYBRID MODELS FOR PREDICTING AND CONTROLLING FOOD PROPERTIES. Acta Hortic. 566, 129-134
spray drying, cheese ripening, food properties