The aim of the work was to develop the classical and hybrid neural models of fluidised bed drying process and to examine the effect of noise level imposed in the training data on the models performance. The analysis was carried out on the basis of one classical neural model (CNM) and two hybrid neural models (HNM). Fast mapping deterioration was observed for the CNM for higher learning data noise levels. However the performance of the CNM is acceptable for small and medium noise level values. Hybrid neural models show astonishing resistance to noisy learning data. An excellent mapping of the fluidised bed drying process, even for the 30% noise level, was proved for the HNM.