General rules to create a family of neural models for chemical reactors have been proposed and described. Main problems of practical use of neural networks to model chemical reactors such as: determination of the input - output variables relevant for the investigated system, choice of the learning method as well as a determination of size of the learning data have been presented and discussed.