Nowadays the increasing complexity of most processes increases the demand for performant models. Most of these processes are highly non-linear and dynamic ones, which require complex modelling techniques. Neural networks are eligible modelling candidates for such processes, since they have the ability to map a variety of input-output patterns quite easily. Moreover certain types of networks (the so-called spatio-temporal networks) can not only model spatial but also temporal patterns. Nevertheless a continuous search for improvement is mandatory. Therefore in this paper combinations of spatio-temporal neural network types with other modelling techniques are discussed whilst applied to a complex problem from the chemical process industry, i.e. a polymerisation reactor.