The large amount of interest in using neural network representations for the modelling, prediction and control of plant processes has resulted in the development of a large number of techniques.
Mechanistic models are costly and time consuming to produce and are still subject to approximations and simplifications of the manufacturing process.
Although linear techniques often provide reasonable process descriptions, in increasingly competitive markets where plants are being pushed to their production limits, recent trends have been to move to more complex non-linear empirical and hybrid modelling approaches.