A Strategy for the Nonlinear Control of Affine Systems using Multiple Neural Networks

This paper presents a control scheme whereby feedforward decoupling of a control affine system is undertaken from the utilisation of plant input/output data. The development of the control scheme comprises two stages. Firstly, a nonlinear dynamic model is identified. In this study, a multiple input multiple output (MIMO) dynamic model is identified using a multiple neural network approach. The second stage involves the utilisation of the neural network based model within the differential geometry based nonlinear control method of input/output linearisation. A simulated case study of the dual composition control of a distillation column is used to illustrate the method. A preliminary robustness evaluation illustrates control performance with respect to the model/plant mismatch.