Multiloop nonlinear IMC strategy design under PLS framework using ARX-neural network model

A novel multi-loop nonlinear Internal Model Control (IMC) strategy is developed for MIMO systems under the Partial Least Squares (PLS) framework, which automatically decomposes the MIMO process into several univariate systems in the latent space. The ARX-neural network (ARX-NN) model is incorporated in the PLS subspace and identified via optimizing two parameter sets so that the plant-model mismatch is minimized. The cascaded mode of ARX-neural network model facilitates the parameters estimation in the process identification, as well as nonlinear IMC design. With perfect modeling, the nonlinear closed-loop system can be reduced to IMC robust filter with a system pure delay factor under the proposed nonlinear IMC scheme. By tuning the IMC filter, the robust property of the control system is adjusted to account for the model-plant mismatch, and the offset free performance is obtained through perfectly inverting the steady-state gain of the model. The simulation result of pH neutralization process demonstrated the effectiveness of the proposed approach in disturbance rejection and tracking performance.