INTERNAL MODEL CONTROL PERFORMANCE EVALUATION IN NONLINEAR COUPLED TANK SYSTEM

Abstract

Most industrial process control systems are associated with non-linearity behaviour.
Control of such industrial process system is a challenging task according to the degree of
non-linearity of the system. A typical process control that represent part of such complex
industrial process is the principle used in Coupled Tank System (CTS) level control. This
paper investigated the performance of Neural Network Internal Model Control (NNIMC)
and Two Degree of Freedom Internal Model Control (2DOF-IMC) with application to level
control in CTS. For the NNIMC, the Neural Network (NN) forward and inverse models of
the plant were identified from process input and output sample data generated from
laboratory experimental setup. The network architecture for both models is nonlinear
autoregressive network with exogenous inputs (NARX) while the training algorithm selected
was Levenberg Marquad (LM). The models developed are then used to design NNIMC.
Additionally, the 2DOF-IMC was designed with advanced tracking and disturbance
rejection filter. Moreover, conventional PI controller was optimized by integrating neural
network feedforward control strategy to serve as advanced feedforward PI control strategy.
Performance analysis of the investigated controls algorithms were evaluated. Simulation
results indicated that both NNIMC and 2DOF-IMC controllers have similar performance
and as well outperforms their traditional control techniques counterpart hence could be used
to control levels in a CTS effectively.