Virtual Unmodeled Dynamics Modeling for Nonlinear Multivariable Adaptive Control with Decoupling Design
Yajun Zhang,Tianyou Chai,Dianhui Wang,Xinkai Chen
DOI: https://doi.org/10.1109/tsmc.2016.2602826
2016-01-01
IEEE Transactions on Systems Man and Cybernetics Systems
Abstract:For a class of complex industrial processes with nonlinear, strongly coupled multivariable properties, a new multivariable decoupling design framework which based on the concepts of virtual unmodeled dynamics (VUD) and lower order linear models is proposed in this paper. First, a self-tuning multivariable decoupling controller is constructed based on a lower order model. Then based on the compensator of the VUD, a nonlinear multivariable decoupling controller is designed, where a decomposition estimation algorithm is employed for modeling the VUD. In our proposed scheme, it solves the problem that the current input signal is embedded in the VUD and the true input data vector used by the learner model is difficult to be obtained in time. The linear and nonlinear decoupling controllers are integrated by an adaptive switching control algorithm to take advantage of their complementary features. Finally, the stability and convergence of the proposed algorithm is analyzed. Experimental tests on a heavily coupled nonlinear twin-tank system are carried out to demonstrate the effectiveness and the practicability of the proposed method.