Nonlinear Decoupling Control With ANFIS-Based Unmodeled Dynamics Compensation for a Class of Complex Industrial...

Area of Science:

• Control Engineering
• Artificial Intelligence
• Process Systems Engineering

Background:

• Complex industrial processes often feature multivariable, strongly coupled, and nonlinear characteristics, hindering accurate modeling.
• Conventional and data-driven control methods face challenges due to the inherent difficulties in modeling these complex systems.
• Existing decoupling control algorithms often rely on complicated switching mechanisms, increasing implementation complexity.

Purpose of the Study:

• To propose a novel multivariable decoupling control algorithm for complex industrial processes.
• To effectively compensate for unmodeled dynamics (UD) using an adaptive neural-fuzzy inference system (ANFIS).
• To simplify the control algorithm and its realization by eliminating the need for complex switching mechanisms.

Main Methods:

• Development of a nonlinear multivariable decoupling controller incorporating UD compensation.
• Utilizing a decomposition estimation algorithm with ANFIS for estimating unmodeled dynamics.
• Analysis of closed-loop system stability and convergence using novel lemmas and theorems to ensure uniform boundedness of variables.

Main Results:

• Successful estimation of unmodeled dynamics and achievement of desired control effects through ANFIS-based compensation.
• Significant simplification of the decoupling algorithm and its practical implementation by avoiding complex switching mechanisms.
• Demonstration of uniform boundedness for all system variables, confirming stability and convergence.

Conclusions:

• The proposed ANFIS-based multivariable decoupling control algorithm is effective and practical for complex industrial processes.
• The method offers a simplified approach to control system design by adeptly handling unmodeled dynamics.
• Experimental validation on a twin-tank system confirms the algorithm's performance and robustness.