A nonlinear regression model-based predictive control algorithm.

Area of Science:

• Control Engineering
• Process Control
• Nonlinear Systems

Background:

• Industrial processes often exhibit complex nonlinear dynamics.
• Existing control strategies may struggle with accurate prediction and tight control of these systems.
• Accurate modeling and predictive control are crucial for efficient industrial operations.

Purpose of the Study:

• To develop and validate a novel nonlinear regression model-based predictive controller (NRPC).
• To enable online nonlinear open-loop modeling during closed-loop control execution.
• To demonstrate the controller's effectiveness for both single-input-single-output (SISO) and multi-input-multi-output (MIMO) industrial applications.

Main Methods:

• Designing a controller structure that allows nonlinear open-loop modeling during closed-loop control.
• Regenerating the system matrix at each sampling instant using a continuous function for accurate plant prediction.
• Implementing and testing the proposed algorithm via computer simulations on nonlinear plants.
• Conducting real-time experiments on a DC motor, a plastic injection molding machine, and a nonlinear MIMO thermal system.

Main Results:

• Computer simulations confirmed the approach's ease of implementation and ability to provide tight control.
• Experimental results on SISO and MIMO systems showed improved performance compared to multi-model dynamic matrix controller (MPC).
• The NRPC demonstrated superior tracking of multi-set point profiles and effective disturbance rejection.

Conclusions:

• The proposed nonlinear regression model-based predictive controller (NRPC) offers a unique and effective approach for industrial process control.
• The controller's ability to perform online nonlinear modeling enhances prediction accuracy and control performance.
• The NRPC provides a robust and improved alternative to existing control methods like multi-model MPC for complex industrial applications.