Online discrete-time LQR controller design with integral action for bulk Bucket Wheel Reclaimer operational processes...

Area of Science:

• Robotics and Automation
• Control Systems Engineering
• Artificial Intelligence

Background:

• Bulk solids resumption using bucket wheel reclaimers (BWRs) faces challenges due to stack geometry and material variations.
• Precise control is difficult due to process imprecision and environmental uncertainties.
• Existing methods struggle with the dynamic and unpredictable nature of bulk material handling.

Purpose of the Study:

• To present a novel online optimal control system design for BWR bulk resumption.
• To implement a reinforcement learning approach for real-time control.
• To address the challenges of imprecision and uncertainty in bulk material handling.

Main Methods:

• Utilized Action Dependent Heuristic Dynamic Programming (ADHDP), a reinforcement learning paradigm.
• Developed an online learning system for the Discrete Linear Quadratic Regulator (DLQR) optimal control solution with integral action.
• Designed a control system independent of the plant model, with self-adjustable controller gains.

Main Results:

• The ADHDP-based DLQR controller effectively learns optimal control solutions in real-time.
• The control system demonstrated adaptability to varying process conditions and uncertainties.
• Achieved self-adjustable controller gains and model-independent decision rules for BWR operation.

Conclusions:

• The proposed ADHDP approach offers a robust and adaptive solution for BWR bulk resumption control.
• This method enhances process efficiency and stability despite inherent system uncertainties.
• The real-time, model-independent nature of the controller provides significant advantages for bulk material handling.