Self-Tuning Control Using an Online-Trained Neural Network to Position a Linear Actuator.

Area of Science:

• Robotics and Control Systems
• Mechatronics Engineering

Background:

• Linear actuators are crucial in industrial applications, converting rotational to linear motion for tasks like lifting and positioning.
• Traditional control systems struggle with the inherent dynamic changes and disturbances affecting linear actuators.
• Existing methods often require precise system models, limiting their adaptability.

Purpose of the Study:

• To develop and evaluate a robust control strategy for electric linear actuators.
• To address real-time positioning challenges, sudden load variations, and parameter uncertainties.
• To improve actuator performance without relying on a detailed system model.

Main Methods:

• Implementation of a general-purpose controller with self-tuning gains.
• Integration of a neural network with Proportional-Integral-Derivative (PID) control for enhanced robustness.
• Utilizing an online training neural network for adaptive control, eliminating the need for a pre-defined engine model.

Main Results:

• Achieved a 42% reduction in Root Mean Square Error (RMSE) for trajectory tracking.
• Demonstrated a 25% saving in energy consumption.
• Validated performance through both simulation and real-world tests, confirming robustness.

Conclusions:

• The proposed neural network-PID control offers a robust and adaptive solution for electric linear actuators.
• This approach effectively manages uncertainties and disturbances, outperforming traditional methods.
• Significant improvements in accuracy and energy efficiency were confirmed in practical applications.