Research on self-adaptive height adjustment control of shearer based on deep deterministic policy gradient

Area of Science:

• Mining Engineering
• Artificial Intelligence
• Control Systems

Background:

• Intelligent control of shearers is crucial for unmanned mining and equipment reliability in mechanized mining faces.
• Traditional optimization and deep reinforcement learning algorithms face limitations in rapid, accurate self-adaptive control for shearer drum height.

Purpose of the Study:

• To propose a novel shearer drum height control strategy using the Deep Deterministic Policy Gradient (DDPG) algorithm.
• To enhance the self-adaptive control capabilities of shearers for improved efficiency and reliability in mining operations.

Main Methods:

• Developed a DDPG-based self-adaptive hydraulic height adjustment model.
• Employed a hybrid SVD-CWT and AlexNet transfer learning method for coal and rock cutting state identification (95.06% accuracy).
• Validated the model through co-simulation (Matlab/Simulink, AMESim) and a physical test platform.

Main Results:

• The DDPG strategy significantly outperformed conventional and fuzzy PID controls.
• Achieved a response time of 0.091s and a steady-state error of 0.00052mm.
• Demonstrated superior response, stability, and anti-interference compared to TD3 and SAC algorithms, with a mean maximum simulation-experiment error of 3.14%.

Conclusions:

• The proposed DDPG-based control strategy is feasible and robust for intelligent, adaptive height control of shearers.
• Offers a reliable approach for complex coal seam conditions, advancing unmanned mining technology.
• Significantly improves shearer control performance metrics like response time and steady-state error.