Tracking performance optimization of balancing machine turntable servo system based on deep deterministic policy gradient fractional order proportional integral derivative control strategy

Abstract

In automotive manufacturing, brake disc balance accuracy is critical for braking system reliability. The tracking accuracy of the balancing machine’s turntable servo system directly influences production efficiency and disc balance. To enhance turntable servo control in position and velocity tracking, this paper proposes a fractional order proportional integral derivative (FOPID) controller using a deep deterministic policy gradient (DDPG) algorithm inspired by deep reinforcement learning (DRL). A dynamic model of the servo system is developed to support the design of the DDPG FOPID control strategy. Anti-interference and anti-noise experiments are conducted to compare control strategies including fuzzy logic (Fuzzy), genetic algorithm (GA) PID, particle swarm optimization (PSO) PID, Q-learning PID, DDPG PID and DDPG FOPID through the physical experimental platform of the turntable servo system. Experimental results demonstrate that the DDPG FOPID strategy offers superior robustness and tracking performance, suggesting its potential to advance intelligent control methods in automotive manufacturing.