Experiment-based comparison of nature-inspired algorithms for optimal tuning of PI-fuzzy controlled nonlinear DC servo systems

摘要

This paper proposes the comparison of seven nature-inspired optimization algorithms (NIOAs) applied to the tuning of proportional-integral (PI)-fuzzy controllers for a class of nonlinear direct current (DC) servo systems. The servo systems are modeled by second order dynamics with a saturation and dead zone static nonlinearity specific to the actuator. Seven NIOAs are considered, namely Simulated Annealing, Particle Swarm Optimization (PSO), Gravitational Search Algorithm (GSA), hybrid PSOGSA, Charged System Search (CSS), adaptive GSA and adaptive CSS, to solve optimization problem. The objective function is the weighted sum of time multiplied by squared control error plus squared output sensitivity function. The output sensitivity function is obtained from the state sensitivity models of the fuzzy control systems with respect to the modification of the process gain leading to a reduced process gain sensitivity. Three parameters of the PI-fuzzy controllers are tuned as variables of the objective function. Simulations and experimental results related to the angular position control of a laboratory nonlinear DC servo system are included.