Computação evolucionária em identificação e controle de processos: fundamentos, análise e aplicações

Abstract

In this paper several evolutionary computation paradigms in process identification and control are utilized. The following methodologies are addressed: i) genetic algorithms (with floating point representation), ii) hybrid algorithms composed by genetic algorithms with simulated annealing, and iii) evolution strategies (without and with self-adaptation mechanisms). Experiments in identification were conducted in mono-tank level and temperature processes. Experimental tests in control are conducted in order to find the design parameter of the PID control when controlling a non-linear level process, composed of coupled twin-tanks, which are submitted to reference change and load disturbance, and steering a trailer truck while backing up to a loading dock.