The authors are looking at the behaviour of the mineral processing circuits by means of the simulators operating under random disturbance. The present paper deals with the fuzzy control systems in comparison with PID controllers. A series of tests are conducted on simplified prototype processes with (1) dead time plus first-order lag model and (2) diffusive model. In the next series of tests the self-sustained process model was introduced to improve control against dead time. In both cases fuzzy and PID controllers are optimised by EVOP method. Particular attention is paid to the effect of observation noise to analyse for the problems which can be fatal to practical applications in mineral processing. Finally the methods developed are applied to a closed-circuit wet grinding system simulator operating under random disturbance.