Using neural network to control cup-shaped pelletizer

Abstract

Cup-shaped pelletizer is complex machine in terms of automatic control. A large number of incoming (humidity mixture, cup rotation speed, the angle of cup, mixture and water discharge into the cup), the outgoing (diameter, strength and moisture of pellet) parameters and external influences (air humidity and temperature, the content of iron in the mixture) dependent on each other non-linear and in some cases not explicitly, that creates great difficulties in controlling cup-shaped pelletizer. The definition of a suitable system of automatic control cup-shaped pelletizer series of papers, including this one. In this paper, justified the choice of direction to find a suitable control system. To do this, a mathematical model is made up of single-drive control pan rotation cup-shaped pelletizer. In the model for comparison of automatic control systems use the classic PID controller and a single-layer neural network trained by the algorithm “delta rule”. Comparison of control systems for different incoming influences. The main parameters of transient processes (the first matching tу, overshoot σ, time of transient process tпп) and found error rates positioning с0, с1 speed and acceleration of с2 for each type of control system. Showing laws, which are found appropriate error rates. For the visibility of the different control systems are graphs errors с0, с1 and с2.