Multiple-page-mapping backpropagation neural network for constant tension control

Abstract

Constant tension control is widely required in industrial applications. Because of random tension interference, motor running speed vibration usually occurs. Tension control is a way to minimise the effect of random tension interference, however, the tension modification reference is another interference to the speed control system. The rewinding roll drive of a metal-film coating machine is a system with multiple closed loops and multiple input and output variables. Desired speed and tension responses are difficult to achieve by implementing conventional analogue proportional-plus-integral (PI) control. The paper introduces an artificial neural network algorithm that can successfully isolate cross coupling between the speed and tension control loops, and both loops can operate quasi-independently. It overcomes the disadvantages of traditional PI control systems. To handle the variation of the rewinding roll diameter, multiple pages of the network are applied. This technique can treat a dynamic nonlinear system as a quasistatic linear control system. Therefore this method decouples the speed and tension-control paths, and both control paths can independently operate in a quasistatic state. Simulation results show the effectiveness of this control algorithm.