Design Optimization Using Genetic Algorithms of Web Handling Systems: The Case of the Pendulum Dancer Mechanism

Web systems handling elastic webs are very common in industry. The plant considered in this paper is an industrial winding system with pendulum dancer mechanism. Such a studied dancer moves along an arc of circle. Its utility is double: it imposes the web tension (in steady state operation) and filters the variations of tension mechanically. It is actuated by an air jack with adjustable pressure and has a stiffness and viscous dynamic behavior. Pendulum dancers have been rarely presented and studied in publications. The non-linear model built in Matlab software environment is used as a simulator. Moreover, the state space model useful for modern controller computing can be found thanks to linearization around an operating point. In order to improve the unwinder-winder control performances, the mechanical parameters of the pendulum dancer have to be optimized. The more constant the dynamic behavior over a large frequency band will be, the better the performance of a linear time invariant controller will be obtained. This is the main objective in desensitizing the dancer. The optimization is achieved by using genetic algorithms. The obtained results are discussed and the benefits of this design optimization for industry applications are presented.