Tension control optimization of industrial large-scale roll-to-roll system under parametric uncertainties
Roll-to-roll systems are very common in industry for metal, paper, textile and polymer material treatments. The key variables to monitor and control are the web speed and web tension in each span. The objective is to reach the expected web speed while maintaining web tension in an acceptable range around the tension reference. Cascading control is used on each motor; a first loop permit to control the motor torque, a second loop is used for speed control. Finally, the third control loop is used for web tension control. Torque and speed control can be easily hand-calculated using identification approaches. However, web tension control is studied for several years. The tension controllers are usually PI controllers in industry. The H∞ approach is very common to synthesize web tension controller. This kind of approach has the drawback to produce high order controllers. The synthesis of PI controller using H∞ approach results to the resolution of a fixed order and fixed structure H∞ synthesis problem. Recent advance in optimization algorithm permits to resolve the fixed order and fixed structure synthesis problem . This synthesis method gives good results in term of performances but the obtained controllers are highly sensitive regarding parameters variations. In this work, multi-scenarios stochastic programming is used to ensure robustness regarding parameters variations. The key idea is to generate a set of uncertain parameters and to evaluate the H∞ norm for each value. Nevertheless, the H∞ norm has not a physical representation, the influence of a variation of the norm on the outputs performances cannot be known directly. Several designs are then simulated in time domain, the time domain simulation has an important computational cost, and a performance criterion is computed to have a physical representation of the system representation. The design to implement is then chosen considering the frequency domain and time domain criteria.
Scheme of the S/KS/T synthesis with reference model M0