Tension optimal control of a multivariable winding process

In this paper, a multivariable control design is developed and applied to a winding pilot plant. This process represents a subsystem of complex industrial processes, known to be multivariable, non-linear and time-variant, which make more difficult their identification and control design. The interaction between the system variables, as tensions and linear velocity, is outlined and taken into account in the multivariable dynamical model, estimated by subspace identification methods. The main goal of this study is to control the strip tensions of this process using a LQG controller associated with a weighted integral error. Experimental results show interesting control performances on the whole winding conditions even if the state-space model is only estimated on a local operating zone.