State estimation and advanced control of the 2D temperature field in an experimental oscillating annealing device

Annealing plays a crucial role in industrial steel strip production lines. Laboratory annealing devices are experimental furnaces that allow the simulation of the annealing process in large-scale production lines and are employed, e.g., to design new or improve existing heat treatment cycles. The furnace considered in this paper is equipped with individually controlled infrared heaters. It is used to reheat flat specimens of steel strips accurately and homogeneously in space according to predefined temperature trajectories. In view of the complex furnace geometry with highly specular surfaces and thermal radiation as the main heat transfer mode, the operation of this furnace constitutes a 2-dimensional nonlinear distributed-parameter thermal control problem. The basic control inputs are the electric powers of infrared heating lamps, which are controlled by six phase-fired thyristors. For temperature tracking, a two-degree-of-freedom control concept is applied, which comprises an optimal feedforward controller and a state feedback controller. The feedforward controller is based on the solution of a dynamic optimization problem. The feedback part contains a Linear-Quadratic-Gaussian controller, which requires knowledge of the actual temperature field of the specimen. Since in the considered annealing device, this temperature field cannot be completely measured, an extended Kalman filter is used for the estimation of spatial temperature profiles. This estimation is based on just three local measurements of the surface temperature of the specimen. The proposed control approach was implemented and experimentally validated in several annealing runs. The effect of an oscillating motion of the specimen on the temperature homogeneity is investigated by comparisons of measurement results with a fixed specimen position. It is shown that the temperature inhomogeneity can be significantly reduced if the specimen oscillation is systematically taken into account in the mathematical model, the state estimation, and the control design.