Medium density fibreboard (MDF) is a popular and cheap wood-based panel for which an accurate control of the fiber humidity is key to get the desired product quality. The drying dynamics is nonlinear and spatially dependent, with strong interactions between air humidity, temperature, flow, type of wood and chips size. This translates in several complex phenomena, difficult to model accurately by first principles. The end aim in this work is implementing a model predictive control (MPC) for an industrial MDF dryer which reaches the desired fiber humidity by the most energy-efficient control path. To this aim, a suitable grey-box model of the dryer is developed first. Then, based on it, a moving-horizon estimator is used to infer the unmeasured process inputs and model parameters, which provides the required inputs to a tracking MPC with economic considerations. An efficient implementation of the optimization problems has been done using orthogonal collocation for model discretization and modern software for large-scale problems, including automatic differentiation and a sparse interior point solver. Previously to the actual implementation on-site, the proposed controller is tested in simulation with a detailed distributed-parameter model of a generic MDF dryer.
No information added