Model Identification of a Paper Machine Cross-Directional Process under Model Predictive Control

The paper machine cross-directional (CD) process is an industrial spatially distributed system. Paper properties are controlled by a set of different actuator arrays acting in the cross direction (CD) as the paper sheet moves along the machine direction (MD). The industrial custom is to identify CD models from bump tests that are run in open-loop. This article presents a technique for CD response shape, alignment and dynamics model identification. The spatial response is identified by a no causal spatial FIR model that accounts for the actuator response in the cross-direction (CD). A novel technique for CD alignment detection is presented through disconsolation of the measurement profile. The spatial model identification is followed by estimating the CD process dynamics. The CD model identification technique is extended to feedback loops running under model predictive control (MPC). CD process models are identified in closed-loop from short identification experiments in a low signal-to-noise ratio (SNR). The proposed technique is validated by conducting identification experiments on an industrial paper machine model running under MPC.