Adaptable neural network model for cross directional control

A two-stage neural network model has been developed for cross-direction control of the papermaking process. The Neuro-CD model has been tested on a simulator with promising results. The basis function is a parametric single-output function called the CD-Neuron. A CD-Neuron was built to represent the quality response of each actuator. The output is the actual quality value that results from a change in the actuator setting.