Quality Assurance at Continuous Hot-Dip Galvanizing Lines by Neuro-Model Assisted Fuzzy-Control

Abstract

The importance of improving product quality at continuous hot-dip galvanizing lines with air knives steadily grows. So the developed solutions have to be intelligent, adaptive and modular. This paper describes the revision of a conventional non-adaptive control strategy towards a modern solution using methods of computational intelligence. The already existing feedforward control is complemented by a neural process model and a neuro-fuzzy controller replaces the previously used conventional process controller. Both components are embedded carefully into the control environment so that consumption of time and material for the installation period can be held low. The neural process model is optional and is used for model-based control so that the process inherent measurement dead-time is avoided. The new control arrangement is adaptive, saves zinc, guarantees a more constant coating and relieves the operators.