| About this Abstract |
| Meeting |
MS&T22: Materials Science & Technology
|
| Symposium
|
Advances in Zinc-coated Sheet Steel Processing and Properties
|
| Presentation Title |
An Exploration of a Neural Network Approach for the On-line Prediction of Steel Strip Radiative Properties |
| Author(s) |
Nishant S. Narayanan, Fatima K. Suleiman, Kyle J. Daun |
| On-Site Speaker (Planned) |
Nishant S. Narayanan |
| Abstract Scope |
Spectral emissivity variations across Advanced High Strength Steel (AHSS) coils may cause non-uniform temperature evolutions and pyrometric temperature errors on continuous galvanising lines (CGL), producing non-homogeneous mechanical properties. This study explores spectral emissivity variations across AHSS coils in their pre-annealed state and finds these variations to be strongly influenced by the presence of surface cavities. The radiative properties are theoretically modelled with the Geometric Optics Approximation (GOA) ray-tracing algorithm, using surface height profiles obtained through 3D depth mapping of optical images. The GOA algorithm provides accurate spectral emissivity predictions within its validity regime, encompassing the typical pyrometric wavelengths; however, it excludes wavelengths important for predicting strip temperature evolutions. For improved spectral emissivity predictions, the paper explores the use of a data-driven regression neural net for correlating the number of cavities and global roughness with spectral emissivity, to possibly serve as a useful tool for on-line spectral emissivity predictions. |