Online topography measurement of uncoated/coated deterministic or random surfaces : final report

More stringent customer requirements and the need to increase productivity are two of the numerous reasons that push steelmakers to deliver higher surface quality strips. Even with the recent process improvements, the strip surface is not yet fully under control. To support a better control of the process, an online topography sensor has been developed based on a completely new approach. The neural network installed in the process computer of the galvanising line at Aviles now predicts the level of force needed to comply with some given elongation, as well as the top and bottom roughness, with a precision of around 20 %. The use of SRM and of the neural network will be pursued to improve the process. Different process models have been tuned at AM Gent allowing, among other benefits, the online calculation giving the skinpass operator a real advantage compared to manual offline stylus roughness measurements on sheet samples. The models can be used to predict roughness values for the coils coming next, leading to a decrease in workroll changes with a substantial financial advantage. Concerning the 3D parameters, due technical deadlock, no efficient optical solution has been found for multiple line projections. To solve this, work has been done on one line profile and on background image leading, through the use of a model, to the calculation of new parameters. The new proposed statistical parameters allow a better classification of the surface structure of strip material regarding its formability and coating properties.