12th International Conference on Zinc & Zinc Alloy Coated Steel Sheet - GALVATECH 2021

Measurements at high-temperature facilities such as galvanizing lines are often possible only to a limited extent. Moreover, many of the processes involved are too complex to apply existing numerical methods such as CFD to provide the desired information. Therefore, an empirical method was developed to estimate process data which is not measured. The method was demonstrated by determining the surface temperature of radiant tubes. During operation these tubes are exposed to high thermal loads, which can result in failure. For this reason, close monitoring of their state is necessary to improve operation and increase the lifetime of the tubes. Using detailed measurements from a single tube and thermodynamical principles, the most significant influences on the tube’s surface temperature were identified. Subsequently, these transport equations were approximated by introducing stochastic parameters. Lastly, the parameters were adjusted by applying a direct search algorithm. Validation of the results was performed by using an extensive control set. It has been shown that the model can produce realistic results, even when exceeding the range of data that was used when training the model. Furthermore, guided by thermodynamic principles, the model can retain a more efficient and comprehensible structure when compared to competing machine learning methods e.g. neural networks. Ultimately, the modular concept of the presented method allows for it to be transferred to other facilities or processes for which only limited information is available.