Electrolytically galvanized steel strapping is considered a premium product of Voestalpine, which holds a leading global position for the most challenging types of thin sheet application. Galvanized steel panel is used due to its particular material property, e.g. in the field of car body production. During the production process, the anode condition is critical for the galvanization quality. Steel strip is galvanized by 24 pairs of anodes, which are subject to regular visual and manual inspection and in case of doubt are replaced at an early stage. Defective anodes are the main cause for quality losses during galvanization.
Reliable prediction optimizes maintenance
Aim was to determine the perfect moment for anodes replacement in order to optimize maintenance timing without risking quality degradation. Initial validation showed machine learning may support this decision making process.
An interdisciplinary expert group of ten members developed a hybrid model which is embedded directly within the process units’ control systems. To achieve this, the relevant influencing factors had to be deduced out of 10 million of performance data and 400 variables based on measurements conducted throughout a period of three years.
Based on an existing physical model and data analyses, cell tension was identified as the crucial variable for galvanization quality. Although sound results had already been achieved by machine learning (ML), the ML model was far too complex to be implemented within the existing online system. By merging the ML model with the improved physical model it was possible to develop a reliable prediction model that allowed easy interpretation by process experts.
Better quality and more cost efficiency
The hybrid model has been performing successfully at the existing plant since the end of 2019. Defects within anodes are detected directly during operation. Process experts will utilize the hybrid model for better quality comparison of individual anodes. The involvement of all relevant employees early on resulted in high user acceptance.
The project was funded within the framework of COMET – Competence Centers for Excellent Technologies by BMK, BMDW, Land Steiermark. The COMET program is processed by the FFG.