Reducción de problemas de adherencia en procesos de galvanizado mediante técnicas de minería de datos

This paper presents an example of the application of data mining techniques to obtain hidden knowledge from the historical data of a hot dip galvanizing process and to establish rules to improve quality in the final product and to reduce errors in the process. For this purpose, the tuning records of a hot dip galvanizing line where coils with adherence problems in the zinc coating had been identified were used as a starting point. From the database of the process, the classical data mining approach was applied to obtain and analyze a number of decision trees that classified two types of coils, i.e. those with the right adherence and those with irregular adherence. The variables and values that might have influenced the quality of the coating were extracted from these trees. Several rules that may be applied to reduce the number of faulty coils with adherence problems were also established.