Deep drill-down analysis for failures detection in the production line

This paper is devoted to the identification of parameters which influence the emergence of failures in the production process in industrial production. The selected assembly line is semi-automatic, which means that an entrusted operator is required to operate some workplaces and other operations work automatically. During running the processes of the assembly line, various types of failures occur. These failures can be divided into six groups according to the severity of the impact of their occurrence on the operation of the assembly line. In the data files, which we work with, files are containing records of the occurrence of these failures. There are types of failure records that require immediate line stops up to type of records indicating only alarms and warnings. Therefore, it is convenient to be able to predict in advance the emergence of the failures in the running process. In the production process, there are also failures, which occurs more often and therefore need more attention. The results of our article were compiled on the basis of reference data provided from the selected assembly line, which represented the assembly of the rear axle wheel unit line. The work is focused on the analysis of real data from the production process, which, in contrast to the data obtained through the simulation model, capture the real state of production. From the obtained results it is possible to predict the emergence of accuracy of outstanding failures. Besides that, it is able to predict the accuracy of each concreate failure. As an illustration example, we choose one failure, which occurs most often and predicted its emerging in the specific day time. On the data analysis and failure detection, we used several methods, but in this article, we would like to devote to the deep drill-down analysis and present the results from this concreate method.