Data mining approachs for machine failures: real case study

Abstract

As in all manufacturing systems, in production of corrugated cardboard, the machine errors and stops caused by the different properties of the raw materials are observed. These errors are reflected to the manufacturers as high costs. Different methods have been used to minimize these costs. One of them is to detect the error in production with machine learning methods. Since the machine learning techniques are at a level that can be applied with different techniques in order to understand the data and use it according to its purpose, has also been used in the corrugated cardboard production data. In this study, the machine fails during manufacturing of corrugated cardboard are determined by using the machine learning methods that they depend or not on the raw material used. At the same time, machine learning methods were compared according to the results of the training and test data, and it was observed which one gave the best results for this problem.