Neural network comparison for paint errors classification for automotive industry in compliance with industry 4.0 concept

Abstract

The proposed paper focuses on utilization of neural networks for paint errors classification in the area of automotive industry. The paper utilizes hypothesis, that outdoor weather has significant impact on the number of paint errors, as a basis for comparison of neural network algorithms. For the neural network algorithms comparison we used real production data from the paint shop process. The paper deals also with definition of classification classes and attributes selection, as well as the data integration process itself that utilizes Hadoop platform as an intermediate data storage.