Machine Learning-Based Identification of Root Causes for Defective Units in Manufacturing Processes

Abstract

AbstractTo achieve a high overall equipment effectiveness in a manufacturing process, reducing the number of defective units is crucial. It is therefore vital to identify the root causes of defects to be able to rectify them. However, the analysis of defective units can be a time-consuming and costly task.By using machine learning, we can leverage data of the manufacturing process, like process states and different measurement values, to identify the root causes for the defects. We propose to use this data as features for the classification of a unit as defective or as belonging to a specific defect class. We can then identify the root causes for the defects by calculating the importance of the features.In this paper, we compare our feature-based approach with deep learning methods based on the attention mechanism. The evaluation of our approach on data of a complex production process shows, that our approach clearly outperforms the deep learning methods. It also revealed the challenges in the collection of meaningful data