Predicting overall equipment effectiveness measure using machine learning algorithms

As markets become more competitive and production becomes globalized, every production facility needs to closely monitor and improve its processes to remain competitive. In line with this goal, production facilities use performance evaluation systems to identify areas of focus to improve their performance and efficiency. Overall equipment effectiveness (OEE) is a widely used metric for evaluating the performance of production equipment. OEE defines production efficiency by combining availability, performance and quality. In this PhD thesis, approaches based on deep learning architectures are proposed to predict OEE using a dataset obtained from the corrugated cardboard department of a box factory operating in Turkey. The use of these approaches is intended to help business managers identify disruptions in production lines and make data-based decisions to improve OEE. Model architectures based on Long Short Term Memory (LSTM), Bi-directional Long Short Term Memory (Bi-LSTM) and Gated Recurrent Units (GRU) architectures were developed for data analysis and OEE prediction. The findings obtained with the developed model architectures show that these algorithms have the potential to predict OEE in a production environment and can be used to improve equipment performance. Keywords: Overall Equipment Effectiveness, Deep Learning, Feature Selection, Corrugated Cardboard