Modeling overall equipment effectiveness as a key performance indicator using deep learning

Key performance indicators (KPI) are essential monitoring measures that can indicate the performance of an organization in various sectors. To ensure progress toward organizational goals and to enhance productivity in different areas of an organization, accurate prediction of these indicators is mandatory. In operational processes, overall equipment effectiveness (OEE) is an essential monitoring indicator that can determine the effectiveness of machines. Thus, this study utilized a new set of features including planned downtime, unplanned downtime, total quantity produced, and total quantity defective for accurate OEE prediction, and for this purpose, the deep neural network (DNN) method and the long short-term memory (LSTM) method were employed. The models were trained based on k-fold cross-validation technique and optimal hyperparameters were selected for these methods. The models were evaluated and compared based on root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination (R۲) metrics. The results show that the LSTM method outperformed the DNN method for OEE prediction. The RMSE, MAE, and R۲ metrics of the LSTM method are ۱.۴۳۶, ۲.۴۷۷, and ۰.۹۵۴, respectively, showing the robustness of the model for OEE prediction.

Keywords:

Key performance indicators , Overall equipment effectiveness , Deep learning , Long short-term memory