Development of Fault Diagnosis Models Based on Predicting Energy Consumption of a Machine Tool Spindle

In recent years, as the cost of energy has increased, the importance of managing energy efficiency in the manufacturing environment has also grown. In this study, energy consumption of a machine tool spindle for future period is predicted based on the past energy consumption patterns. Models for diagnosing abnormal conditions of the spindle is then developed using the predicted energy consumption. We use the energy consumption data to detect and track the spindle abnormalities and to support an environment wherein machine conditions can be reviewed. The energy consumption data of the spindle are collected, and then pre-processing is performed as follows. We proceed with an overlapping sampling of time series data that consist of timestamps and energy values to form the supervised learning-enabled data structure. The pre-processed dataset is divided into a training set and a test set in consideration of time series characteristics. Then random forest, one of the ensemble machine learning methods, is applied. The forecast performance is summarized by the mean absolute percentage error (MAPE). When the accuracy of the random forest model derived from machine learning is greater than a certain MAPE value, the detection of outliers in the predicted data is performed. Outlier detection uses the inter quartile range (IQR) method and considers data out of the defined range to be outliers. In conclusion, this paper presents data analysis models to predict the energy consumption of the machine tool spindle and to detect abnormalities by applying the predicted data.