A Novel Clustering Anomaly Detection of PCA Based Time Series Features with CNC Machines Data

Abstract

With the 4th industrial revolution, the manufacturing technology of computer numerical control (CNC) has been one of the irreplaceable important technologies in the manufacturing industry. If the factory’s machinery stops working for some reason, such as overheating or wear and tear, the damage to the factory is very high. Therefore, in order to prevent this in advance, research on a method for detecting anomalies using data collected from a machine is currently being actively conducted. This paper presents a method for detecting anomalies using only the data received from the machine without using additionally installed sensors and visually expressing them. Anomaly detection was successfully performed on both the actually collected dataset and the open dataset for validation using the clustering method after principal component analysis using the tendency. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.