In manufacturing, maintaining process stability and reducing machine downtime are critical for achieving high productivity and reliable product quality. This study aims to develop a robust anomaly detection framework integrated within condition monitoring for computer numerical control (CNC) turning centres for processes such as turning, drilling, and reaming. The work begins with sensor selection, followed by the integration of acoustic emission (AE) sensors onto the machine. These sensors capture high-frequency data that reveal subtle changes in tool conditions, workpiece interactions, and machine performance due to tool wear, material inconsistencies, or variations in machine data. The initial focus is on single-spindle machines serving as a proof of concept, with the goal of extending the approach to multi-spindle configurations. The framework effectively distinguishes between normal operations and deviations by processing the AE signals and extracting key features in the time and frequency domains. The extracted features are assessed using supervised machine learning (ML) models such as support vector machine (SVM) and decision tree (DT). Additionally, the system enables real-time identification and aids in removal of faulty components, ensuring that only high-quality parts proceed through the production line. The investigations on the single-spindle machine provide a solid foundation for algorithm development, facilitating precise adjustments to the detection framework. These algorithms are then adapted to the more complex multi-spindle scenario, which involves concurrent operations and increased signal interference, utilizing transfer learning to leverage the knowledge gained from the single-spindle setup for efficient adaptation in the multi-spindle context. By integrating acoustic emission sensor technology, condition monitoring, and artificial intelligence (AI)-driven analysis, this approach ensures a reliable solution for process monitoring, significantly improving productivity and operational reliability in manufacturing environment.
Nadkarni et al. (Fri,) studied this question.