Machine learning-based CPS for clustering high throughput machine cycle conditions

Cyber-physical systems (CPS) have opened up a wide range of opportunities in terms of performance analysis that can be applied directly to the machine tool industry and are useful for maintenance systems and machine designers. High-speed communication capabilities enable the data to be gathered, pre-processed and processed for the purpose of machine diagnosis. This paper describes a complete real-world CPS implementation cycle, ranging from machine data acquisition to processing and interpretation. In fact, the aim of this paper is to propose a CPS for machine component knowledge discovery based on clustering algorithms using real data from a machining process. Therefore, it compares three clustering algorithms —k-means, hierarchical agglomerative and Gaussian mixture models— in terms of their contribution to spindle performance knowledge during high throughput machining operation