Diagnosis of High-Speed Ball-Bearing Spindles by Data Mining of Dynamic Responses from Various Rotating Elements

The increasing demand for high-speed machine tools has elicited the widespread adoption of specially designed spindles that incorporate built-in motors and ball bearings. To ensure the durability and reliability of these spindles during high-speed operations, various sensors and control devices are employed to actively regulate their dynamic characteristics. However, the direct measurement of internal heat generation or high-frequency bearing vibrations poses substantial challenges, hindering the reliable diagnosis of failure signals. Herein, we propose a data mining technology to predict failures in high-speed ball spindles by leveraging a comparative analysis of dynamic responses with reference patterns based on specific failure types. For this purpose, dynamic signals combined with various failure responses are rigorously analyzed and characterized in the frequency domain. Thereafter, the resulting individual failure responses are subjected to data mining analysis wherein four feature identification scores are evaluated within six reference frequency ranges. The entire lifetime of the spindle is categorized into four distinct degradation stages, namely the initial, propagation, developed, and deterioration stages. This categorization enables efficient and reliable estimation of spindle-failure progression by precisely measuring and analyzing the dynamic responses of the high-speed ball spindle. The proposed data mining technology enhances the ability to predict and diagnose failures in high-speed ball spindles, thereby facilitating timely maintenance and reducing downtime in manufacturing processes.