Machine Tool Data Correlation for Predictive Maintenance

The patent segments the collected information into multiple categories including sensor information (vibration, temperature, acoustic emission), machine tool operation information (spindle speed, feed rate, depth of cut), and tool information (tool ID, tool life). This segmentation allows for targeted analysis and utilization of specific information types rather than treating all data uniformly, thereby improving effectiveness despite increased data quantity.

The patent transforms raw collected information into processed information by changing parameters such as calculating tool life from operation data, determining vibration levels from sensor data, and assessing tool condition through multiple indicators. This parameter transformation enables more effective utilization of the collected information for predictive maintenance and optimization.

The patent creates a unified information processing framework that handles multiple types of information (sensor data, operation data, tool information) through a single integrated system. The processing unit applies universal algorithms to diverse data types, such as using correlation analysis for both vibration data and operation parameters, thereby managing complexity while maintaining versatility.

The patent introduces an information processing unit as an intermediary between data collection and utilization. This intermediary processes raw information from multiple sources, transforms it into standardized formats, and presents processed results to users. The intermediary layer simplifies the complexity by handling data integration, validation, and transformation centrally rather than requiring complex point-to-point processing.

The patent performs preliminary correlation learning between part information and sensor information during periods when full processing capacity is available, building predictive models in advance. By pre-establishing these correlations, the system can quickly apply the learned relationships to new data without requiring extensive real-time computation, thereby reducing processing time while maintaining predictive capability.

The patent implements partial correlation learning by focusing on the most critical correlations between specific sensor types and part conditions rather than analyzing all possible combinations. This selective approach captures the essential predictive relationships while significantly reducing the computational burden and processing time required.