Machine Anomaly Monitoring by Operating-Point Clustering

The patent segments the continuous operating space into discrete operating-point clusters based on similarity metrics. Each cluster represents a group of similar operating conditions, allowing monitoring to be tailored to specific operating points rather than applying a single general threshold across all conditions. This segmentation enables precise fault detection within each cluster while maintaining comprehensive coverage across the entire operating range.

The patent applies local quality by creating operating-point-specific monitoring characteristics and thresholds for each cluster. Instead of using uniform monitoring parameters globally, the system adapts the monitoring criteria to local operating conditions. Each operating-point cluster has its own anomaly recognition model trained on locally relevant data, improving detection precision for faults occurring under specific operating conditions.

The patent implements universality by creating a unified clustering framework that groups operating points based on their similarities. The same clustering algorithm and anomaly recognition approach are applied across all operating-point clusters, providing a universal methodology that handles diverse operating conditions. This multi-functional system can adapt to any operating point by assigning it to the appropriate cluster, avoiding the need for completely separate monitoring systems for each operating point.

The patent uses parameter changes by dynamically adjusting monitoring thresholds and characteristics based on the identified operating-point cluster. The system changes its monitoring parameters according to the current operating conditions, selecting appropriate thresholds and anomaly criteria from the trained models corresponding to the active cluster. This allows precise fault detection adapted to local conditions without requiring manual configuration for each operating point.

The patent implements self-service by enabling the monitoring system to automatically adapt to different operating points through unsupervised clustering and automated model training. The system performs self-configuration by identifying operating-point clusters from operational data and training appropriate anomaly recognition models without requiring manual intervention. This automated approach maintains implementation simplicity while achieving high adaptability to varying operating conditions.

The patent applies dynamics by creating a flexible monitoring system that can dynamically adjust to changing operating conditions. The clustering framework allows the system to automatically reorganize operating-point groups as new operating conditions are encountered, and the anomaly recognition models can be retrained to adapt to evolving operational patterns. This dynamic capability enables the system to maintain simplicity while being highly adaptable to different and changing operating points.