Feature Relevance Modeling for Early Operating State Detection

The patent transforms the monitoring approach by changing from fixed threshold parameters to dynamic baseline parameters that adapt to gradual equipment degradation. The system continuously updates baseline values based on historical data, allowing detection of deviations from normal behavior patterns rather than comparing against static thresholds. This enables reliable detection of abnormal states while maintaining system simplicity through automated parameter adaptation.

The patent implements self-service through automated baseline generation and update mechanisms. The system automatically learns normal operation patterns from historical data and continuously refines its own prediction rules without requiring manual expert intervention. This self-learning capability enables early detection of abnormal states while eliminating the time-consuming process of manual rule establishment and confirmation, as the system autonomously adapts to equipment degradation patterns.

The patent applies dynamics by implementing a living baseline that continuously adapts to changing equipment behavior. Instead of static rules, the system dynamically updates baseline values based on ongoing operation data, automatically adjusting to gradual wear patterns and post-maintenance states. This dynamic adaptation maintains ease of operation through automated processes while achieving high adaptability to behavioral changes, eliminating the need for manual rule updates.