A Correlation Driven Approach with Edge Services for Predictive Industrial Maintenance.

Area of Science:

• Industrial Engineering
• Computer Science
• Data Science

Background:

• Current predictive maintenance models rely on build-time data, failing to account for fault propagation.
• Sensor and device anomalies can spread through hidden correlations, impacting maintenance effectiveness.
• Predicting this propagation is challenging due to multiple potential pathways and uncertainty.

Purpose of the Study:

• To develop a predictive industrial maintenance approach that accounts for anomaly/fault propagation.
• To leverage sensor correlations for more accurate fault prediction and proactive maintenance scheduling.
• To address the uncertainty in anomaly/fault propagation pathways.

Main Methods:

• Proposes a correlation-based event routing approach, building on previous work.
• Maps physical sensors to proactive data services, encapsulating anomalies/faults as events.
• Utilizes a service hyperlink model to represent correlations and maps propagation to event routing.
• Develops a heuristic algorithm for routing events among services based on service hyperlinks.

Main Results:

• The proposed correlation-based event routing approach demonstrates high performance.
• Achieved 100% precision in predicting anomaly/fault propagation.
• Reached a maximum recall of 88.89% in the experiments.

Conclusions:

• The developed method effectively predicts anomaly/fault propagation by considering sensor correlations.
• This approach enhances predictive industrial maintenance by enabling proactive scheduling and minimizing unexpected failures.
• The event routing strategy offers a robust solution for managing fault propagation uncertainty.