An Industrial Fault Diagnostic System Based on a Cubic Dynamic Uncertain Causality Graph

Area of Science:

• Industrial Systems Engineering
• Artificial Intelligence
• Reliability Engineering

Background:

• Traditional industrial fault diagnosis often requires extensive training data, limiting its application in systems with insufficient data.
• Cloud-native technologies offer scalable and flexible platforms for developing advanced diagnostic systems.

Purpose of the Study:

• To develop and validate an industrial fault diagnosis system capable of handling data-scarce environments.
• To leverage cloud-native technology and the cubic dynamic uncertain causality graph (cubic DUCG) for enhanced fault detection and diagnosis.

Main Methods:

• Development of a cloud-native industrial fault diagnosis system comprising a modular knowledge base and an inference engine.
• Implementation of the cubic dynamic uncertain causality graph (cubic DUCG) algorithm for dynamic causal graph generation and probabilistic inference.
• Modular knowledge base construction by domain experts, representing causal relationships visually.

Main Results:

• The system successfully diagnosed all 24 fault cases in a secondary circuit system at the Ningde nuclear power plant.
• The cubic DUCG algorithm dynamically generated causal graphs from real-time data for accurate fault evolution visualization.
• The developed system demonstrated high diagnostic accuracy and feasibility in a real-world industrial application.

Conclusions:

• The cubic DUCG-based industrial fault diagnosis system is feasible and effective, particularly for systems with limited training data.
• Cloud-native architecture provides a robust platform for implementing complex diagnostic algorithms like cubic DUCG.
• The system offers a promising approach for improving the reliability and safety of industrial operations through accurate fault diagnosis.