The Cubic Dynamic Uncertain Causality Graph: A Methodology for Temporal Process Modeling and Diagnostic Logic...

Area of Science:

• Engineering
• Computer Science
• Systems Science

Background:

• Complex systems require dynamic and reliable real-time fault diagnosis.
• Existing methods often rely on restrictive assumptions about graph structure and static time slices.

Purpose of the Study:

• To extend Dynamic Uncertain Causality Graphs (DUCG) with novel temporal causality modeling and reasoning.
• To develop a new methodology, Cubic DUCG, for accurate and efficient dynamic causal reasoning in fault spreading.

Main Methods:

• Developed the Cubic DUCG methodology for compact representation and accurate reasoning of dynamic causalities.
• Incorporated continuous causality graph generation across time slices, handling complex causalities including negative feedback loops.
• Utilized a rigorous inference algorithm based on complete real-time causalities and implemented solutions for causality simplification and reasoning complexity reduction.

Main Results:

• The Cubic DUCG allows causal connections across multiple time slices, discarding restrictive structural assumptions.
• It models complex causalities, including dynamic negative feedback loops, intuitively.
• The inference algorithm provides real-time fault situation reflection, outperforming static aggregation.

Conclusions:

• The Cubic DUCG methodology offers accurate, robust, and efficient real-time fault diagnosis for complex systems.
• Experiments on a nuclear power plant simulator validated the proposed approach's effectiveness.