Optimal Diagnosis Strategy via Functional Extremum Transformation and Fault Elimination Neural Network: A Minimum...

Area of Science:

• Complex Systems Analysis
• Artificial Intelligence in Engineering
• Fault Diagnosis

Background:

• Expert knowledge-based methods are common for complex systems with limited fault data.
• Traditional multi-signal flow graph models face scalability challenges in generating minimum-cost diagnostic strategies.
• Growing system complexity leads to high diagnostic costs and computation times.

Purpose of the Study:

• To reformulate the minimum-cost diagnostic strategy generation problem (MDP) for complex systems.
• To develop an efficient method that overcomes the limitations of existing qualitative approaches.
• To reduce diagnostic cost and computation time while maintaining diagnostic accuracy.

Main Methods:

• Reformulated the minimum-cost diagnostic strategy generation problem (MDP) as a functional extremum value problem (FEVP) using a multipartite graph model.
• Developed a fault elimination neural network with an intelligent algorithm (ENI) structure.
• The ENI learns an elimination function to retain promising fault subsets and discard redundant ones, optimizing a non-differentiable objective.

Main Results:

• The proposed ENI method reduces diagnostic cost by over 30% compared to existing methods.
• Acceptable computation time is maintained, even for large-scale systems.
• Demonstrated significant advantages in reducing diagnostic costs for larger and more complex systems.

Conclusions:

• The FEVP reformulation and ENI approach offer an effective solution for minimum-cost diagnostic strategy generation in complex systems.
• This method significantly improves efficiency and reduces costs, particularly for large-scale systems.
• The approach provides a scalable and computationally efficient alternative to traditional methods.