Spatio-Temporal Collaborative Perception-Enabled Fault Feature Graph Construction and Topology Mining for Variable...

Area of Science:

• Mechanical Engineering
• Artificial Intelligence
• Data Science

Background:

• Industrial equipment fault diagnosis is challenged by data distribution discrepancies across operating conditions, limiting generalization.
• Multi-source data often contains underutilized spatio-temporal information, hindering effective feature extraction.

Purpose of the Study:

• To develop a spatio-temporal collaborative perception-driven methodology for constructing fault feature graphs and mining topology for variable-condition diagnosis.
• To enhance the generalization capabilities and diagnostic accuracy of industrial equipment fault diagnosis systems.

Main Methods:

• Constructed fault feature graphs using single-source data and similarity clustering, validating operational condition invariance.
• Revealed spatio-temporal correlations within multi-source feature topologies via collaborative perception.
• Developed a graph residual convolutional network for mining multi-source spatio-temporal features.

Main Results:

• Feature graphs successfully integrated multi-source information despite operational variations.
• The methodology accurately captured spatio-temporal delays from vibrational path discrepancies.
• The proposed model achieved high diagnostic accuracy and generalization under complex conditions.

Conclusions:

• The developed framework offers a highly reliable approach for rotating machinery fault diagnosis.
• The methodology effectively addresses challenges in variable-condition industrial equipment diagnosis.
• Spatio-temporal collaborative perception enhances feature representation and diagnostic performance.