Intelligent Fault Diagnosis of Machinery Using BPSO-Optimized Ensemble Filters and an Improved Sparse Representation...

Area of Science:

• Engineering
• Computer Science

Background:

• Machinery fault diagnosis is crucial for industrial maintenance.
• Existing methods may lack accuracy or efficiency.

Purpose of the Study:

• To propose an ensemble approach for intelligent machinery fault diagnosis.
• To introduce a novel aggregation strategy, Cumulative Reconstruction Residual (CRR).

Main Methods:

• An ensemble approach combining six feature selection filters with an Improved Sparse Representation Classifier (ISRC).
• Hyper-parameter optimization using Binary Particle Swarm Optimization.
• Aggregation of base model outputs using CRR, replacing traditional voting.

Main Results:

• The proposed ensemble method demonstrates effectiveness on mechanical datasets (bearings, gears).
• CRR shows superiority over the voting strategy in ensemble fault diagnosis.
• ISRC offers high classification accuracy and reduced computation time.

Conclusions:

• The developed ensemble approach with CRR is a promising technique for intelligent machinery fault diagnosis.
• The CRR aggregation method provides a more effective alternative to voting.