Adaptive Diagnosis for Rotating Machineries Using Information Geometrical Kernel-ELM Based on VMD-SVD

Area of Science:

• Mechanical Engineering
• Signal Processing
• Machine Learning

Background:

• Rotating machinery operates under challenging, variable conditions, complicating fault diagnosis.
• Existing methods often require significant prior knowledge or human intervention, limiting their adaptability.

Purpose of the Study:

• To develop a novel adaptive fault diagnosis algorithm for rotating machinery under variable operating conditions.
• To enhance the self-adaptability and generalization ability of existing machine learning models for fault diagnosis.

Main Methods:

• Proposed a new algorithm: Information Geometrical Extreme Learning Machine with Kernel (IG-KELM).
• Utilized variation mode decomposition (VMD) for signal decomposition and singular value decomposition (SVD) for feature extraction.
• Applied IG-KELM for self-adaptive fault diagnosis with minimal prior knowledge.

Main Results:

• The proposed IG-KELM method, combined with VMD and SVD, demonstrated superior performance in adaptive fault diagnosis.
• Achieved accuracy improvements of up to 7.25% for bearing faults and 7.78% for hydraulic pump faults compared to conventional methods.

Conclusions:

• The developed IG-KELM based approach offers a robust and self-adaptive solution for fault diagnosis in rotating machinery.
• This method significantly improves diagnostic accuracy under variable operating conditions without extensive human intervention.