Fault Detection of Bearing Systems through EEMD and Optimization Algorithm

Area of Science:

• Mechanical Engineering
• Signal Processing
• Machine Learning

Background:

• Bearing systems are critical in machinery, and their failure can lead to significant downtime and costs.
• Early and accurate fault detection in bearings is essential for predictive maintenance and operational reliability.
• Existing methods may struggle with complex vibration signatures and require robust feature extraction.

Purpose of the Study:

• To develop an advanced fault detection and diagnosis method for bearing systems.
• To enhance the accuracy and visualization of bearing damage classification.
• To improve the performance of bearing fault diagnosis through optimized feature selection.

Main Methods:

• Utilizing ensemble empirical mode decomposition (EEMD) for vibration signal decomposition into intrinsic mode functions (IMFs).
• Extracting damage-sensitive statistical features to form a parameter vector.
• Applying particle swarm optimization (PSO) for optimal weighting of the parameter vector.
• Employing principal component analysis (PCA) and Isomap for classification and 3D visualization of bearing conditions.

Main Results:

• Successful generation of vibration signals from simulated damaged bearing components (inner-race, outer-race, rolling elements).
• Development of a damage-sensitive parameter vector through EEMD-based feature extraction.
• Effective classification and visualization of healthy versus damaged bearing components using PCA and Isomap.
• Demonstrated improvement in classification performance via PSO-based optimization for enhanced separation and grouping of parameter vectors.

Conclusions:

• The proposed EEMD-based method, combined with PSO, PCA, and Isomap, offers a robust approach for bearing fault detection and diagnosis.
• The technique provides superior visualization capabilities, aiding in the clear differentiation of bearing health states.
• This method contributes to more reliable predictive maintenance strategies in mechanical systems.