Smart Fault-Detection Machine for Ball-Bearing System with Chaotic Mapping Strategy

Area of Science:

• Mechanical Engineering
• Signal Processing
• Nonlinear Dynamics

Background:

• Industrial ball-bearing systems exhibit complex vibration signals.
• Distinguishing between normal and various fault states (inner race, outer race, ball) is challenging using traditional time-series analysis.
• Effective fault detection is crucial for predictive maintenance and operational efficiency.

Purpose of the Study:

• To develop a smart fault-detection approach for industrial ball-bearing systems.
• To enhance the classification accuracy of different bearing states.
• To leverage chaotic dynamics for improved signal analysis.

Main Methods:

• A nonlinear error dynamic system and chaotic mapping strategy were developed to convert time-series vibration signals into the chaotic domain.
• 3D phase portraits were generated to visualize the converted signals.
• Clustering methods, including Euclidean distance (ED) and kernel K-means (KM), were employed for classification.
• Autonomously adjusted feature value ranges were used for identification.

Main Results:

• The proposed approach effectively converts vibration signals into distinguishable 3D phase portraits.
• Clustering methods successfully identified different bearing states based on these portraits.
• Experimental results demonstrated high effectiveness and feasibility of the developed method.
• Near 100% accuracy was achieved in the testing stage for fault detection.

Conclusions:

• The smart fault-detection approach using chaotic mapping is effective for industrial ball-bearing systems.
• The method provides a robust way to classify normal and various fault conditions.
• This technique offers a promising solution for accurate and efficient bearing health monitoring.