Automatic Fault Detection and Isolation Method for Roller Bearing Using Hybrid-GA and Sequential Fuzzy Inference

Area of Science:

• Mechanical Engineering
• Vibration Analysis
• Condition Monitoring

Background:

• Accelerometers are crucial for bearing condition diagnosis but often must be placed far from the bearing in industrial equipment.
• Signals measured at a distance are susceptible to higher noise levels, complicating accurate fault detection.
• Existing methods struggle with detecting bearing faults when sensor placement is suboptimal due to signal degradation.

Purpose of the Study:

• To propose a novel automated method for accurate bearing fault detection using signals from remote accelerometers.
• To overcome the challenges posed by increased noise in vibration signals acquired at a distance from the bearing.
• To enable simple and precise automatic diagnosis of bearing faults irrespective of sensor proximity.

Main Methods:

• Utilized a hybrid genetic algorithm (GA) and tabu search to optimize high-pass filter cutoff frequencies for extracting fault signals.
• Employed possibility theory and fuzzy inference for precise diagnosis of bearing faults.
• Applied the developed methods to vibration signals measured from a distance to validate effectiveness.

Main Results:

• Successfully extracted bearing fault signals from noisy data acquired at a distance.
• Achieved accurate and automatic diagnosis of bearing faults using the proposed signal processing and inference techniques.
• Demonstrated the practical applicability and efficiency of the method in real-world scenarios.

Conclusions:

• The proposed automated method effectively detects bearing faults using vibration signals from distant sensors.
• The hybrid GA approach for filter optimization and fuzzy inference provide a robust solution for challenging diagnostic scenarios.
• This technique enhances the reliability of condition monitoring for bearings in industrial settings where optimal sensor placement is not feasible.