A novel quantitative diagnosis method for rolling bearing faults based on digital twin model

Area of Science:

• Mechanical Engineering
• Condition Monitoring
• Signal Processing

Background:

• Quantitative diagnosis of rolling bearings relies on analyzing dual-impulse behaviors.
• Extracting precise fault moments from noisy signals remains a significant challenge.

Purpose of the Study:

• To propose a novel quantitative diagnosis method using a digital twin model for assessing bearing fault severity.
• To accurately extract fault information from noise-contaminated raw signals.

Main Methods:

• Derivation of a quantitative diagnostic criterion for bearing faults to analyze instantaneous response characteristics.
• Construction of a digital twin model to characterize fault signatures using noise-free twin signals.
• Real-time optimization of the twin model using a recursive parameter optimization strategy based on cosine similarity (RPOS-CS).
• Application of fault parameters from the optimal signal to evaluate bearing fault size.
• Utilization of kernel density estimation for uncertainty analysis and interval estimation.

Main Results:

• The proposed digital twin method accurately assesses bearing fault severity from raw signal waveforms.
• The RPOS-CS strategy effectively optimizes the twin model in real time.
• Kernel density estimation enhances the reliability of diagnosis results through uncertainty analysis.
• Validation with simulated and experimental signals demonstrates high diagnostic accuracy and reliability.

Conclusions:

• The novel digital twin-based method offers a reliable approach for quantitative diagnosis of rolling bearing faults.
• The technique effectively overcomes challenges associated with noise-contaminated signals.
• The study highlights the potential for improved condition monitoring and predictive maintenance in machinery.