A sparrow search algorithm-optimized LSTM framework with EMD denoising for rolling element bearing remaining useful...

Area of Science:

• Mechanical Engineering
• Artificial Intelligence
• Signal Processing

Background:

• Accurate remaining useful life (RUL) prediction is crucial for predictive maintenance in rotating machinery.
• Challenges include severe signal noise and the complex, nonlinear nature of bearing degradation.
• Existing methods often provide point estimates, lacking quantified uncertainty.

Purpose of the Study:

• To propose a novel hybrid deep learning framework for enhanced bearing RUL prediction.
• To address limitations of signal noise and complex degradation processes.
• To provide probabilistic RUL predictions with quantified uncertainty.

Main Methods:

• Empirical Mode Decomposition (EMD) for signal denoising and feature extraction.
• Sparrow Search Algorithm (SSA) for optimizing Long Short-Term Memory (LSTM) network hyperparameters.
• First-passage-time model (inverse Gaussian distribution) for probabilistic RUL estimation.

Main Results:

• The proposed EMD-SSA-LSTM framework achieved superior prediction accuracy compared to benchmark methods (GA-LSTM, PSO-LSTM).
• Demonstrated faster convergence and enhanced robustness in experimental validation.
• Successfully provided probabilistic RUL predictions with quantified uncertainty.

Conclusions:

• The EMD-SSA-LSTM framework offers a comprehensive and effective solution for data-driven bearing prognostics.
• The integration of adaptive signal processing, intelligent optimization, and probabilistic modeling significantly improves RUL prediction.
• Contributes to more reliable predictive maintenance strategies in industrial applications.