Remaining Useful Life Prediction Using Dual-Channel LSTM with Time Feature and Its Difference

Area of Science:

• Mechanical Engineering
• Artificial Intelligence
• Signal Processing

Background:

• Current machinery remaining useful life (RUL) prediction methods often rely on multi-sensor feature extraction, which can be complex and inaccurate in noisy or abnormal operational conditions.
• Challenges include increased model complexity and reduced prediction accuracy due to noise, alongside difficulties in effectively utilizing temporal sensor data.
• The need for robust and accurate RUL prediction is critical for predictive maintenance and operational efficiency in complex systems.

Purpose of the Study:

• To propose a novel dual-channel long short-term memory (LSTM) neural network model for enhanced machinery RUL prediction.
• To address the limitations of existing methods by adaptively selecting and processing temporal sensor features.
• To improve the accuracy and stability of RUL predictions, particularly in challenging operational environments.

Main Methods:

• A dual-channel LSTM neural network architecture is developed to process sensor data.
• The model adaptively selects relevant temporal features and applies first-order processing to these features.
• A creative momentum-smoothing module is incorporated to stabilize the predicted RUL curve and enhance accuracy.

Main Results:

• The proposed dual-channel LSTM model demonstrates superior performance in RUL prediction compared to existing methods.
• Adaptive feature selection and first-order processing effectively utilize temporal sensor characteristics.
• The momentum-smoothing module significantly improves the smoothness and accuracy of the predicted RUL curve.

Conclusions:

• The dual-channel LSTM model offers an effective and stable approach for machinery RUL prediction.
• The method overcomes noise-induced complexities and improves accuracy in diverse operational scenarios.
• Validation on the C-MAPSS dataset confirms the practical applicability and robustness of the proposed technique.