Data-driven remaining useful life prediction via multiple sensor signals and deep long short-term memory neural...

Area of Science:

• Engineering
• Computer Science
• Data Science

Background:

• Remaining Useful Life (RUL) prediction is critical for system availability and cost reduction.
• Traditional methods may struggle with complex, long-term dependencies in sensor data.

Purpose of the Study:

• To propose a novel Deep Long Short-Term Memory (DLSTM) network for accurate RUL prediction.
• To leverage multi-sensor time series data for enhanced predictive maintenance.

Main Methods:

• Utilized a DLSTM network architecture to process and fuse multi-sensor time series signals.
• Employed a grid search strategy and adaptive moment estimation for efficient DLSTM model tuning.
• Validated the proposed method on two distinct turbofan engine datasets.

Main Results:

• The DLSTM model demonstrated robust performance in RUL prediction.
• Achieved competitive results compared to existing state-of-the-art methods and other neural network models.
• Effectively captured hidden long-term dependencies within sensor data.

Conclusions:

• The DLSTM network offers a powerful and accurate approach for RUL prediction in complex systems.
• This deep learning method enhances predictive maintenance capabilities by analyzing multi-sensor data.
• The findings support the adoption of DLSTM for improving system reliability and reducing operational costs.