Remaining useful life prediction techniques for electric valves based on convolution auto encoder and long short term...

Area of Science:

• Nuclear Engineering
• Artificial Intelligence
• System Maintenance

Background:

• Nuclear power systems require optimized operation and maintenance for safety and economic efficiency.
• Electric valves are critical components whose failure can impact system performance.
• Accurate prediction of component degradation is essential for proactive maintenance strategies.

Purpose of the Study:

• To develop an advanced Remaining Useful Life (RUL) prediction method for electric valves in nuclear power systems.
• To enhance the safety and economic operation of nuclear plants through improved maintenance strategies.
• To leverage deep learning for more accurate and reliable component health monitoring.

Main Methods:

• A hybrid deep learning model combining Convolutional Auto-Encoder (CAE) for feature extraction and Long Short-Term Memory (LSTM) for time-series analysis.
• Enrichment of input features for LSTM by designing a parallel structure integrating CAE outputs with original data.
• Systematic comparison of network architectures and hyper-parameters to optimize model performance.

Main Results:

• The proposed CAE-LSTM method demonstrates high accuracy in RUL prediction for electric valves.
• Feature enrichment through parallel data structures significantly improves prediction performance.
• The developed model outperforms other conventional machine learning algorithms in accuracy and reliability.

Conclusions:

• The integrated CAE-LSTM approach offers a robust and accurate solution for RUL prediction in critical systems.
• This innovation contributes to enhanced safety and economic operation of nuclear power plants.
• The method has potential applications in the maintenance of other complex industrial systems.