A RUL Estimation System from Clustered Run-to-Failure Degradation Signals

Area of Science:

• Engineering
• Computer Science
• Data Science

Background:

• Prognostics and health management (PHM) enhance system durability and operational lifespan.
• Remaining Useful Life (RUL) estimation is crucial for predictive maintenance planning.
• Accurate RUL prediction minimizes downtime and optimizes resource allocation.

Purpose of the Study:

• To develop and evaluate a novel prognostics approach for RUL estimation.
• To assess the performance of recurrent neural networks and the Prophet forecasting method in RUL prediction.
• To apply the developed method to real-world degradation data from new generation telescopes.

Main Methods:

• Development of a recurrent neural network (RNN) model for prognostics.
• Integration of the Prophet forecasting tool for RUL estimation.
• Application of the combined approach to non-monotonical degradation signals.

Main Results:

• The developed RNN-Prophet model demonstrated effective RUL estimation capabilities.
• The approach successfully handled complex degradation patterns.
• Validation using real-world telescope data confirmed the system's practical applicability.

Conclusions:

• The proposed RNN-Prophet framework offers a robust solution for RUL estimation in PHM.
• This method provides valuable insights for predictive maintenance strategies.
• The study highlights the potential of advanced machine learning techniques in system health management.