Remaining Useful Life Prediction Method for Stochastic Degrading Devices Considering Predictive Maintenance

Area of Science:

• Engineering
• Reliability Engineering
• Data Science

Background:

• Predictive maintenance is crucial for device longevity but often overlooks maintenance's effect on sensor data and RUL for repairable systems.
• Existing RUL prediction methods may lack accuracy due to unaddressed impacts of preventive maintenance on degradation.
• Repairable degrading devices present unique challenges in RUL estimation due to maintenance interventions.

Purpose of the Study:

• To develop an adaptive remaining useful life (RUL) prediction method for repairable degrading devices.
• To improve the accuracy and adaptability of RUL predictions by considering the impact of imperfect maintenance.
• To provide a robust framework for health management of degrading systems.

Main Methods:

• Established a phased degradation model using an adaptive Wiener process to incorporate imperfect maintenance effects.
• Derived the probability distribution of RUL based on the first hitting time (FHT) concept, integrating maintenance impacts on degradation.
• Developed a model parameter identification and updating method using maximum likelihood estimation and Bayesian inference, accounting for device variations.

Main Results:

• The proposed adaptive RUL prediction method effectively accounts for the influence of maintenance on degradation.
• The model demonstrates improved accuracy and adaptability in predicting RUL for repairable devices.
• Validation through numerical simulations and real-world gyroscope degradation data confirms the method's effectiveness.

Conclusions:

• The adaptive RUL prediction method offers a significant advancement in managing repairable degrading devices.
• Accounting for maintenance impacts is essential for accurate RUL prediction in such systems.
• The developed approach enhances device lifespan management and operational reliability.