Estimating Residual Life Distributions of Complex Operational Systems Using a Remaining Maintenance Free Operating...

Area of Science:

• Condition monitoring research
• Predictive maintenance algorithms
• Machinery health prognostics

Background:

• Industrial implementation of condition monitoring is limited by a lack of real-world operational data.
• Existing prognosis algorithms are often validated using idealized lab data, not reflecting operational complexities.
• Bridging the gap between academic research and industrial application is crucial for effective preventative maintenance.

Purpose of the Study:

• To present a novel prognosis methodology that integrates academic findings with industrial needs.
• To introduce and define the remaining maintenance-free operating period (RMFOP) for improved degradation data utilization.
• To validate a new methodology using real-world operational data from railway switch systems.

Main Methods:

• Definition and application of the remaining maintenance-free operating period (RMFOP).
• Extraction of degradation patterns and fitting them into linear or exponential random coefficients regression models.
• Computation and updating of system residual life distributions via parameter statistics estimation.

Main Results:

• The RMFOP-based methodology was validated using real-world degradation data from operational railway switches.
• Both linear and exponential regression models demonstrated sufficient prediction accuracy for residual life.
• The exponential model provided superior predictions, with accuracy increasing as more of the system's life elapsed.

Conclusions:

• The proposed RMFOP methodology effectively bridges the gap between condition monitoring research and industrial practice.
• The methodology enables accurate prediction of system residual life distributions, enhancing preventative maintenance strategies.
• The approach successfully identified and predicted incipient overdriving faults in railway switch systems.