Validation of Probabilistic Regime Recognition and Damage Estimation with Large Fleet Datasets

Regime recognition (RR) and damage estimation is a critical step in the creation of usage and damage spectra and in development of component replacement timelines. While a variety of methodologies exist for regime recognition based on deterministic classification, recent work has demonstrated the benefit of approaching the problem of regime classification and damage estimation from a probabilistic standpoint. However, while studies have shown the successful deployment of probabilistic algorithms on simulated data, little work has been done to validate their use on large fleet datasets that contain actual flight data. This paper seeks to address two main points. First, the damage estimation scheme proposed in prior work is extended to incorporate uncertainty in the component damage rates, allowing for a complete treatment of the damage estimation problem that fully captures all major sources of uncertainty. Second, this paper exercises the probabilistic regime recognition and damage estimation algorithms published in previous work on 500 hours of actual fleet data, validating the efficacy of the approach. Detailed analysis of the results show that the probabilistic algorithms are effective in estimating flown regimes and component damage fractions and offer unique insights that are not available from comparable deterministic algorithms.