Adaptive parameter-estimation-method selection and life-distribution optimization for censored aero-engine data

In aero-engine reliability analysis, estimator selection for censored data and life-distribution choice often rely on empirical judgement. This paper proposes an engineering-oriented framework integrating standardized data preprocessing, multi-criteria distribution selection, and adaptive final parameter estimation for multi-source operational data. Heterogeneous records are mapped to complete, right-censored, left-censored, and interval-censored observations. Candidate distributions are fitted consistently by maximum likelihood and screened by a parametric-bootstrap goodness-of-fit test. AIC, BIC, and the probability-plot correlation coefficient (CC) are combined using entropy-derived weights, with a prespecified mechanism-based physical-consistency constraint. After selecting a distribution family, the final estimator is selected from a validated rule or a rule constructed and validated for the target scenario. The framework is illustrated using 15 failures and 51 right-censored observations from the PW4000-100 No. 3 bearing-compartment heat shield operated by China Southern Airlines. Within the examined Weibull simulation grid, the 25-cell adaptive lookup rule reduced average regret relative to always using MLE without increasing the maximum scenario-wise mean NMSE. Entropy-weight fusion performed comparably to AIC-only, BIC-only, and equal-weight scoring while avoiding manually specified criterion weights. The framework provides traceable statistical support for reliability assessment under explicitly stated censoring and model assumptions.