在如 IC 產品的一些電子零件中，由於製程或其他因素產生瑕疵，導致產品前期早衰（infant mortality）失效，而非瑕疵品則最終會因磨耗（wear-out）而失效。廣義有限失效母體（Generalized Limited Failure Population，GLFP）模型，可用於分析同時受因製程產生的瑕疵導致產品前期早衰，與後期終因長期磨損失效的產品或電子零件之失效時間數據。另一方面，自 E-M 演算法後，資料增補 (data augmentaion) 在統計學中的應用極為廣泛。本研究於對數常態分布 GLFP 模型，以貝氏方法結合資料增補建構完整概似函數，在各參數具共軛先驗分布下，以吉布斯抽樣（Gibbs sampling）加速馬可夫鍊蒙地卡羅（Markov chain Monte Carlo，MCMC）演算法的計算過程，從而提升計算效率。同時藉由資料增補的隱藏變數之後驗抽樣過程中，為每筆失效資料之失效模式和其是否為瑕疵品進行預測，並針對 Backblaze 公司所提供的硬碟資料，考慮不同的先驗資訊，進行壽命之可靠度相關分析。

Some electronic components such as integrated circuits contain latent defects introduced during manufacturing or other processes, causing a subset of units to fail prematurely (infant mortality), while non-defective units eventually fail due to wear-out. The generalized limited failure population (GLFP) model captures this dual behavior by jointly modeling early failures arising from manufacturing defects and later failures driven by long-term degradation. Based on the widely used data augmentation technique in statistics, this study develops a Bayesian GLFP framework with log-normal lifetime distributions. By augmenting the latent data, we construct the complete likelihood, assign conjugate priors to all parameters, and employ Gibbs sampling to accelerate Markov chain Monte Carlo (MCMC). The posterior draws of the latent augmentation variables simultaneously yield predictive classifications of each observed failure mode and defect status. Finally, using hard-drive failure data released by Backblaze, we perform reliability analysis under various prior assumptions to illustrate the practical utility of the proposed method.

[1] Albert, J. R. G. and Baxter, L. A. (1995). Applications of the EM algorithm to the
analysis of life length data. Journal of the Royal Statistical Society Series C: Applied
Statistics, 44, 323–341.
[2] Balakrishnan, N., So, H. Y., and Ling, M. H. (2015). EM algorithm for one-shot
device testing with competing risks under Weibull distribution. IEEE Transactions
on Reliability, 65, 973–991.
[3] Basu, S., Sen, A., and Banerjee, M. (2003). Bayesian analysis of competing risks with
partially masked cause of failure. Journal of the Royal Statistical Society Series C:
Applied Statistics, 52, 77–93.
[4] Berger, J. O. (1985). Statistical Decision Theory and Bayesian Analysis (Second
Edition). Springer-Verlag, New York, US.
[5] Chan, V. and Meeker, W. Q. (1999). A failure-time model for infant mortality and
wearout failure modes. IEEE Transactions on Reliability, 48, 377–387.
[6] Dempster, A. P., Laird, N. M., and Rubin, D. B. (1977). Maximum likelihood from
incomplete data via the EM algorithm. Journal of the Royal Statistical Society: Series
B, 39, 1–22.
[7] Diebolt, J., and Robert, C. P. (1994). Estimation of finite mixture distributions
through Bayesian sampling. Journal of the Royal Statistical Society, 56, 363–375.
[8] Fan, T. H. and Hsu, T. M. (2014). Statistical inference of a two-component series sys-
tem with correlated log-normal lifetime distribution under multiple type-I censoring.
IEEE Transactions on Reliability, 64, 376–385.
46
[9] Gelman, A., Meng, X. L., and Stern, H. (1996). Posterior predictive assessment of
model fitness via realized discrepancies. Statistica Sinica, 6, 733–807.
[10] Louis, T. A. (1982). Finding the observed information matrix when using the EM
algorithm. Journal of the Royal Statistical Society Series B: Statistical Methodology,
44, 226–233.
[11] McLachlan, G. J. and Peel, D. (2000). Finite Mixture Models. John Wiley and Sons,
New York, US.
[12] Meeker, W. Q. (1987). Limited failure population life tests: application to integrated
circuit reliability. Technometrics, 29, 51–65.
[13] Meeker, W. Q. and Escobar, L. A. (1998). Statistical Methods for Reliability Data,
John Wiley and Sons, New York, US.
[14] Mittman, E., Lewis-Beck, C., and Meeker, W. Q. (2019). A hierarchical model for
heterogeneous reliability field data. Technometrics, 61, 354–368.
[15] Miyakawa, M. (1984). Analysis of incomplete data in competing risks model. IEEE
Transactions on Reliability, 33, 293–296.
[16] Orchard, T. and Woodbury, M. A. (1972). A missing information principle: theory
and applications. In Proceedings of the Sixth Berkeley Symposium on Mathematical
Statistics and Probability, 1, 697–716.
[17] Park, C. (2005). Parameter estimation of incomplete data in competing risks using
the EM algorithm. IEEE Transactions on Reliability, 54, 282–290.
[18] Rubin, D. B. (1984). Bayesianly justifiable and relevant frequency calculations for
the applied statistician. The Annals of Statistics, 12, 1151–1172.
[19] Tai, C. Y. and Fan, T. H. (2025). Reliability inference in GLFP models based on
EM algorithm with related application. Accepted by Applied Stochastic Models in
Business and Industry.
[20] Tan, C. M. and Raghavan, N. (2007). An approach to statistical analysis of gate
oxide breakdown mechanisms. Microelectronics Reliability, 47, 1336–1342.
47
[21] Tobias, P. A. and Trindade, D. C. (1995). Applied Reliability. 2nd Edition, Chapman
and Hall/CRC, New York, US.
[22] Usher, J. S. and Hodgson, T. J. (1988). Maximum likelihood analysis of component
reliability using masked system life-test data. IEEE Transactions on Reliability, 37,
550–555.
[23] 陳睦璿 (2022). 多重失效模式微電子資料之可靠度分析，國立中央大學碩士論文。