Abstract
A general method of lifetime distribution selection for multi-level complete and censored testing data is developed. To alleviate the scarcity of test samples at each individual stress levels, a mathematical model is incorporated to correlate the lifetimes and stresses allowing for utilizing the multi-level lifetime data as a whole. Probabilistic modeling of the model prediction error is made and the equivalence between the lifetime distribution and the error distribution is shown. Based on the error modeling and likelihood functions, a Bayesian framework for lifetime distribution selection is proposed. It is shown that the classical information criterion is a local measure and the Bayes factor is a global measure over the likelihood space. A two-step assessment procedure integrating the goodness-of-fit and asymptotic Bayes factors is given. The proposed method is demonstrated using three engineering examples. Comparisons between the proposed method and the conventional methods such as ratio test and the difference of information criteria are made. The influence of the distribution selection on the subsequent probability of failure estimates is discussed.
Similar content being viewed by others
References
Adnan MAS, Roy S (2017) A sequential discrimination procedure for two almost identically shaped wrapped distributions. J Appl Stat 44(5):872–881
Akaike H (1974) A new look at the statistical model identification. IEEE Trans Autom control 19(6):716–723
Anderson TW, Darling DA, et al. (1952) Asymptotic theory of certain goodness of fit criteria based on stochastic processes. Ann Math Stat 23(2):193–212
Ayyub BM, McCuen RH (2016) Probability, statistics, and reliability for engineers and scientists. CRC Press, Boca Raton
Ba I, Ashkar F (2017) Discrimination between a group of three-parameter distributions for hydro-meteorological frequency modeling. Can J Civ Eng 45(5):351–365
Bain LJ, Engelhardt M (1980) Probability of correct selection of weibull versus gamma based on likelihood ratio. Commun Stat - Theory Methods 9(4):375–381
Balakrishnan N, Cohen AC (2014) Order statistics & inference: estimation methods. Elsevier, Amsterdam
Balakrishnan N, Ng H, Kannan N (2004) Goodness-of-fit tests based on spacings for progressively Type-II censored data from a general location-scale distribution. IEEE Trans Reliab 53(3):349–356
Burnham KP, Anderson DR (2003) Model selection and multimodel inference: a practical information-theoretic approach. Springer, Berlin
Chen P, Xu A, Ye ZS (2016) Generalized fiducial inference for accelerated life tests with Weibull distribution and progressively Type-II censoring. IEEE Trans Reliab 65(4):1737–1744
Chick SE (2001) Input distribution selection for simulation experiments: accounting for input uncertainty. Oper Res 49(5):744–758
Coin D (2017) A goodness-of-fit test for generalized error distribution. Commun Stat - Theory Methods 46 (23):11485–11499
D’Agostino RB (1986) Goodness-of-fit-techniques, vol 68. CRC Press, Boca Raton
Dey AK, Kundu D (2009) Discriminating among the log-normal, Weibull, and generalized exponential distributions. IEEE Trans Reliability 58(3):416–424
Dick E (2004) Beyond ‘lognormal versus gamma’: discrimination among error distributions for generalized linear models. Fish Res 70(2-3):351–366
Dumonceaux R, Antle CE (1973) Discrimination between the log-normal and the Weibull distributions. Technometrics 15(4):923–926
El Adlouni S, Hammami D, Bobée B (2015) Discrimination test between the Halphen (a and b) and the gamma distributions. Stochastic Environmental Research and Risk Assessment 29(1):13–26
Fearn D, Nebenzahl E (1991) On the maximum likelihood ratio method of deciding between the Weibull and gamma distributions. Commun Stat - Theory Methods 20(2):579–593
Guan X, Jha R, Liu Y (2011) Model selection, updating, and averaging for probabilistic fatigue damage prognosis. Struct Saf 33(3):242–249
Guan X, He J, Jha R, Liu Y (2012) An efficient analytical bayesian method for reliability and system response updating based on laplace and inverse first-order reliability computations. Reliability Engineering & System Safety 97(1):1–13
Guan X, Zhang J, Zhou S, Rasselkorde EM, Abbasi W (2014) Probabilistic modeling and sizing of embedded flaws in ultrasonic non-destructive inspections for fatigue damage prognostics and structural integrity assessment. NDT & E International 61:1–9
Gupta RD, Kundu D (2003) Discriminating between Weibull and generalized exponential distributions. Computational Statistics & Data Analysis 43(2):179–196
Hannan E, Quinn B (1979) The determination of the order of an autoregression. Journal of the Royal Statistical Society Series B (Methodological) 41(2):190–195
Hazra NK, Kuiti MR, Finkelstein M, Nanda AK (2018) On stochastic comparisons of minimum order statistics from the location–scale family of distributions. Metrika 81(2):105–123
He J, Guan X (2017) Uncertainty sensitivity analysis for reliability problems with parametric distributions. IEEE Trans Reliab 66(3):712–721
He J, Guan X, Jha R (2016) Improve the accuracy of asymptotic approximation in reliability problems involving multimodal distributions. IEEE Trans Reliab 65(4):1724–1736
Ioannides E, Harris TA (1985) A new fatigue life model for rolling bearings. J Tribol 107(3):367–377
Jeffreys H (1998) The theory of probability. Oxford University Press, New York
Kang YJ, Lim OK, Noh Y (2016) Sequential statistical modeling method for distribution type identification. Struct Multidiscip Optim 54(6):1587–1607
Kappenman RF (1982) On a method for selecting a distributional model. Commun Stat - Theory Methods 11(6):663–672
Karadağ Ö, Aktaş S (2016) Goodness of fit tests for generalized gamma distribution. In: AIP conference proceedings, vol 1738. AIP Publishing, p 480119
Kundu D, Manglick A (2004) Discriminating between the Weibull and log-normal distributions. Naval Research Logistics (NRL) 51(6):893–905
Kundu D, Raqab MZ (2007) Discriminating between the generalized Rayleigh and log-normal distribution. Statistics 41(6):505–515
Kundu D, Gupta RD, Manglick A (2005) Discriminating between the log-normal and generalized exponential distributions. Journal of Statistical Planning and Inference 127(1-2):213–227
Lawless JF (2011) Statistical models and methods for lifetime data, vol 362. Wiley, New York
Liao H, Elsayed EA (2010) Equivalent accelerated life testing plans for log-location-scale distributions. Naval Research Logistics (NRL) 57(5):472–488
Lin YJ, Lio Y (2012) Bayesian inference under progressive Type-I interval censoring. J Appl Stat 39 (8):1811–1824
Ma H, Meeker WQ (2008) Optimum step-stress accelerated life test plans for log-location-scale distributions. Naval Research Logistics (NRL) 55(6):551–562
Marshall A, Meza JC, Olkin I (2001) Can data recognize its parent distribution? J Comput Graph Stat 10(3):555–580
McKane SW, Escobar LA, Meeker WQ (2005) Sample size and number of failure requirements for demonstration tests with log-location-scale distributions and failure censoring. Technometrics 47(2):182–190
Miner M (1945) Cumulative damage in fatigue. J Appl Mech 12:A159–A164
Morrison F, McCool J, Yonushonis T, Weinberg P (1984) The load-life relationship for m50 steel bearings with silicon nitride ceramic balls. Lubrication Eng 40(3):153–159
Nelson WB (2005a) Applied life data analysis, vol 577. Wiley, New York
Nelson WB (2005b) A bibliography of accelerated test plans part ii-references. IEEE Transactions on Reliability 54(3):370–373
Nelson WB (2009) Accelerated testing: statistical models, test plans, and data analysis, vol 344. Wiley, New York
Noh Y, Choi K, Lee I (2010) Identification of marginal and joint cdfs using Bayesian method for rbdo. Struct Multidiscip Optim 40(1-6):35
Nohut S, Lu C (2012) Fracture statistics of dental ceramics: discrimination of strength distributions. Ceram Int 38(6):4979–4990
Noughabi HA (2017) Goodness-of-fit tests for lifetime distributions based on Type II censored data. J Stat Comput Simul 87(9):1787–1798
Noughabi HA, Balakrishnan N (2014) Goodness of fit using a new estimate of Kullback-Leibler information based on Type II censored data. IEEE Trans Reliab 64(2):627–635
Ortiz-Yañez JF, Piña-Monarrez MR (2018) Discrimination between the lognormal and Weibull distributions by using multiple linear regression. Dyna 85(205):9–18
Paris P, Erdogan F (1963) A critical analysis of crack propagation laws. J Basic Eng 85(4):528–534
Posada D, Buckley TR (2004) Model selection and model averaging in phylogenetics: advantages of Akaike information criterion and Bayesian approaches over likelihood ratio tests. Systematic biology 53(5):793–808
Raqab MZ (2013) Discriminating between the generalized Rayleigh and Weibull distributions. J Appl Stat 40 (7):1480–1493
Schwarz G, et al. (1978) Estimating the dimension of a model. The Annals of Statistics 6(2):461–464
Tallian T (1962) Weibull distribution of rolling contact fatigue life and deviations therefrom. ASLE Transactions 5(1):183–196
Taylor JA, Jakeman AJ (1985) Identification of a distributional model. Communications in Statistics-Simulation and Computation 14(2):497–508
Thereska E, Abd-El-Malek M, Wylie JJ, Narayanan D, Ganger GR (2006) Informed data distribution selection in a self-predicting storage system. In: 2006 IEEE international conference on autonomic computing. IEEE, pp 187–198
Wang D, He J, Guan X, Yang J, Zhang W (2018) A model assessment method for predicting structural fatigue life using lamb waves. Ultrasonics 84:319–328
Welham S, Thompson R (1997) Likelihood ratio tests for fixed model terms using residual maximum likelihood. Journal of the Royal Statistical Society: Series B (Statistical Methodology) 59(3):701–714
Wiens BL (1999) When log-normal and gamma models give different results: a case study. The American Statistician 53(2):89–93
Xu HY, Fei HL (2007) Planning step-stress accelerated life tests with two experimental variables. IEEE Transactions on Reliability 56(3):569–579
Zhang Y, Wang H, Wang Z, Yang Y, Blaabjerg F (2017) Impact of lifetime model selections on the reliability prediction of igbt modules in modular multilevel converters. In: 2017 IEEE energy conversion congress and exposition (ECCE). IEEE, pp 4202–4207
Acknowledgments
The support is greatly acknowledged. The authors would like to thank the anonymous reviewers for their constructive comments.
Funding
The work in this study was supported by the Science Challenge Project, No.TZ2018007, and by National Natural Science Foundation of China, Nos. 51975546, 11872088.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The authors declare that they have no conflict of interest.
Replication of results
The calculations in this study are made using a MATLAB script-compatible software package GNU Octave version 5.1.0.
-
1.
Asymptotic Bayes factors
In the calculation of asymptotic Bayes factors, the modes of the logarithm of a distribution is found using function fminsearch1 with a relative error tolerance 10− 8 and an initial value obtained using the least square method. The censored data are treated as complete data in the least square method only for the purpose of getting an approximate initial value for fminsearch. Once the modes are found, the Hessian matrices evaluated at the modes are obtained using a numerical differentiation script suite called DERIVEST2 provided by John R. D’Errico, which runs directly in GNU Octave.
1https://octave.sourceforge.io/octave/function/fminsearch.html
-
2.
Confidence intervals and histograms
The calculation of confidence intervals of stress-life model predictions and σ histograms are made using MCMC samples drawn using the standard Metropolis-Hastings (MH) algorithm. The starting sample in the MH is set as the initial value used in fminsearch. The proposal density of the MH is the Gaussian distribution with a zero mean and a standard deviation set as half of the estimated standard deviation by the least square method mentioned above. A total of 5 × 106 samples are obtained for each cases.
Additional information
Responsible Editor: Felipe A. C. Viana
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
About this article
Cite this article
He, J., Chen, J. & Guan, X. Lifetime distribution selection for complete and censored multi-level testing data and its influence on probability of failure estimates. Struct Multidisc Optim 62, 1–17 (2020). https://doi.org/10.1007/s00158-019-02465-7
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00158-019-02465-7