高可靠度的時代為了提升產品競爭力,製造商必須即時地提供顧客有關產品可靠度的訊息(如平均失效時間)。唯高可靠度產品之壽命推論,即使利用加速壽命試驗, 亦很難在有限的測試時間內獲得產品失效資料。此時, 若存在與壽命相關之品質特徵值(quality characteristics, QC)且會隨時間逐漸衰變,則可藉由衰變路徑來推估產品壽命相關資訊。因此, 對於生產製造商而言,如何建構衰變模
型及推論產品壽命分配,將顯得格外重要。文獻上有關衰變模型之建構,大都以隨機效應(random effect) 或Wiener 過程之衰變模型來描述之。然而對於產品的衰變路徑為單調遞增,如金屬疲勞(metal fatigue), 此時若採用隨時間遞增之gamma過程來描述產品衰變路徑將更為合理。本論文主要針對gamma衰變模型探討以下三個研究主題:
(1) 此研究主題以laser產品為動機例子,主要探討當gamma衰變模型被誤判為Wiener衰變模型時,對於產品平均壽命估計值之準確度(accuracy)與精確度(precision)的影響。具體來說,本文導出在大樣本下,當模型發生誤判時,產品平均壽命估計量的漸近分配,藉此結果可探討模型誤判對於產品平均壽命之影響。結果顯示臨界值(critical value)與gamma衰變模型中的尺度(scale)參數之比值,會影響產品平均壽命估計值之準確度。而當gamma衰變模型中的形狀(shape)或尺度參數大時,模型誤判對於產品平均壽命估計值之精確度的影響是不容忽視的。此外, 在小樣本或較短測試時間下,本主題以模擬方式探討誤判對於產品平均壽命估計值之影響, 其模擬結果與
大樣本下的結論是差異不大的。
(2) 逐步應力加速衰變試驗(step-stress accelerated degradation test, SSADT)是推估高可靠度產品壽命之一可行方法, 其優點是以較少的測試樣本及試驗設備, 在較短的實驗時間內, 可迅速推估高可靠度產品的壽命分配資訊。本主題首先建構出gamma過程之SSADT衰變模型,並在試驗總成本不超過事先給定的預算下,尋找一最佳試驗計畫(optimal test plan)。換言之,極小化產品平均失效時間估計值之近似變異數,以獲得最佳樣本數、每一應力水準下的量測次數與量測頻率。藉由敏感度分析(sensitivity analysis)可知
最佳試驗計畫的表現是相當穩健的。此外,在節省成本考量下,可對SSADT之最佳化問題增加限制條件, 以解得次最佳(suboptimal) 試驗計畫。本主題亦將上一主題的模型誤判問題推廣到SSADT上,探討誤判對於產品MTTF之壽命推估的影響。結果顯示在SSADT下,誤判gamma衰變模型為Wiener衰變模型對於產品壽命估計精確度的影響是相當嚴重的。
(3) 預燒試驗(burn-in test)可有效地在產品運送到顧客手中之前,將產品製造過程中所產生的潛在失效與不良產品加以剔除。對於現今高可靠度產品,可收集與產品可靠度相關的QC,藉由收集到的衰變資料來進行篩選試驗。本主題假設產品之衰變路徑來自混合gamma 衰變模型, 並訂定如何區別正常產品與不良產品的準則,接著推導出每個時間點下的最佳切斷點(cutoff point), 再以極小化總成本函數得到最佳之預燒時間。除此之外, 本文亦探討混合gamma衰變模型被誤判為混合Wiener衰變模型之誤判問題,模擬結果顯示模型誤判對於預燒試驗的錯誤分類機率將有很大的影響。

Nowadays, degradation analysis has been widely used to assess the lifetime information of highly reliable products. In the literature, degradation models are
mostly built by the randomeffect models and/orWiener process. However, for materials that lead to fatigue data, it is more appropriately modeled by a gamma process
which exhibits a monotonic increasing pattern. Hence, it is of great interest to design and analyze the degradation tests based on gamma process. In this thesis, we study the following three topics.
(1) Motivated by laser data,we discuss the effects ofmodelmis-specificationwhen the truemodel comes fromgamma degradationprocess, and iswrongly treated as Wiener degradation process. We derive the asymptotic distribution of quasi MLE (QMLE) so that the penalty formodelmis-specification can be addressed sequentially. The result demonstrates that the effect on the accuracy of the product’smean-time-to failure (MTTF) prediction strongly depends on the ratio of critical value and scale parameter of gamma degradation process. Besides, the effects on the precision of the product’s MTTF prediction are critical
when the shape and scale parameters of gamma degradation process are large enough. Furthermore,we use a simulationstudy to address the penalty of model mis-specification. The simulation result is quite close to the theoretical result even when the sample size and termination time are not large enough.
(2) Step-stress accelerated degradation test (SSADT) is a useful tool for assessing the lifetime distribution of highly reliable products when the available test items are very few. Firstly, we propose the SSADT model based on a gamma process with a linear degradation path. Under the constraint that the total experimental cost not exceeding a pre-specified budget, the optimal test plan can be obtained by minimizing the approximate variance of the estimated MTTF
of the lifetime distribution of the product. The sensitivity analysis reveals that the optimal test plan is quite robust to moderate departures from the model
parameters. Furthermore,we address the effect ofmodelmis-specification under SSADT. The results show that the effect on the accuracy of product’s MTTF prediction is not so critical, while the effect on the precision of product’s MTTF prediction is substantial when gamma degradation process ismis-specified as Wiener degradation process.
(3) Burn-in test has been widely used to eliminate latent failures or weak components in the factory before the products are shipped to the customers. The traditional burn-in test over a short period of time to collect time-to-failure data is rather inefficient. This decision problem can be solved by degradation models. We first propose amixture gamma process to describe the degradation path of the product. Next, a decision rule for classifying a unit as a normal or a weak unit is proposed. Finally, a cost model is used to determine the optimal burn-in time. Furthermore,we address themis-specification issue of treating a mixture gamma degradation process as amixtureWiener degradation process by a simulation study. The results reveal that the effect on the probabilities of
misclassification is significant.

[1] 彭健育(2008). “高可靠度產品之衰變試驗分析,”國立清華大學統計學研究
所博士論文.
[2] Abdel-Hameed, M. (1975). “A gamma wear process,” IEEE Transactions
on Reliability, 24, 152–153.
[3] Bae, S. J., Kim, S. J., Kim, M. S., Lee, B. J. and Kang, C. W. (2008). “Degradation
analysis of nano-contamination in plasma display panels,” IEEE
Transactions on Reliability, 57, 222–229.
[4] Bae, S. J. and Kvam, P.H. (2004). “A nonlinear random-coefficientsmodel
for degradation testing,” Technometrics, 46, 460–469.
[5] Bae, S. J. and Kvam, P. H. (2006). “A change-point analysis for modeling
incomplete burn-in for light displays,” IIE Transactions, 38, 489–498.
[6] Bagdonavicius, V. and Nikulin, M. (2002). Accelerated Life Models : Modeling
and Statistical Analysis, Chapman & Hall/CRC, New York.
[7] Balka, J., Desmond, A. F. and McNicholas, P. D. (2009). “Review and implementation
of curemodels based on first hitting times forWiener processes,”
Lifetime Data Analysis, 15, 147–176.
[8] Berman, M. (1981). “Inhomogeneous and modulated gamma processes,”
Biometrika, 68, 143–152.
[9] Birnbaum, Z. W. and Saunders, S. C. (1969). “A new family of life distributions,” Journal of Applied Probability, 6, 319–327.
[10] Boulanger,M. and Escobar, L. A. (1994). “Experimental design for a class
of accelerated degradation tests,” Technometrics, 36, 260–272.
[11] Cha, J. H. (2006). “A stochastic model for burn-in procedures in accelerated
environment,” Naval Research logistics, 53, 226–234.
[12] Cha, J. H. and Mi, J. (2007). “Optimal burn-in procedure for periodically
inspected systems,” Naval Research logistics, 54, 720–731.
[13] Chang, D. S. (2000). “Optimal burn-in decision for products with an unimodel
failure rate function,” European Journal of Operational Research,
126, 534–540.
[14] Chao, M. T. (1999). “Degradation analysis and related topics: Some
thoughts and a review,” The Proceedings of the National Science Council.
A, 23, 555–566.
[15] Chen, Z. and Zheng, S. (2005). “Lifetime distribution based degradation
analysis,” IEEE Transactions on Reliability, 54, 3–10.
[16] Chhikara, R. S. and Folks, L. (1989). The Inverse Gaussain Distribution.
Theory,Methodology, and Applications,Marcel Dekker, New York.
[17] Chien, W. T. K. and Kuo, W. (1997). “A nonparametric Bayes approach to
decide system burn-in time,” Naval Research logistics, 44, 655–671.
[18] Corless, R.M., Gonnet, D. E. G., Jeffrey, D. J. and Knuth, D. E. (1996). “On
the Lambert W function,” Advances in Computational Mathematics, 5,
329–359.
[19] Crowder, M. and Lawless, J. F. (2007). “On a scheme for predictive maintenance,”
European Journal of Operational Research, 16, 1713–1722.
[20] Doksum, K. A. andH´oyland, A. (1992). “Model for variable-stress accelerated
life testing experiments based onWiener processes and the inverse
Gaussian distribution,” Technometrics, 34, 74–82.
[21] Doksum, K. A. and Normand, S.-L. T. (1995). “Gaussianmodels for degradation
processes - Part I: methods for the analysis of biomarker data,”
Lifetime Data Analysis, 1, 135–144.
[22] Horrocks, J. and Thompson, M. E. (2004). “Modeling event times with
multiple outcomes using the Wiener process with drift,” Lifetime Data
Analysis, 10, 29–49.
[23] Hsieh, M. H. and Jeng, S. L. (2007). “Accelerated discrete degradation
models for leakage current of ultra-thin gate oxides,” IEEE Transactions
on Reliability, 56, 369–380.
[24] Ishwaran, I. and James, L. F. (2004). “Computational methods for multiplicative
intensity models using weighted gamma processes: proportional
hazards, marked point processes, and panel count data,” Journal
of the American Statistical Association, 99, 175–190.
[25] Jensen, F. and Petersen, N. E. (1982). Burn-in: An Engineering Approach
to The Design and Analysis of Burn-in Procedures, John Wiley & Sons,
New York.
[26] Kalbfleisch, J. D. and Prentice, R. I. (2002). The Statistical Analysis of Failure
Time Data, John Wiley & Sons, New York.
[27] Kim, K. O. and Kuo, W. (2003a). “A general model for heterogeneous system
lifetime and conditions for system burn-in,” Naval Research Logistics,
50, 364–380.
[28] Kim, K. O. and Kuo,W. (2003b). “Percentile residual life and system reliability as performancemeasures in the optimal system design,” IIE Transactions,
35, 1133–1142.
[29] Kuo,W. (1984). “Reliability enhancement through optimal burn-in,” IEEE
Transactions on Reliability, 3, 145–156.
[30] Kuo, W. and Kuo, Y. (1983). “Facing the headaches of early failures: a
state-of-the-art review of burn-in decisions,” Proceedings of the IEEE, 71,
1257–1266.
[31] Lawless, J. F. (2002). Statistical Models and Methods for Lifetime Data,
JohnWiley & Sons, New York.
[32] Lawless, J. F. and Crowder, M. (2004). “Covariates and random effects in
a gamma process model with application to degradation and failure,”
Lifetime Data Analysis, 10, 213–227.
[33] Lawrence, M. J. (1966). “An investigation of the burn-in problem,” Technometrics,
8, 61–71.
[34] Lee, M. -L. T. and Whitmore, G. A. (1993). “Stochastic processes directed
by randomized time,” Journal of Applied Probability, 30, 302–314.
[38] Leemis, L. M. and Beneke, M. (1990). “Burn-in models and methods: a
review,” IIE Transactions, 22, 172–180.
[36] Li, Q. and Kececioglu, D. B. (2004). “Optimal design of accelerated degradation
tests,” International Journal ofMaterials and Product Technology,
20, 73–90.
[37] Liao, C. M. and Tseng, S. T. (2006). “Optimal design for step-stress accelerated
degradation tests,” IEEE Transactions on Reliability, 55, 59–66.
[38] Lindstrom, M. J. and Bates, D. M. (1990). “Nonlinear mixed effects models for repeatedmeasures data,” Biometrics, 46, 673–687.
[39] Lu, J. C. and Meeker, W. Q. (1993). “Using degradation measures to estimate
a time-to-failure distribution,” Technometrics, 35, 161–173.
[40] Mahmoud, M. A. W. and Moustafa, H. M. (1993). “Estimation of a discriminant
function fromamixture of two gamma distributions when the
sample size is small,”Mathematical and computermodelling, 18, 87–95.
[41] Meeker,W. Q. and Escobar, L. A. (1998). StatisticalMethods for Reliability
Data, JohnWiley & Sons, New York.
[42] Meeker, W. Q., Escobar, L. A. and Lu, C. J. (1998). “Accelerated degradation
tests: modeling and analysis,” Technometrics, 40, 89–99.
[43] Moran, P. A. P. (1954). “A probability theory of dams and storage systems,”
Australian Journal of Applied Science, 5, 116–124.
[44] Nelson, W. (1990). Accelerated Testing: Statistical Models, Test Plans, and
Data Analysis, John Wiley & Sons, New York.
[45] Nguyen, D. G. andMurthy, D. N. P. (1982). “Optimal burn-in time tominimize
cost for products sold under warranty,” IIE Transactions, 14, 167–
174.
[46] Noortwijk, J. M. V. (2009). “A survey of the application of gamma processes
in maintenance,” Reliability Engineering & System Safety, 94, 2–
21.
[47] Padgett, W. J. and Tomlinson, M. A. (2004). “Inference from accelerated
degradation and failure data based on Gaussian process models,” Lifetime
Data Analysis, 10, 191–206.
[48] Park, C. and Padgett, W. J. (2005). “Accelerated degradation models for failure based on geometric Brownianmotion and gamma process,” Lifetime
Data Analysis, 11, 511–527.
[49] Park, C. and Padgett, W. J. (2006). “Stochastic degradation models with
several accelerating variables,” IEEE Transactions on Reliability, 55, 379–
390.
[50] Park, J. I. and Yum, B. J. (1997). “Optimal design of accelerated degradation
tests for estimatingmean lifetime at the use condition,” Engineering
Optimization, 28, 199–230.
[51] Park, S. J., Yum, B. J. and Balamurali, S. (2004). “Optimal design of stepstress
degradation tests in the case of destructivemeasurement,” Quality
Technology & Quantitative Management, 1, 105–124.
[52] Pascual, F. G. (2005). “Maximum likelihood estimation under misspecified
lognormal and weibull distributions,” Communications in Statistics
– Simulation and Computation, 34, 503–524.
[53] Pascual, F. G. (2006). “Accelerated life test plans robust to misspecification
of the stress-life relationship,” Technometrics, 48, 11–25.
[54] Pascual, F. G. andMontepiedra, G. (2003). “Model-robust test plans with
applications in accelerated life testing,” Technometrics, 45, 47–57.
[55] Pascual, F.G. andMontepiedra, G. (2005). “Lognormal andWeibull accelerated
life test plans under distributionmisspecification,” IEEE Transactions
on Reliability, 54, 43–52.
[56] Peng, C. Y. and Tseng, S. T. (2009). “Mis-specification analysis of linear
degradationmodels,” IEEE Transactions on Reliability, To appear.
[57] Seshadri, V. (1999). Inverse Gaussian Distribution: Statistical Theory and Applications, Springer-Verlag, New York.
[58] Shiau, H. J. J. and Lin, H. H. (1999). “Analyzing accelerated degradation
data by nonparametric regression,” IEEE Transactions on Reliability, 48,
149–158.
[59] Singpurwalla, N. D. (1995). “Survival in dynamic environments,” Statistical
Science, 10, 86–103.
[60] Tseng, S. T., Balakrishnan, N. and Tsai, C. C. (2009). “Optimal step-stress
accelerated degradation test plan for gamma degradation process,” IEEE
Transactions on Reliability, To appear.
[61] Tseng, S. T., Hamada, M. S. and Chiao, C. H. (1995). “Using degradation
data from a fractional factorial experiment to improve fluorescent lamp
reliability,” Journal of Quality Technology, 27, 363–369.
[62] Tseng, S. T. and Liao, C. M. (1998). “Optimal design for a degradation
test,” International Journal ofOperations andQuantitativeManagement,
4, 293–301.
[63] Tseng, S. T. and Peng, C. Y. (2004). “Optimal burn-in policy by using an
integratedWiener process,” IIE Transactions, 36, 1161–1170.
[64] Tseng, S. T. and Peng, C. Y. (2007). “Stochastic diffusion modeling of
degradation data,” Journal of Data Science, 5, 315–333.
[65] Tseng, S. T. and Tang, J. (2001). “Optimal burn-in time for highly reliable
products,” International Journal of Industrial Engineering, 8, 329–338.
[66] Tseng, S. T., Tang, J. and Ku, I. H. (2003). “Determination of burn-in parameters
and residual life for highly reliable products,” Naval Research
Logistics, 50, 1–14.
[67] Tseng, S. T. and Wen, Z. C. (2000). “Step-stress accelerated degradation
analysis of highly-reliable products,” Journal of Quality Technology, 32,
209–216.
[68] Tseng, S. T. and Yu, H. F. (1997). “A termination rule for degradation experiments,”
IEEE Transactions on Reliability, 46, 130–133.
[69] Venturini, S.,Dominici, F. and Parmigiani,G. (2008). “Gamma shapemixtures
for heavy-tailed distributions,” The Annals of Applied Statistics, 2,
756–776.
[70] Wang, X. (2008). “A pseudo-likelihood estimationmethod for nonhomogeneous
gamma process model with random effects,” Statistica Sinica,
18, 1153–1163.
[71] Watson, G. S. and Wells, W. T. (1961). “On probability of improving the
mean useful life of items by eliminating thosewith short lives,” Technomterics,
3, 281–298.
[72] White, H. (1982). “Maximumlikelihood estimation ofmisspecifiedmodels,”
Econometrica, 50, 1–25.
[73] Whitmore, G. A. (1995). “Estimating degradation by a Wiener diffusion
process subject to measurement error,” Lifetime Data Analysis, 1, 307–
319.
[74] Whitmore, G. A., Crowder, M. I. and Lawless, J. F. (1998). “Failure inference
from a marker process based on a bivariate model,” Lifetime Data
Analysis, 4, 229–251.
[75] Whitmore, G. A. and Schenkelberg, F. (1997). “Modelling accelerated
degradation data using Wiener diffusion with a time scale transformation,”
Lifetime Data Analysis, 3, 27–45.
[76] Wu, C. J. and Chang, C. T. (2002). “Optimal design of degradation tests in
presence of cost constraint,” Reliability Engineering & System Safety, 76,
109–115.
[77] Wu, S. and Xie,M. (2007). “Classifying weak, and strong components using
ROC analysis with application to burn-in,” IEEE Transactions on Reliability,
56, 552–561.
[78] Yu, H. F. (2002). “Optimal selection of the most reliable product with
degradation data,” Engineering Optimization, 34, 579-590.
[79] Yu, H. F. (2003). “Optimal selection of the most reliable product whose
degradation path satisfies Wiener process,” International Journal of
Quality and Reliability Management, 20, 1084-1095.
[80] Yu, H. F. (2006). “Designing an accelerated degradation experiment with
a reciprocal Weibull degradation rate,” Journal of Statistical Planning
and Inference, 136, 282–297.
[81] Yu, H. F. (2007a). “Mis-specification analysis between normal and extreme
value distribution for a linear regressionmodel,” Communications
in Statistics-Theory andMethods, 36, 499–521.
[82] Yu, H. F. (2007b). “Designing a screening experiment with reciprocal
weibull degradation rates,” Computers&Industrial Engineering, 52, 175–
191.
[83] Yu, H. F. and Chiao, C. H. (2002). “An optimal designed degradation experiment
for reliability improvement,” IEEE Transactions on Reliability,
51, 427–433.
[84] Yu, H. F. and Tseng, S. T. (1998). “On-line procedure for determining an appropriate termination time for an accelerated degradation experiment,”
Statistica Sinica, 8, 207–220.
[85] Yu, H. F. and Tseng, S. T. (1999). “Designing a degradation experiment,”
Naval Research logistics, 46, 689–706.
[86] Yu, H. F. and Tseng, S. T. (2002). “Designing a screening degradation experiment,”
Naval Research Logistics, 49, 514–526.
[87] Yu, H. F. and Tseng, S. T. (2004). “Designing a degradation experiment
with a reciprocal Weibull degradation rate,” Quality Technology and
Quantitative Management, 1, 47–63.