偏常態量測誤差模型下之加速破壞衰變試驗

Accelerated Destructive Degradation Test based on Skew-Normal Measurement Error Model

林姿吟(Tzu-Yin Lin)

指導教授 ： 蔡志群

淡江大學/理學院/數學學系碩士班/碩士(2013年)

https://doi.org/10.6846/TKU.2013.00604

摘要

加速破壞衰變試驗，其量測過程中，需破壞測試樣本，以便量測到產品的品質特徵值，並配合提高環境應力，以加速產品衰變，進而有效地推估產品的壽命資訊。Tsai et al. (2013) 建構一量測誤差服從常態分配的非線性 ADDT 衰變模型。然而，當量測誤差為非對稱的分配時，此時使用偏常態分配來描述較為合適。因此，本文以聚合物材料為動機例子，首先建構一偏常態 ADDT 衰變模型，並探討其壽命資訊，其結果顯示使用本文所建構的衰變模型，來推估產品壽命較為精確。接下來，模擬一組偏常態 ADDT 衰變資料，並執行最佳化設計，其結果顯示在不同預算下，偏常態 ADDT 衰變模型的近似變異數皆較小。最後，模擬分析結果可知，模擬結果與理論結果是相近的，而模型誤判對於產品壽命的推估影響很大。

關鍵字

加速破壞衰變試驗 ； 偏常態分配 ； 最佳試驗配置

並列摘要

Accelerated destructive degradation tests (ADDTs) that measurement process of quality characteristic (QC) would destroy the tested units at higher stress environment are powerful and useful tools for lifetime assessment of highly reliable products. Motivated by a polymer data, Tsai et al. (2013) proposed a nonlinear ADDT model with measurement error that follows a normal distribution. However, the skew normal distribution that generalizes the normal distribution to allow for non-zero skewness is more appropriate for describing the measurement error with non-symmetrical pattern. Hence, this article used skew-normal distribution to construct the nonlinear ADDT model. The results show that the skew normal ADDT model has better precision on the estimated lifetime of the products than normal ADDT model. Moreover, the optimal design problem based on the proposed ADDT model is discussed on this article. Finally, Monte Carlo simulations are used to evaluate the asymptotical results and the effect of model misspecification on the estimator of the products’ lifetime.

並列關鍵字

accelerated destructive degradation test ； skew-normal distribution ； optimal test plan

參考文獻

[2] Boulanger, M. and Escobar, L. A. (1994). “Experimental design for a class of accelerated degradation tests,” Technometrics, vol. 36, 260-272.

[3] Carey, M. B. and Koenig, R. H. (1991). “Reliability assessment based on accelerated degradation,” IEEE Transactions on Reliability, vol. 40, 499-506.

[5] Gomez, H. W. and Salinas, H. S. (2010). “Information matrix for generalized skew-normal distributions,” Proyecciones Journal of Mathematics, vol. 29, 83-92.

[6] Gupta, R. C. and Brown, N. (2001). “Reliabilitystudies of the skew-normal distribution and its application to a strengthstress model,” Communications in Statistics - Theory and Methods, vol. 30, 2427-2445.

[7] Henze, N. (1986). “A probabilistic representation of the skew-normal distribution,” Scandinavian Journal of Statistics, vol. 13, 271-275.