由於現今多數產品的壽命時間較長，使得在開發短期內評估產品可靠度及訂定保固期等任務變得較為艱難，為了解決此問題，加速壽命試驗 (accelerated life tests, ALT) 被廣泛應用於產品可靠度研究中。由於實驗成本和時間因素的考量，如何設計一既有效率又能準確地推估產品在正常使用條件下壽命的試驗計畫是一重要課題。過去已有許多文獻針對一因子及二因子加速壽命模型提出找尋最佳試驗計畫之方法。本研究將延伸討論當兩個因子具有交互作用項時，如何找尋最佳試驗計畫。首先透過粒子群演算法 (particle swarm optimization, PSO) 來尋找最佳試驗計畫，並輔以一般等價定理 (general equivalence theorem, GET) 驗證其為全域最佳解。本研究亦將文獻上提出之試驗配置策略與最佳試驗計畫比較，其中二因子不具交互作用項之加速模型找尋最佳試驗計畫的方法，結果顯示由降維計畫推廣至非降維計畫此方法於具有交互作用項之模型下不可行。而透過演算法找尋之最佳試驗計畫和傳統的2^2、3^2因子設計進行比較，結果顯示使用PSO找到的試驗計畫更有效率。最後應用依序設計，即逐點依序決定測試樣本之實驗應力水準，結果顯示和使用PSO找到的最佳試驗計畫有相似結論。

Due to the long lifespan of most products nowadays, it is more difficult to evaluate product reliability and set warranty period in a short period of development. In order to solve this problem, accelerated life tests (ALT) are widely used on product reliability research. Due to the consideration of experimental cost and time factors, how to design a test plan that is both efficient and can accurately estimate the lifetime of the product under normal use conditions is an important topic. In the past, many literatures have proposed methods to find the optimum test plan for one-factor and two-factor accelerated life models. This study will extend the discussion on how to find the optimum test plan when the model considers the interaction term of two factors. To determine the optimum plan, the particle swarm optimization (PSO) is used to find the optimum test plan, and the general equivalence theorem (GET) is supplemented to verify that it is the global best solution. This study also compares the method proposed in the literature with the optimum test plan. The results show that the extension of the degenerate plan to the nondegenerate plan is not feasible under the model with interaction terms. The optimum test plan is also compared with the traditional 2^2 & 3^2 factorial design, and the results show that the test plan found using PSO is more efficient. Finally, a sequential design is applied, that is, the experimental stress level of the test unit is determined point by point, and the result shows that there is a similar conclusion to the optimum test plan found using PSO.

[1] Atkinson, A. C., and Donev, A. N. Optimum Experimental Designs. 1992.
[2] Chaloner, K., and Larntz, K. Optimal bayesian design applied to logistic regression experiments. Journal of Statistical Planning and Inference 21, 2 (1989), 191–208.
[3] Chen, P.-Y. GitHub - globpso. https://github.com/PingYangChen/globpso, 2018.
[4] Dubov, E. D-optimal designs for non-linear models under the bayesian approach. Regression Experiments (1977), 103–111.
[5] Escobar, L. A., and Meeker, W. Q. Planning accelerated life tests with two or more experimental factors. Technometrics 37, 4 (1995), 411–427.
[6] Hakamipour, N. Optimal plan for step-stress accelerated life test with two stress variables for lognormal data. Iranian Journal of Science and Technology, Transactions A: Science 42, 4 (2018), 2259–2271.
[7] Hakamipour, N. Approximated optimal design for a bivariate step-stress accelerated life test with generalized exponential distribution under type-i progressive censoring. International Journal of Quality & Reliability Management (2020).
[8] Hakamipour, N., and Rezaei, S. Optimal design for a bivariate simple step-stress accelerated life testing model with type-ii censoring and gompertz distribution. International Journal of Information Technology & Decision Making 14, 06 (2015), 1243–1262.
[9] Hong, Y., Ma, H., and Meeker, W. Q. A tool for evaluating time-varying-stress accelerated life test plans with log-location-scale distributions. IEEE Transactions on Reliability 59, 4 (2010), 620–627.
[10] Hu, I. On sequential designs in nonlinear problems. Biometrika 85, 2 (1998), 496–503.
[11] Kennedy, J., and Eberhart, R. Particle swarm optimization. In Proceedings of ICNN’95-international conference on neural networks (1995), vol. 4, IEEE, pp. 1942–1948.
[12] Meeker, W., and Escobar, L. Statistical Methods for Reliability Data. Wiley Series in Probability and Statistics. Wiley, 1998.
[13] Nelson, W. Accelerated Testing: Statistical Models, Test Plans, and Data Analysis. Wiley Series in Probability and Statistics. Wiley, 1990.
[14] Pascual, F., Meeker, W. Q., and Escobar, L. A. Accelerated life test models and data analysis. Handbook of Engineering Statistics (H. Pham, ed.) Chapter 22 (2006).
[15] Pukelsheim, F., and Rieder, S. Efficient rounding of approximate designs. Biometrika 79, 4 (1992), 763–770.
[16] Shi, Y., and Eberhart, R. A modified particle swarm optimizer. In 1998 IEEE international conference on evolutionary computation proceedings. IEEE world congress on computational intelligence (Cat. No. 98TH8360) (1998), IEEE, pp. 69–73.
[17] Sun, J., Wu, X., Palade, V., Fang, W., Lai, C.-H., and Xu, W. Convergence analysis and improvements of quantum-behaved particle swarm optimization. Information Sciences 193 (2012), 81–103.
[18] Whittle, P. Some general points in the theory of optimal experimental design. Journal of the Royal Statistical Society: Series B (Methodological) 35, 1 (1973), 123–130.
[19] Xu, H., Fei, H., and Chen, Z. Design of bivariate accelerated life tests with one main effect and one interaction effect for log-location-scale distributions. Communications in Statistics - Theory and Methods 40, 15 (2011), 2781–2798.
[20] Xu, X., and Krzeminski, M. Optimal designs for accelerated life tests with multiple stress factors and heteroscedasticity. Communications in Statistics - Theory and Methods 47, 6 (2018), 1273–1306.
[21] Zhang, Y., and Meeker, W. Q. Bayesian methods for planning accelerated life tests. Technometrics 48, 1 (2006), 49–60.