在高可靠度電子元件製造中，預燒測試 (Burn-in Test) 是用來篩選出早期失效品並確保產品品質的關鍵製程。然而，現行預燒策略多依賴統計衰退模型推估最佳測試時長，當失效樣本稀缺且衰退行為複雜時，容易因統計分布假設偏差而影響停止決策，造成測試成本增加與資源浪費。

為解決上述問題，本研究提出一套結合深度時間序列學習與成本導向決策之資料驅動 (Data-driven) 預燒終止決策架構。針對砷化鎵 (GaAs) 雷射元件之電流衰退資料，本研究建立時間序列分類模型，使模型能夠根據不同測試時間下已觀測到的衰退資訊，逐步判斷產品狀態，模擬實際預燒過程中的終止決策。此外，考量預燒資料通常具有樣本數有限與不良品比例偏低的特性，本研究透過資料擴增方法增加訓練資料的多樣性，降低模型偏向多數類別的情形，並提升其在有限資料下的穩定性。在決策層面，結合成本模型並進行成本軌跡穩定化處理，同時考量測試運行成本與誤判風險損失，建立最佳終止決策機制。

實驗結果顯示，所提出之方法能有效提升早期不良品辨識能力，並在維持產品可靠度的同時，有效降低整體預燒成本。相較於傳統統計方法與既有深度學習方法，本研究提出之架構可使預燒停止時間縮短約六成，並降低整體預燒成本。此外，研究結果亦顯示，當統計分布假設存在偏差時，容易造成分類效能下降，而本研究所提出之架構因不依賴特定統計分布假設，因此在不同衰退情境下具備較佳之決策穩健性與實務應用潛力。

Burn-in test is a critical quality assurance process in high-reliability electronic manufacturing. However, conventional burn-in strategies often rely on statistical degradation models, which may lead to unreliable stopping decisions when defective samples are scarce and degradation behaviors are complex. To address this issue, this study proposes a data-driven and cost-driven framework for estimating the optimal stopping time in burn-in tests. A time-aware GRU model is developed to learn degradation patterns from partially observed sequences, and prefix slicing is used to simulate sequential decisions. SMOTE, transformation-based augmentation, and TimeGAN are applied to mitigate small-sample and class-imbalance problems. The prediction results are integrated with a cost model, while Savitzky--Golay smoothing and PAV-based slope stabilization are used to obtain a stable cost trajectory. Experimental results show that the proposed framework improves early defective-product identification and reduces total burn-in cost. Compared with statistical and CNN-based methods, the proposed approach reduces the stopping time to 757.5 hours and the total cost to 165.25.

Ayer, M., Brunk, H. D., Ewing, G. M., Reid, W. T., and Silverman, E. (1955). An empirical distribution function for sampling with incomplete information. The annals of mathematical statistics, pages 641–647.

Block, H. W. and Savits, T. H. (1997). Burn-in. Statistical Science, 12(1):1–19.

Cha, J. H. (2011). A survey of burn-in and maintenance models for repairable systems. In Replacement Models with Minimal Repair, pages 179–203. Springer.

Cha, J. H. and Finkelstein, M. (2010). Stochastically ordered subpopulations and optimal burn-in procedure. IEEE Transactions on Reliability, 59(4):635–643.

Chawla, N. V., Bowyer, K. W., Hall, L. O., and Kegelmeyer, W. P. (2002). Smote: synthetic minority over-sampling technique. Journal of artificial intelligence research, 16:321–357.

Choi, K., Yi, J., Park, C., and Yoon, S. (2021). Deep learning for anomaly detection in time-series data: Review, analysis, and guidelines. IEEE access, 9:120043–120065.

Chung, J., Gulcehre, C., Cho, K., and Bengio, Y. (2014). Empirical evaluation of gated recurrent neural networks on sequence modeling. arXiv preprint arXiv:1412.3555.

De Leeuw, J., Hornik, K., and Mair, P. (2010). Isotone optimization in r: pool-adjacent-violators algorithm (pava) and active set methods. Journal of statistical software, 32:1–24.

Feng, Q., Peng, H., and Coit, D. W. (2010). A degradation-based model for joint optimization of burn-in, quality inspection, and maintenance: a light display device application. The International Journal of Advanced Manufacturing Technology, 50(5):801–808.

Foumani, N. M., Tan, C. W., Webb, G. I., and Salehi, M. (2024). Improving position encoding of transformers for multivariate time series classification. Data mining and knowledge discovery, 38(1):22–48.

Gebraeel, N. (2006). Sensory-updated residual life distributions for components with exponential degradation patterns. IEEE Transactions on Automation Science and Engineering, 3(4):382–393.

Goodfellow, I. J., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., Courville, A., and Bengio, Y. (2014). Generative adversarial nets. Advances in neural information processing systems, 27.

Hochreiter, S. and Schmidhuber, J. (1997). Long short-term memory. Neural computation, 9(8):1735–1780.

Iqbal, A. and Amin, R. (2024). Time series forecasting and anomaly detection using deep learning. Computers & Chemical Engineering, 182:108560.

Iwana, B. K. and Uchida, S. (2021). An empirical survey of data augmentation for time series classification with neural networks. Plos one, 16(7):e0254841.

Janssens, O., Slavkovikj, V., Vervisch, B., Stockman, K., Loccufier, M., Verstockt, S., Van de Walle, R., and Van Hoecke, S. (2016). Convolutional neural network based fault detection for rotating machinery. Journal of Sound and Vibration, 377:331–345.

Jensen, F. (1982). Burn-in: an engineering approach to the Design and Analysis of Burn-in Procedures. New York: Wiley.

Kaiyuan, H. and Na, L. (2023). Improved timegan based on attention for time series prediction method with few shot. Journal of East China University of Science and Technology, 49(6):890–899.

Kazemi, S. M., Goel, R., Eghbali, S., Ramanan, J., Sahota, J., Thakur, S., Wu, S., Smyth, C., Poupart, P., and Brubaker, M. (2019). Time2vec: Learning a vector representation of time. arXiv preprint arXiv:1907.05321.

Kuo, W. (1984). Reliability enhancement through optimal burn-in. IEEE Transactions on Reliability, R-33(2):145–156.

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(11):1257–1266.

Ling, M. H., Tsui, K. L., and Balakrishnan, N. (2014). Accelerated degradation analysis for the quality of a system based on the gamma process. IEEE Transactions on Reliability, 64(1):463–472.

Lipton, Z. C., Berkowitz, J., and Elkan, C. (2015). A critical review of recurrent neural networks for sequence learning. arXiv preprint arXiv:1506.00019.

Lyu, Y., Gao, J., Chen, C., Jiang, Y., Li, H., Chen, K., and Zhang, Y. (2019). Optimal burn-in strategy for high reliable products using convolutional neural network. IEEE Access, 7:178511–178521.

Markovic, T., Dehlaghi-Ghadim, A., Leon, M., Balador, A., and Punnekkat, S. (2023). Time-series anomaly detection and classification with long short-term memory network on industrial manufacturing systems. In 2023 18th Conference on Computer Science and Intelligence Systems (FedCSIS), pages 171–181. IEEE.

Meeker, W. Q., Escobar, L. A., and Pascual, F. G. (1998). Statistical methods for reliability data. John Wiley & Sons.

Pascanu, R., Mikolov, T., and Bengio, Y. (2013). On the difficulty of training recurrent neural networks. In International conference on machine learning, pages 1310–1318. Pmlr.

Safaei, F. and Taghipour, S. (2024). Integrated degradation-based burn-in and maintenance model for heterogeneous and highly reliable items. Reliability Engineering & System Safety, 244:109942.

Savitzky, A. and Golay, M. J. (1964). Smoothing and differentiation of data by simplified least squares procedures. Analytical chemistry, 36(8):1627–1639.

Schafer, R. W. (2011). What is a savitzky-golay filter? IEEE Signal processing magazine, 28(4):111–117.

Shao, S., Wang, P., and Yan, R. (2019). Generative adversarial networks for data augmentation in machine fault diagnosis. Computers in Industry, 106:85–93.

Sheu, S.-H. and Chien, Y.-H. (2004). Minimizing cost-functions related to both burn-in and field-operation under a generalized model. IEEE Transactions on Reliability, 53(3):435–439.

Sheu, S.-H. and Chien, Y.-H. (2005). Optimal burn-in time to minimize the cost for general repairable products sold under warranty. European Journal of Operational Research, 163(2):445–461.

Shi, Y., Xiang, Y., Liao, Y., Zhu, Z., and Hong, Y. (2020). Optimal burn-in policies for multiple dependent degradation processes. IISE Transactions, 53(11):1281–1293.

Siewiorek, D. P. and Swarz, R. S. (1982). The theory and practice of reliable system design.

Tsai, C.-C., Tseng, S.-T., and Balakrishnan, N. (2011). Optimal burn-in policy for highly reliable products using gamma degradation process. IEEE Transactions on Reliability, 60(1):234–245.

Tseng, S.-T. and Peng, C.-Y. (2004). Optimal burn-in policy by using an integrated wiener process. Iie Transactions, 36(12):1161–1170.

Um, T. T., Pfister, F. M., Pichler, D., Endo, S., Lang, M., Hirche, S., Fietzek, U., and Kulić, D. (2017). Data augmentation of wearable sensor data for parkinson’s disease monitoring using convolutional neural networks. In Proceedings of the 19th ACM international conference on multimodal interaction, pages 216–220.

Wen, Q., Sun, L., Yang, F., Song, X., Gao, J., Wang, X., and Xu, H. (2020). Time series data augmentation for deep learning: A survey. arXiv preprint arXiv:2002.12478.

Yoon, J., Jarrett, D., and Van der Schaar, M. (2019). Time-series generative adversarial networks. Advances in neural information processing systems, 32.

Zakaria, M. F., Kassim, Z. A., Ooi, M.-L., and Demidenko, S. (2006). Reducing burn-in time through high-voltage stress test and weibull statistical analysis. IEEE Design & Test of computers, 23(2):88–98.

Zamanzadeh Darban, Z., Webb, G. I., Pan, S., Aggarwal, C., and Salehi, M. (2024). Deep learning for time series anomaly detection: A survey. ACM Computing Surveys, 57(1):1–42.

Zhang, C., Song, D., Chen, Y., Feng, X., Lumezanu, C., Cheng, W., Ni, J., Zong, B., Chen, H., and Chawla, N. V. (2019). A deep neural network for unsupervised anomaly detection and diagnosis in multivariate time series data. In Proceedings of the AAAI conference on artificial intelligence, volume 33, pages 1409–1416.