在半導體封裝製程中，由於產品規格、材料特性與多階段製程條件之間的複雜相互作用，導致翹曲（Warpage）問題難以預測與控制。翹曲會影響晶片可靠性與良率，成為製程品質中的重要挑戰。儘管傳統基於物理的模型可進行翹曲模擬，其難以捕捉高階非線性行為，且建模與參數設定成本高昂；統計迴歸與機器學習方法則受限於實際產線中樣本數不足的限制，難以進行有效預測。因此，在有限資料情境下，發展一套具高準確性且可擴展的翹曲預測系統，對於加速封裝開發與製程優化具關鍵意義。
本研究提出一個整合式深度學習預測與最佳化框架，該系統結合基因演算法與人工神經網路（Genetic Algorithm Optimized Artificial Neural Network, GA-NN），藉由演化式搜尋方式同時進行模型超參數與輸入特徵的優化，以提升模型的預測準確性與訓練效率。為強化模型在實務應用中的透明度與可理解性，本研究引入 SHapley Additive exPlanations (SHAP) 方法，以解析各輸入特徵對翹曲預測結果的貢獻度，進而協助產線工程師進行參數調控與設計改善。此外，為克服樣本不足的問題，我們運用條件式表格生成對抗網路（Conditional Tabular GAN, CTGAN）進行資料擴增，並透過資料合成後的回饋機制，重新導入至優化模型中，進一步強化參數搜尋的效果與模型的泛化能力。
本研究以實際先進封裝案例進行驗證，實驗結果顯示所提出之方法能將翹曲預測誤差控制於 4% 以內，相較傳統方法具備更高的準確率、更快的收斂速度，並能透過 SHAP 分析視覺化呈現產品結構參數與翹曲行為之間的內在關聯，提供決策支持的依據。整體而言，本研究不僅提出一套具可解釋性與資料擴充能力之智慧預測系統，更為半導體製程中的翹曲控制提供一條有效且具應用潛力的技術路徑。

In semiconductor packaging processes, warpage remains difficult to predict and control due to the complex interactions among product specifications, material properties, and multi-stage process conditions. Warpage can severely impact chip reliability and production yield, posing a critical challenge to manufacturing quality. Although traditional physics-based methods can simulate warpage behavior, it often fails to capture higher-order nonlinear effects and entails high costs in modeling and parameter calibration. On the other hand, statistical regression and machine learning approaches are constrained by the limited size of available datasets in real manufacturing settings, making accurate prediction impractical. Therefore, under data-scarce conditions, developing a scalable and accurate warpage prediction system is essential for accelerating packaging development and process optimization.
This study proposes an integrated deep learning-based prediction and optimization framework that combines Genetic Algorithms and Artificial Neural Networks (GA-NN). By employing evolutionary search strategies, the framework simultaneously optimizes model hyperparameters and input features to enhance prediction accuracy and training efficiency. To improve the model’s interpretability in practical applications, we adopt SHapley Additive exPlanations (SHAP) to evaluate the impact of each input feature on the warpage prediction results. This provides engineering insights for process parameter tuning and design improvement. In addition, to address the challenge of limited data, we adopt the Conditional Tabular Generative Adversarial Network (CTGAN) for synthetic data augmentation. A feedback mechanism is implemented wherein the generated data are reintegrated into the optimization loop, further enhancing parameter search performance and the generalization ability of the model.
The proposed approach is validated using real-world advanced packaging cases. Experimental results demonstrate that prediction error can be maintained within 4%, significantly outperforming conventional methods in terms of prediction accuracy and convergence speed. Moreover, the SHAP-based analysis facilitates intuitive visualization of the relationships between structural parameters and warpage behavior, supporting data-driven decision-making. Overall, this research presents an interpretable and data-augmented intelligent prediction system, offering a promising solution for improving warpage prediction accuracy, shortening design cycles, and enhancing manufacturing quality in semiconductor packaging.

Arunadevi, M., Veeresha, G., Kharche, A. W., Suryawanshi, V. P., Sollapur, S. B., Mhatre, M. S., Kapadani, K. R., & Nalawade, D. (2024). Enhancing surface quality of metal parts manufactured via LPBF: ANN classifier and bayesian learning approach. International Journal of Interactive Design and Manufacturing - Ijidem, 18(6), 4093-4101.
Baek, J. W., Yang, W. S., Hur, M. J., Yun, J. C., & Park, S. J. (2019). Representative volume element analysis for wafer-level warpage using Finite Element methods. Materials Science in Semiconductor Processing, 91, 392-398.
Batool, U., Shapiai, M. I., Tahir, M., Ismail, Z. H., Zakaria, N. J., & Elfakharany, A. (2021). A Systematic Review of Deep Learning for Silicon Wafer Defect Recognition. IEEE Access, 9, 116572-116593.
Bin Hameed, M. A., & Alamgir, Z. (2022). Improving mortality prediction in Acute Pancreatitis by machine learning and data augmentation. Computers in Biology and Medicine, 150, Article 106077.
Chen, C., Yu, D. Q., Wang, T., Xiao, Z. Y., & Wan, L. X. (2019). Warpage Prediction and Optimization for Embedded Silicon Fan-Out Wafer-Level Packaging Based on an Extended Theoretical Model. IEEE Transactions on Components Packaging and Manufacturing Technology, 9(5), 845-853.
Chen, T., & Wang, Y. C. (2022). A two-stage explainable artificial intelligence approach for classification-based job cycle time prediction. International Journal of Advanced Manufacturing Technology, 123(5-6), 2031-2042.
Chen, Z. W., Zhang, Z., Dong, F., Liu, S., & Liu, L. (2021). A Hybrid Finite Element Modeling: Artificial Neural Network Approach for Predicting Solder Joint Fatigue Life in Wafer-Level Chip Scale Packages. Journal of Electronic Packaging, 143(1), Article 011001.
Chuang, W. C., Huang, Y., & Chen, P. E. (2023). Exploring the Influence of Material Properties of Epoxy Molding Compound on Wafer Warpage in Fan-Out Wafer-Level Packaging. Materials, 16(9), Article 3482.
Dina, A. S., Siddique, A. B., & Manivannan, D. (2022). Effect of Balancing Data Using Synthetic Data on the Performance of Machine Learning Classifiers for Intrusion Detection in Computer Networks. IEEE Access, 10, 96731-96747.
Goodfellow, I., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., Courville, A., & Bengio, Y. (2020). Generative Adversarial Networks. Communications of the Acm, 63(11), 139-144.
Habibi, O., Chemmakha, M., & Lazaar, M. (2023). Imbalanced tabular data modelization using CTGAN and machine learning to improve IoT Botnet attacks detection. Engineering Applications of Artificial Intelligence, 118, Article 105669.
Hsiao, H. H., & Wang, K. J. (2024). GAGAN: Global Attention Generative Adversarial Networks for Semiconductor Advanced Process Control. IEEE Transactions on Semiconductor Manufacturing, 37(1), 115-123.
I. Ho, P.-Y. L., T. Shih, A. Kang and Y. -P. Wang. (2023). Package Warpage of Whole Strip Evaluation with Interface Analysis in the Flip-Chip Die Bonding Process 2023 IEEE 73rd Electronic Components and Technology Conference (ECTC), Orlando, FL, USA.
Jeon, J. E., Hong, S. J., & Han, S. S. (2024). Utilization of Machine Learning and Explainable Artificial Intelligence (XAI) for Fault Prediction and Diagnosis in Wafer Transfer Robot. Electronics, 13(22), Article 4471.
Kim, J., Nam, Y., Kang, M. C., Kim, K., Hong, J., Lee, S., & Kim, D. N. (2021). Adversarial Defect Detection in Semiconductor Manufacturing Process. IEEE Transactions on Semiconductor Manufacturing, 34(3), 365-371.
Kong, J. W. Y., Kim, J. K., & Yuen, M. M. F. (2003). Warpage in plastic packages: Effects of process conditions, geometry and materials. IEEE Transactions on Electronics Packaging Manufacturing, 26(3), 245-252.
Lau, J. H. (1999). Thermal Stress and Strain in Microelectronics Packaging.
Lau, J. H. (2022). Recent Advances and Trends in Advanced Packaging. IEEE Transactions on Components Packaging and Manufacturing Technology, 12(2), 228-252.
Le, L. T., Nguyen, H., Dou, J., & Zhou, J. (2019). A Comparative Study of PSO-ANN, GA-ANN, ICA-ANN, and ABC-ANN in Estimating the Heating Load of Buildings' Energy Efficiency for Smart City Planning. Applied Sciences-Basel, 9(13), Article 2630.
Lei Xu, M. S., Alfredo Cuesta-Infante, Kalyan Veeramachaneni. (2019). Modeling Tabular data using Conditional GAN NeurIPS 2019, Vancouver CANADA.
Li, Z., Huang, J., Wang, J. C., & Ding, M. (2022). Comparative study of meta-heuristic algorithms for reactor fuel reloading optimization based on the developed BP-ANN calculation method. Annals of Nuclear Energy, 165, Article 108685.
Lin, K. Y., & Jamrus, T. (2024). Industrial data-driven modeling for imbalanced fault diagnosis. Industrial Management & Data Systems, 124(11), 3108-3137.
Liu, C. L., & Hsieh, P. Y. (2020). Model-Based Synthetic Sampling for Imbalanced Data. IEEE Transactions on Knowledge and Data Engineering, 32(8), 1543-1556.
Naqvi, M. R., Elmhadhbi, L., Sarkar, A., Archimede, B., & Karray, M. H. (2024). Survey on ontology-based explainable AI in manufacturing. Journal of Intelligent Manufacturing, 35(8), 3605-3627.
Ozturk, S., Kayabasi, E., Celik, E., & Kurt, H. (2018). Determination of lapping parameters for silicon wafer using an artificial neural network. Journal of Materials Science-Materials in Electronics, 29(1), 260-270.
Pramanik, P., Jana, R. K., & Ghosh, I. (2024). AI readiness enablers in developed and developing economies: Findings from the XGBoost regression and explainable AI framework. Technological Forecasting and Social Change, 205, Article 123482.
Salahouelhadj, A., Gonzalez, M., Vanstreels, K., van der Plas, G., Beyer, G., & Beyne, E. (2023). Analysis of warpage of a flip-chip BGA package under thermal loading: Finite element modelling and experimental validation. Microelectronic Engineering, 271, Article 111947.
Sethanan, K., & Jamrus, T. (2020). Hybrid differential evolution algorithm and genetic operator for multi-trip vehicle routing problem with backhauls and heterogeneous fleet in the beverage logistics industry. Computers & Industrial Engineering, 146, Article 106571.
Solati, A., Hamedi, M., & Safarabadi, M. (2019). Combined GA-ANN approach for prediction of HAZ and bearing strength in laser drilling of GFRP composite. Optics and Laser Technology, 113, 104-115.
Tsai, C. H., Liu, S. W., & Chiang, K. N. (2020). Warpage Analysis of Fan-Out Panel-Level Packaging Using Equivalent CTE. IEEE Transactions on Device and Materials Reliability, 20(1), 51-57.
Tsai, M. Y., Chang, H. Y., & Pecht, M. (2009). Warpage Analysis of Flip-Chip PBGA Packages Subject to Thermal Loading. IEEE Transactions on Device and Materials Reliability, 9(3), 419-424.
Wang, G. D., Ledwoch, A., Hasani, R. M., Grosu, R., & Brintrup, A. (2019). A generative neural network model for the quality prediction of work in progress products. Applied Soft Computing, 85, Article 105683.
Wang, H.-K., Lin, Y.-C., Chen, T.-Y., Wang, C.-C., & Hung, C.-P. (2025). Optimizing warpage in semiconductor packaging using GA-NN and conditional GAN with explainable AI. Computers & Industrial Engineering, 206, 111246.
Wang, J. L., Yang, Z. L., Zhang, J., Zhang, Q. H., & Chien, W. T. K. (2019). AdaBalGAN: An Improved Generative Adversarial Network With Imbalanced Learning for Wafer Defective Pattern Recognition. IEEE Transactions on Semiconductor Manufacturing, 32(3), 310-319.
Wang, S. Z., Sun, Y. M., Sheng, C., Feng, Z., Li, R., Xue, L. H., Liu, J., & Liu, S. (2021). Warpage Analysis and Prediction of the Advanced Fan-Out Technology Based on Process Mechanics. IEEE Transactions on Components Packaging and Manufacturing Technology, 11(12), 2201-2213.
Wu, M. L., & Wu, J. Y. (2025). Predictive Modeling and Optimization of Wafer Warpage in FOPoP Packaging Using Element Birth and Death Technique. IEEE Transactions on Components Packaging and Manufacturing Technology, 15(3), 454-462.
Xu, L., Skoularidou, M., Cuesta-Infante, A., & Veeramachaneni, K. (2019). Modeling tabular data using conditional GAN. In Proceedings of the 33rd International Conference on Neural Information Processing Systems (pp. Article 659). Curran Associates Inc.
Yu, J. B., & Liu, J. T. (2022). Multiple Granularities Generative Adversarial Network for Recognition of Wafer Map Defects. IEEE Transactions on Industrial Informatics, 18(3), 1674-1683.
Yuki Sugiura, D. M., Yasuhito Fujii, Kenta Imamura, Ippei Yamai, Takashi Yagihashi. (2024). Novel sheet molding compound technology for wafer-level packaging to overcome wafer warpage issue 2024 IEEE 74th Electronic Components and Technology Conference (ECTC), Denver, CO, USA.
Zhu, X. M., Mårtensson, P., Hanson, L., Björkman, M., & Maki, A. (2025). Automated assembly quality inspection by deep learning with 2D and 3D synthetic CAD data. Journal of Intelligent Manufacturing, 36(4), 2567-2582.