基于可解释深度学习的晶圆制造工期预测方法

An Explainable Deep Learning Approach for Cycle Time Prediction in Wafer Fabrication

摘要

在现代晶圆制造系统中，作业工期的精确预测对提高生产计划的可靠性、优化资源配置与提升工厂整体运营效率具有重要价值。然而，由于制造流程高度复杂，作业之间相互依赖且特征数据呈现非线性和高维特性，传统预测方法在实际应用中存在性能瓶颈。为应对上述挑战，本文提出一种基于可解释深度学习的工期预测方法，融合了Transformer神经网络与模型可解释性技术。构建以作业规模、瓶颈前排队长度、工厂在制品量等关键生产特征为输入的数据集，并进行标准化与时间序列处理。利用Transformer模型对特征间复杂关系建模，实现对作业工期的高精度预测。在模型训练完成后，引入LIME和SHAP两种方法从局部与全局角度揭示模型预测的依据，识别影响预测结果的关键因素。实验基于真实晶圆制造数据集开展，对比传统机器学习模型与深度学习基线模型，验证了所提方法在预测准确性与模型可解释性方面的综合优势。研究结果表明，本文方法不仅在数值上显著提升了预测性能，也为工厂调度系统的智能化与透明化提供了技术支持，具有较高的实际应用潜力。

Abstract

In modern wafer fabrication systems, accurate prediction of job cycle time is vital to production planning reliability, resource allocation, and factory operational efficiency. However, the inherent complexity of manufacturing processes, involving strong job interdependencies and nonlinear, high-dimensional data, limits the effectiveness of traditional prediction methods in practice. To address these challenges, this paper proposes an interpretable deep-learning approach for cycle time prediction that integrates a Transformer neural network with model explainability techniques. First, a dataset is constructed using key production features such as job sizes, bottleneck queue lengths, and work-in-process levels, followed by normalization and time-series preprocessing. Then, a Transformer model is employed to capture the complex relationships among these features, enabling highly accurate cycle time prediction. After model training, both LIME and SHAP methods are applied to reveal the prediction logic from local and global perspectives, thereby identifying the key factors that influence prediction results. Experiments conducted on a real-world wafer fabrication dataset, with comparisons against traditional machine-learning models and deep-learning baseline models, demonstrate the superior performance of the proposed approach in both prediction accuracy and model interpretability. Results also indicate that the proposed approach not only achieves excellent prediction performance on complex production data but also provides technical support for the intelligent and transparent scheduling in wafer fabrication systems, with strong application potential.

关键词

晶圆制造 / 工期预测 / 可解释人工智能 / 深度学习 / Transformer模型

Key words

wafer fabrication / cycle time prediction / explainable AI / deep learning / Transformer model

参考文献

[1] Fu W H, Chien C F, Tang L Z. Bayesian network for integrated circuit testing probe card fault diagnosis and troubleshooting to empower Industry 3.5 smart production and an empirical study[J]. Journal of Intelligent Manufacturing, 2022, 33(3): 785-798.

[2] Zhang Y, Tiňo P, Leonardis A, et al. A survey on neural network interpretability[J]. IEEE Transactions on Emerging Topics in Computational Intelligence, 2021, 5(5): 726-742.

[3] 汪俊亮, 高鹏捷, 张洁, 等 . 制造大数据分析综述: 内涵、方法、应用和趋势[J]. 机械工程学报, 2023, 59(12): 1-16.

[4] Wang Junliang, Gao Pengjie, Zhang Jie, et al. A review of manufacturing big data: connotation, methodology, application and trends[J]. Chinese Journal of Mechanical Engineering, 2023, 59(12): 1-16.

[5] 崔岩, 肖泽文, 钟雪. 日本半导体产业的结构转型与新发展战略论析[J]. 现代日本经济, 2025, 44(2): 44-58.

[6] Cui Yan, Xiao Zewen, Zhong Xue. Analysis on structural transformation and new development strategy of Japan's semiconductor industry[J]. Contemporary Economy of Japan, 2025, 44(2): 44-58.

[7] Wang C, Jiang P Y. Deep neural networks based order completion time prediction by using real-time job shop RFID data[J]. Journal of Intelligent Manufacturing, 2019, 30(3): 1303-1318.

[8] Tirkel I. Forecasting flow time in semiconductor manufacturing using knowledge discovery in databases[J]. International Journal of Production Research, 2013, 51(18): 5536-5548.

[9] Chen T T, LIN Y C. Fuzzified deep neural network ensemble approach for estimating cycle time range[J]. Applied Soft Computing, 2022, 130: 109697.

[10] Chen T, Wang Y C. Hybrid big data analytics and Industry 4.0 approach to projecting cycle time ranges[J]. The International Journal of Advanced Manufacturing Technology, 2022, 120(1): 279-295.

[11] 严如强, 商佐港, 王志颖, 等 . 可解释人工智能在工业智能诊断中的挑战和机遇: 先验赋能[J]. 机械工程学报, 2024, 60(12): 1-20.

[12] Yan Ruqiang, Shang Zuogang, Wang Zhiying, et al. Challenges and opportunities of XAI in industrial intelligent diagnosis: priori-empowered[J]. Chinese Journal of Mechanical Engineering, 2024, 60(12): 1-20.

[13] Serradilla O, Zugasti E, Cernuda C, et al. Interpreting remaining useful life estimations combining explainable Artificial Intelligence and domain knowledge in industrial machinery[C]// 2020 IEEE International Conference on Fuzzy Systems (FUZZ-IEEE). Piscataway: IEEE, 2020.

[14] Protopapadakis G, Apostolidis A, Kalfas A I. Explainable and interpretable AI-assisted remaining useful life estimation for aeroengines[C/OL]// Proceedings of ASME Turbo Expo2022: Turbomachinery Technical Conference and Exposition. New York: ASME, 2022. https://doi.org/10.1115/GT2022-80777.

[15] Gupta S, Venugopal A, Mohan M J. Fault detection and diagnosis using autoencoders and interpretable AI-case study on an industrial chiller[C]// 2022 IEEE International Symposium on Advanced Control of Industrial Processes (AdCONIP). Piscataway: IEEE, 2022.

[16] Martakis P, Movsessiar A, Reuland Y, et al. A semi-supervised interpretable machine learning framework for sensor fault detection[J]. Smart Structures and Systems, 2022, 29: 251-266.

[17] Chen T, Wang Y C. A two-stage explainable Artificial Intelligence approach for classification-based job cycle time prediction[J]. The International Journal of Advanced Manufacturing Technology, 2022, 123(5): 2031-2042.

[18] 高鹏捷, 汪俊亮, 张洁. 面向晶圆制造工期预测的可解释深度学习方法[J]. 机械工程学报, 2024, 60(22): 179-191.

[19] Gao Pengjie, Wang Junliang, Zhang Jie. Interpretable deep learning method for wafer manufacturing cycle time forecasting[J]. Chinese Journal of Mechanical Engineering, 2024, 60(22): 179-191.

[20] Chen T. Embedding a back propagation network into fuzzy c-means for estimating job cycle time: wafer fabrication as an example[J]. Journal of Ambient Intelligence and Humanized Computing, 2016, 7(6): 789-800.

[21] 张蓝天, 石宇强. 考虑特征学习的IPSO-LSTM晶圆加工周期预测[J]. 工业工程, 2023, 26(3): 143-150.

[22] Zhang Lantian, Shi Yuqiang. Wafer cycle time prediction of IPSO-LSTM considering feature learning[J]. Industrial Engineering Journal, 2023, 26(3): 143-150.

基金资助

上海市高水平机构建设运行计划“软科学研究”项目(25692116600)

上海高校青年教师培养资助计划(ZZ202203036)