Robust Cross-Chamber One-Class Fault Detection in Semiconductor Manufacturing

Fault detection (FD) is essential for wafer quality control in semiconductor manufacturing (SM), as it can identify abnormal wafers in the early stages. However, frequent chamber discrepancies leads to distribution shifts in the data from different chambers, which causes performance degradation of the existing model. In this paper, we conceive a robust cross-chamber fault detection method in SM, which to the best of our knowledge is the first work that employs domain generalization to address the issue of cross-chamber fault detection in SM. Our basic idea is to map the samples from the source chambers and target chambers to the same hypersphere space, making normal samples cluster around a specific feature center while faulty samples stay away from it. Due to the scarcity of faulty samples, we propose a one-class classification-based fault detection method relying on normal samples to establish classification boundaries. To generalize the model to unseen target chambers, we design a meta-learning-based one-class domain generalization approach. We also devise a strategy to enhance distribution alignment within hypersphere, making the classification boundaries of various chambers be close to each other. The evaluations on real-world data collected from a wet process equipment in SM verify the high robustness of the model to chamber discrepancies with limited faulty samples. Note to Practitioners—The main challenges of fault detection in semiconductor manufacturing include the lack of faulty wafer samples and the frequent data distribution shifts across chambers. Traditional transfer learning-based fault detection methods, such as domain adaptation and domain generalization, often fail to effectively address the aforementioned challenges simultaneously. The key work in this paper is achieving robust cross-chamber fault detection with scarce faulty samples. We design a one-class classification model to learn classification boundaries from normal samples and propose a one-class domain generalization technique that achieves few-shot learning for fault samples through gradient alignment. We also introduce a new batch normalization technique to further align the classification boundaries. Engineers can apply the method to perform a intelligent fault detection on cross-chamber process tools for health management in semiconductor manufacturing.