Abstract:The patterns on wafer maps play a crucial role in helping engineers identify the causes of production issues during semiconductor manufacturing. In order to reduce costs and improve accuracy, automation technology is essential, and recent developments in deep learning have led to impressive results in wafer map pattern recognition. In this context, inspired by the effectiveness of semi-supervised learning and contrastive learning methods, we introduce an innovative approach that integrates the Mean Teacher framework with the supervised contrastive learning loss for enhanced wafer map pattern recognition. Our methodology not only addresses the nuances of wafer patterns but also tackles challenges arising from limited labeled data. To further refine the process, we address data imbalance in the wafer dataset by employing SMOTE and under-sampling techniques. We conduct a comprehensive analysis of our proposed method and demonstrate its effectiveness through experiments using real-world dataset WM811K obtained from semiconductor manufacturers. Compared to the baseline method, our method has achieved 5.46%, 6.68%, 5.42%, and 4.53% improvements in Accuracy, Precision, Recall, and F1 score, respectively.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the challenge of wafer map pattern recognition in the semiconductor manufacturing process, especially how to use a large amount of unlabeled data to improve the performance of the model. Specifically, the paper focuses on the following aspects: 1. **Limited labeled data**: In actual production, due to cost and time limitations, it is impossible to label all generated data. Therefore, how to effectively use a large amount of unlabeled data to improve model performance is a key issue. 2. **Data imbalance problem**: There is a serious class imbalance problem in the wafer map dataset. Some types of defect patterns (such as "Donut" and "NearFull") account for an extremely low proportion, which brings difficulties to model training. 3. **Application of automation technology**: In order to reduce costs and improve recognition accuracy, the application of automation technology (especially deep - learning methods) in wafer map pattern recognition is crucial. To solve the above problems, the author proposes an innovative method that combines the Mean Teacher framework with supervised contrastive loss for enhancing wafer map pattern recognition. Through this method, not only can the complexity of wafer map patterns be dealt with, but also the challenges brought by limited labeled data can be effectively addressed. In addition, the author also adopts SMOTE and undersampling techniques to solve the class imbalance problem in the dataset. Through experimental verification, this method has achieved significant performance improvement on the WM811K dataset, increasing the accuracy, precision, recall, and F1 - score by 5.46%, 6.68%, 5.42%, and 4.53% respectively compared to the baseline method. ### Formula summary - **Exponential Moving Average (EMA) update formula**: \[ \theta_{\text{teacher}}(t)=\alpha\cdot\theta_{\text{teacher}}(t - 1)+(1-\alpha)\cdot\theta_{\text{student}}(t) \] where $\alpha$ is the decay factor, and its value range is within [0, 1]. - **Supervised contrastive loss function**: \[ L_{\text{supcontrast}}=-\sum_{i = 1}^{N}\frac{1}{|P(i)|}\sum_{p\in P(i)}\log\frac{\exp(\hat{f}_i\cdot\hat{f}_p/\tau)}{\sum_{a\in A(i)}\exp(\hat{f}_i\cdot\hat{f}_a/\tau)} \] where: - $P(i)$ represents the set of all positive sample indices of the $i$-th sample. - $A(i)$ represents the set of indices of all samples in the batch, including sample $i$ itself. - $\tau$ is the temperature parameter. - $\hat{f}_i\cdot\hat{f}_p$ is the dot product between normalized feature vectors, which is used as a similarity measure. - **Total loss function**: \[ L = L_{\text{consistency}}+L_{\text{classification}}+L_{\text{supcontrast}} \] By combining these techniques and methods, the author has successfully improved the performance of wafer map pattern recognition and provided new perspectives and possibilities for the field of semi - supervised learning.

Utilizing the Mean Teacher with Supcontrast Loss for Wafer Pattern Recognition

Wafer Map Defect Patterns Semi-Supervised Classification Using Latent Vector Representation

Wafer Defect Pattern Labeling and Recognition Using Semi-Supervised Learning

Semi-Supervised Framework for Wafer Defect Pattern Recognition with Enhanced Labeling.

Applying Data Augmentation and Mask R-CNN-Based Instance Segmentation Method for Mixed-Type Wafer Maps Defect Patterns Classification

Contrastive deep clustering for detecting new defect patterns in wafer bin maps

Unsupervised Learning for Wafer Surface Defect Pattern Recognition

Antecedent hash modality learning and representation for enhanced wafer map defect pattern recognition

Wafer map defect pattern detection method based on improved attention mechanism

Auto-Labeling for Pattern Recognition of Wafer Defect Maps in Semiconductor Manufacturing

When Wafer Failure Pattern Classification Meets Few-shot Learning and Self-Supervised Learning

PaLM: Point Cloud and Large Pre-trained Model Catch Mixed-type Wafer Defect Pattern Recognition

A Deep Convolutional Neural Network for Wafer Defect Identification on an Imbalanced Dataset in Semiconductor Manufacturing Processes

AI classification of wafer map defect patterns by using dual-channel convolutional neural network

An Analog Wafer Map Clustering Model with Deep-Learning Based Data Augmentation and Feature Extraction

Wafer Bin Map Recognition with Autoencoder-based Data Augmentation in Semiconductor Assembly Process

CowSSL: contrastive open-world semi-supervised learning for wafer bin map

Machine Learning Based Wafer Defect Detection

Advances in machine learning and deep learning applications towards wafer map defect recognition and classification: a review

Wafer Map Defect Classification Using Autoencoder-Based Data Augmentation and Convolutional Neural Network

Progressive hierarchical deep reinforcement learning for defect wafer test