Wafer maps contain information about defects and clustered defects that form failure patterns. Failure patterns exhibit the information related to defect generation mechanisms. The accurate classification of failure patterns in wafer maps can provide crucial information for engineers to recognize the causes of the fabrication problems. In this paper, we proposed a decision tree ensemble learning-based wafer map failure pattern recognition method based on radon transform-based features. Radon transform is applied on raw wafer map data to generate the new features which are exhibiting the geometric information of failure patterns in wafer map. Decision tree algorithm is applied to build decision tree ensemble and the final decision is made by aggregating the prediction results of decision trees. The effectiveness of the proposed method has been verified by using the real world wafer map data set (WM-811K).

Defective die on a wafer map tend to cluster in distinguishable patterns, and such defect patterns can provide crucial information to identify equipment problems or process failures in the semiconductor manufacturing. Therefore, it is important to accurately and efficiently classify the defect patterns. In this research, we propose a novel clustering-based defect pattern detection and classification framework for wafer bin map. The proposed framework has many advantages. Outlier detection and defect cluster pattern extraction can be done at the same time; arbitrarily shaped cluster patterns can be detected; and there is no need to specify the number of clusters in advance. Based on wafer bin map property, the parameters used in the clustering algorithm are fixed so that parameter sensitivity can be avoided. Since the defect patterns are classified based on extracted features, no labeled data and no supervised classification training are needed and single-type patterns as well as mixed-type patterns can be found. Extensive experiments conducted on a real-world WM-811K dataset has shown the superiority of proposed framework. The proposed framework can also be used in a big data environment to accelerate performance.

Ensemble data mining methods, also known as classifier combination, are often used to improve the performance of classification. Various classifier combination methods such as bagging, boosting, and random forest have been devised and have received considerable attention in the past. However, data dimensionality increases rapidly day by day. Such a trend poses various challenges as these methods are not suitable to directly apply to high-dimensional datasets. In this paper, we propose an ensemble method for classification of high-dimensional data, with each classifier constructed from a different set of features determined by partitioning of redundant features. In our method, the redundancy of features is considered to divide the original feature space. Then, each generated feature subset is trained by a support vector machine, and the results of each classifier are combined by majority voting. The efficiency and effectiveness of our method are demonstrated through comparisons with other ensemble techniques, and the results show that our method outperforms other methods.

The lack of complete labels and data heterogeneity are obstacles to the application of artificial intelligence-based methods in industrial scenarios, such as machinery fault diagnosis. To address these challenges, this article proposes a federated learning (FL) framework for the industrial Internet of Things based on bidirectional knowledge distillation (KD) and hard sample selection. In the framework, the cloud server provides a pretrained deep learning (DL) model based on the cross-domain public dataset to facilitate the cold start in real-world applications. Then during the training process, each participating factory trains its heterogeneous local DL model according to local data volume and computing resources. Bidirectional KD with feature maps and hard sample selection is then carried out on a shared dataset between the server and factories to share knowledge efficiently. Moreover, all the DL models used in the application of the proposed framework are designed based on expertise and attention mechanism to diagnose multiple types of machinery and faults. Case studies using the vibration data collected from multiple factories show that the proposed framework improves the fault diagnosis accuracy compared to other FL methods while significantly reducing communication overhead.