LOGIC product yield analysis by Wafer Bin Map pattern recognition supervised neural network

Abstract

Wafer Bin Maps (WBMs) are important for yield improvement to trace root causes. The characteristic of WBMs patterns are formed by processes, so process engineers can collect clues from the patterns and correlate them with speciJic processes. and this can save much time and eforts in finding the root causes. However, the existing learning algorithms have the main shortage of product dependency. For this reason, this work adopted a supervised learning methodology to develop an on-line WBMs pattern recognition system that maps WBMs into 70x70 binary images to salve this issue. Furthermore, this work also proposed a learning scheme to recognize repeating failures that are usually viewed as random pattern in the existing approaches.

Unsupervised spatial pattern classification of electrical-wafer-sorting maps in semiconductor manufacturing

“…The third approach is supervised learning (Lee et al, 1996;Chen et al, 2003). For instance, a set of wafers is classified by a human operator and then used to train a neural network.…”

Ra-Unet: A New Deep Learning Segmentation Method for Semiconductor Wafer Defect Analysis on Fine-Grained Scanning Electron Microscope (Sem) Images

“…Various machine learning models, including Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), and autoencoder, have been applied to this field. Chen et al (2003) used supervised learning to develop an online wafer bin map pattern recognition system, which solved the problems of product dependence and excessive derived patterns in unsupervised learning [3]. Huang et al (2019) proposed an intelligent defect detection system for metal products based on machine learning, using generative adversarial networks (GAN) to address the issue of insufficient surface defect datasets and achieve the detection of common stains, edge defects, scratches, and more [4].…”

“…This pattern is often indicative of a particular process flow, which can be used to identify potential sources of defects. In recent years, various machine learning and deep learning methods have been heavily applied in the detection of wafer map defect distribution [3]- [5].…”

An ordinal optimization theory-based algorithm for a class of simulation optimization problems and application

“…wafer may consist of 500-15,000 dies and each die is a chip of integrated circuits. Although there exist techniques such as the statistical methods and machine learning methods (Barnett, Grady, Purdy, & Singh, 2005;Chen, Lin, Doong, & Young, 2003) for monitoring the operations of the wafer probes, the probing errors may still occur in many aspects and cause some good dies being over killed; consequently, the profit is diminished. Fig.…”