Implementation of Silicon Wafer Defect Classification Web application using Deep Learning

• Published in: 2024 International Conference on Advancements in Power, Communication and Intelligent Systems (APCI) pp. 1 - 6
• Main Authors: Akshaya, Bura, Raga, Vavilala Rohitha, Reddy, Gundla Sridhar, Chaitanya, B. Krishna
• Format: Conference Proceeding
• Language: English
• Published: IEEE 21-06-2024
• Subjects: Accuracy; Auto encoders; Convolutional Neural Networks; Deep learning; Defected silicon wafers; Flask; Image classification; Pickle file; Quality control; Semiconductor device manufacture; Semiconductor device modeling; Semiconductor device reliability; Web interface; Web pages
• DOI: 10.1109/APCI61480.2024.10616637

This paper presents a solution for the wafers classification of defected silicon wafers, a critical task in semiconductor manufacturing processes. The classification aims to identify various patterns of defects which are pivotal for quality control. The presented approach in this paper leverages deep learning techniques, employing Convolutional Neural Networks (CNNs) and Autoencoders, to automatically learn discriminative features from wafer map images. The trained model is then serialized into a pickle file for efficient deployment and integration. To democratize access to the solution, a user-friendly web interface is developed using Flask, enabling seamless interaction with the classification model. Users can upload wafer map images through the web page, and in return, receive real-time predictions of the defect class, facilitating quick decision-making in manufacturing processes. The presented method demonstrates promising results, achieving high accuracy in classifying defected silicon wafers across diverse patterns. This contributes significantly to enhancing the efficiency and reliability of semiconductor manufacturing, ultimately leading to improved product quality and reduced production costs.