Detection of bent wires in semiconductors using deep learning

• Main Author: Pang, Ian Yi En
• Other Authors: Qian Kemao
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2023
• Subjects: Engineering::Computer science and engineering
• Online Access: https://hdl.handle.net/10356/166166
• Institution: Nanyang Technological University

The technology of semiconductors has been advancing greatly, with them being smaller and smaller over time. They are essential in every electronic product, and they serve as a basic requirement for electronic devices to be built. As a result, they are mass produced in large amounts and speeds. As a result of mass production, numerous errors can happen during the production process. Hence, quality control is crucial in this production process to prevent semiconductors from being used in bigger components, ultimately lowering costs of production. Hence, a good inspection system is required by companies to deal with this problem. The current inspection systems used by semiconductor production companies are usually automated by rule-based inspection systems. However, these as a result will only work in specific conditions they are built for. After the automation, a human is still required to manually approve the inspection. One such possible defect that can happen in semiconductors is the wire bending effect, whereby the gold or aluminium fine wires are bent in the semiconductors, leading to potential issues. With advances in deep learning technology, we will be attempting to create and experiment on a good inspection pipeline to take in semiconductor images and predict if the wires are bent or straight. The strengths and weaknesses of each created experiment will also be evaluated to provide new insights in this direction.