| Title |
Detection of Semiconductor Test Pins Defects using Neural Networks |
| Authors |
김영주(Young-Ju Kim) ; 진천덕(Cheon-Deok Jin) |
| DOI |
https://doi.org/10.5573/ieie.2024.61.2.43 |
| Keywords |
Deep learning; CNN; Perceptron; Automatic defect classification; Semiconductor |
| Abstract |
Semiconductor test pins, which are thinner than human hair, repeatedly make contact with semiconductor terminals to measure the electrical performance of the semiconductors. To ensure complete contact between the test pins and the semiconductor, the pin is equipped with a spring, allowing the structure on both sides of the spring to be compressed and then released repeatedly. If a defective test pin occurs, it results in incomplete contact, leading to errors in measuring semiconductor performance. This paper utilizes Convolutional Neural Networks (CNN) and Multilayer Perceptron Neural Networks (MLP) techniques to train and detect good and defective pins, thereby detecting defective pins without the use of human vision. Therefore, instead of manually detecting defective pins, the proposed models enable machines to automatically detect defective pins with about 97% accuracy. |