Fault Detection and Classification in Semiconductor Manufacturing for Sensor Screening Using Multi-Layer Deep Neural Network

The fault detection and classification are a prediction model that employs the sensor equipment data to detect if each wafer is healthy or faulty which is significant to attain a greater yield. However, the manufacturing of semiconductors has difficulties in maintaining high-quality production because of potential defects. Hence, effective fault detection and classification is essential to overcome this issue. In this research, the multi-layer Deep Neural Network (DNN) is proposed for fault detection and classification in semiconductor manufacturing for sensor screening. Initially, the image is acquired from a benchmark dataset of Printed Circuit Board (PCB) dataset. The median filter is employed to eliminate the noise in the acquired image. Then, the feature extracted is established using a Gray Level Co-occurrence Matrix (GLCM). Finally, the multi-layer DNN is used to detect and classify as healthy or faulty in semiconductors effectively. The existing techniques like Convolutional Neural Network-VGG16 (CNN-VGG16), Global Contextual attention augmented You Only Look Once (GCC-YOLO), DL-based Path Aggregation Network (PANet), and improved fully CNN are used to compare with proposed technique. When compared to this existing technique, the multi-layer DNN achieves a better accuracy of 99.89% respectively.