Detection of Unknown Defects in Semiconductor Materials from a Hybrid Deep and Machine Learning Approach

• Francisco López de la Rosa ^[1] ; José L. Gómez-Sirvent ^[1] ; Corinna Kofler ^[2] ; Rafael Morales ^[1] ; Antonio Fernández-Caballero ^[1]
  1. [1] Universidad de Castilla-La Mancha

     Universidad de Castilla-La Mancha

     Ciudad Real, España

     
  2. [2] KAI Kompetenzzentrum für Automobil ,Austria
• Localización: Bio-inspired Systems and Applications: from Robotics to Ambient Intelligence: 9th International Work-Conference on the Interplay Between Natural and Artificial Computation, IWINAC 2022, Puerto de la Cruz, Tenerife, Spain, May 31 – June 3, 2022, Proceedings, Part II / José Manuel Ferrández Vicente (dir. congr.) , José Ramón Álvarez Sánchez (dir. congr.) , Félix de la Paz López (dir. congr.) , Hojjat Adeli (aut.), 2022, ISBN 978-3-031-06527-9, págs. 356-365
• Idioma: inglés
• Texto completo no disponible (Saber más ...)
• Resumen

  • Artificial intelligence techniques such as deep learning and machine learning are nowadays implemented in inspection systems in a growing number of industries. These models have reached human-level performance in defect detection and classification tasks when enough data is available. However, most models use supervised learning approaches and, therefore, must have prior knowledge of the number of defect classes that may occur along the production line. This is a major problem in dynamic industries, such as the semiconductor manufacturing industry, where continuous changes in equipment and environment lead to the emergence of new classes of defects. Hence, it is necessary to detect new defect classes and classify them as “unknown” in order to study them meticulously and ensure a good quality of the manufactured semiconductor wafer. This paper presents a novel approach that fuses the ResNet50 convolutional neural network with a Gaussian mixture model for the detection of 100% of the images from the unknown defect class.