Convolutional neural network (CNNs) based image diagnosis for failure analysis of power devices. (September 2019)

• Record Type: Journal Article
• Authors: Watanabe, Akihiko; Hirose, Naoto; Kim, Hyoungseop; Omura, Ichiro
• Is Part Of: Microelectronics and reliability. Volume 100/101(2019)
• Publication Date: 2019-09
• Subjects: Electronic apparatus and appliances -- Reliability -- Periodicals; Miniature electronic equipment -- Periodicals; Appareils électroniques -- Fiabilité -- Périodiques; Équipement électronique miniaturisé -- Périodiques; Electronic apparatus and appliances -- Reliability; Miniature electronic equipment; Periodicals; 621.3815
• DOI: 10.1016/j.microrel.2019.113399
• Languages: English
• ISSNs: 0026-2714

Abstract

Abstract: An image diagnosis by deep learning was applied to failure analysis of power devices. A series of images during a process to failure by power cycling test was used for this method. The images were obtained by a scanning acoustic microscopy of our real-time monitoring system. An image classifier was designed based on a convolutional neural network (CNNs). A developed classifier successfully diagnosed input image into a normal device and an abnormal device. The accuracy of classification was improved by introducing a pre-training and an overlapping pooling into the system. A technique to extract a feature related a failure is essential for the failure analysis based on the real-time monitoring and the deep learning is one likely candidate for it. Highlights: A failure analysis based on real-time monitoring of power devices under acceleration test. An image diagnosis by deep learning was applied to failure analysis of power devices. An image classifier was designed based on a convolutional neural network (CNNs). The accuracy of classification was improved by introducing a pre-training and an overlapping pooling into the system.