Explainable AI-infused ultrasonic inspection for internal defect detection
Explainable AI-infused ultrasonic inspection for internal defect detection
复制标题
DOI：
10.1016/j.cirp.2022.04.036
复制
发表时间：
2022-07-12
影响因子：
4.1
通讯作者：
Bukkapatnam, Satish T. S.
中科院分区：
文献类型：
--
作者：
Karthikeyan, Adithyaa;Tiwari, Akash;Bukkapatnam, Satish T. S.
虽然人工智能和成像技术正在极大地改变工艺和机器状态监测，但产品检测仍然局限于探测几何形状和表面形态。地下和整体检查仍然非常缓慢和不精确。本文提出了一种可解释的人工智能（XAI）注入超声成像方法，用于快速检测包括产品缺陷在内的工件。这种方法导致在远离人工制品的图像中发现相关的空间模式。这一发现使当前图像分割方法无法识别的伪影得以准确检测（bbb80 %），并可能对产品质量和（网络）安全保障技术产生深远影响。(c) 2022年cirp。Elsevier Ltd.出版。版权所有。
While AI and imaging technologies are dramatically transforming the process and machine condition monitoring, product inspection remains confined to probing the geometry and surface morphology. Subsurface and bulk inspection remain prohibitively slow and imprecise. This paper presents an explainable AI (XAI)-infused ultrasound imaging approach for rapid detection of artifacts including product defects. The approach led to the discovery of correlated spatial patterns in the images located away from the artifacts. This discovery enabled accurate (> 80%) detection of artifacts that are not discernible with the current image segmentation methods, and it could profoundly impact product quality and (cyber)security assurance technologies. (C) 2022 CIRP. Published by Elsevier Ltd. All rights reserved.