{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/augmented-ultrasonic-data-for-machine","title":"Augmented Ultrasonic Data for Machine Learning","arxiv_id":"1903.11399","date":"2019-03-26","proceeding":null,"authors":["Iikka Virkkunen","Tuomas Koskinen","Oskari Jessen-Juhler","Jari Rinta-Aho"],"abstract":"Flaw detection in non-destructive testing, especially in complex signals like\nultrasonic data, has thus far relied heavily on the expertise and judgement of\ntrained human inspectors. While automated systems have been used for a long\ntime, these have mostly been limited to using simple decision automation, such\nas signal amplitude threshold. The recent advances in various machine learning\nalgorithms have solved many similarly difficult classification problems, that\nhave previously been considered intractable. For non-destructive testing,\nencouraging results have already been reported in the open literature, but the\nuse of machine learning is still very limited in NDT applications in the field.\nKey issue hindering their use, is the limited availability of representative\nflawed data-sets to be used for training. In the present paper, we develop\nmodern, very deep convolutional network to detect flaws from phased-array\nultrasonic data. We make extensive use of data augmentation to enhance the\ninitially limited raw data and to aid learning. The data augmentation utilizes\nvirtual flaws - a technique, that has successfully been used in training human\ninspectors and is soon to be used in nuclear inspection qualification. The\nresults from the machine learning classifier are compared to human performance.\nWe show, that using sophisticated data augmentation, modern deep learning\nnetworks can be trained to achieve superhuman performance by significant\nmargin.","url_abs":"http://arxiv.org/abs/1903.11399v1","url_pdf":"http://arxiv.org/pdf/1903.11399v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"augmented-ultrasonic-data-for-machine","repo_url":"https://github.com/iikka-v/ML-NDT","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"none","reach":null}],"tasks":[{"task_slug":"machine-learning","task_name":"BIG-bench Machine Learning"},{"task_slug":"data-augmentation","task_name":"Data Augmentation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}