[4B7] The advantage of machine learning to the inspector
T Koskinen, T Tyystjärvi, O Siljama and I Virkkunen
Trueflaw Ltd, Finland
NDT inspectors are highly trained professionals who have taken advantage of new inspection techniques such as phased array ultrasonics and digital radiography. While these new techniques improve the inspection result as a whole, they also produce a considerable amount more data than the older analogue methods. In inspections, the increased amount of data usually leads to more reliable results; however, this also leads to a longer inspection time. The majority of the inspector’s time is usually used to look at data where there are no indications to detect, evaluate or report. Some shortcuts such as merging and adjusting can speed up the data analysis, but they inherently destroy information.
Machine learning offers the opportunity to utilise modern inspection techniques to their full extent while maintaining high reliability, repeatability and efficiency. Machine learning further enables new statistical interpretation options that combine information from numerous indications, which have been unfeasible with the human inspector alone. In this paper, we study the different opportunities machine learning has to offer in the form of data analysis for the inspectors and how they affect the inspector’s task. These include examples from radiography and ultrasonic inspection and different ways to apply machine learning and present results to the inspector.
Machine learning offers the opportunity to utilise modern inspection techniques to their full extent while maintaining high reliability, repeatability and efficiency. Machine learning further enables new statistical interpretation options that combine information from numerous indications, which have been unfeasible with the human inspector alone. In this paper, we study the different opportunities machine learning has to offer in the form of data analysis for the inspectors and how they affect the inspector’s task. These include examples from radiography and ultrasonic inspection and different ways to apply machine learning and present results to the inspector.
