Search

Artificial intelligence is providing new possibilities for analysis in the field of industrial radiography. As capabilities evolve, there is the need for knowledge concerning how to deploy these technologies in practice and benefit from the new automatically generated information. In this study, automatic defect recognition based on machine learning was deployed as an aid in industrial radiography of laser welds in an aerospace component, and utilized to produce statistics for improved quality control. A multi-model approach with an added weld segmentation step improved the inference speed and decreased false calls to improve field use. A user interface with visualization options was developed to display the evaluation results. A dataset of 451 radiographs was automatically analysed, yielding 10037 indications with size and location information, providing capability for statistical analysis beyond what is practical to carry out with manual annotation. The distribution of indications was modeled as a product of the probability of detection and an exponentially decreasing underlying flaw distribution, opening the possibility for model reliability assessment and predictive capabilities on weld defects. An analysis of the indications demonstrated the capability to automatically detect both large-scale trends and individual components and welds that were more at risk of failing the inspection. This serves as a step towards smarter utilization of non-destructive evaluation data in manufacturing. [ABSTRACT FROM AUTHOR]

Probability of detection (POD) as a metric for quantifying the capability of inspection procedures in nondestructive evaluation (NDE), has been applied and evolved in industries since 1970s. Progress had been noted when certain statistical functions were brought up to model POD behavior, including log-normal model (also referred as Probit model). This model had been concluded to be the best fit and therefore has been widely used in many studies, while the involved assumptions and conditions were not carefully addressed and explored. To make flexible POD datasets available for specific inspection procedures and reduce the number of expensive experiments needed, model-assisted POD (MAPOD) is an alternative. This paper addresses a pure simulation-based POD procedure of an inspection scenario involving phased array ultrasonic testing (PAUT) on lack-of-fusion defects in additive manufactured (AM) components. The mathematical simulations are performed by an ultrasonic testing (UT) simulation software, simSUNDT, developed at Chalmers University of Technology in Sweden. Resulted inspection datasets with the proposed data processing steps are evaluated in terms of the assumptions and conditions of log-normal POD model, with the purpose of discussing the POD model validity under different circumstances. Simulation-based POD curves are finally compared with several discrete POD values at some defect sizes, calculated through massive computations from physics-model based metamodel. Comparisons and observations confirm satisfactory application of log-normal POD model despite some violations in model hypotheses. [ABSTRACT FROM AUTHOR]