Advanced Machine Learning for Non-Destructive Testing

and armament system components to identify defects before these products reach the warfighter. Highly skilled and experienced radiographers must interpret the visual results, and in many cases, this process is not only laborious and time consuming, but also subjective and inconsistent. The development of advanced AI algorithms opens the opportunity to develop an automated system to identify potential defects to aid the interpretation of the results to expedite the inspection process. To address this problem, Sky Park Labs (SPL) proposes DeepNDT, a software tool comprised of machine learning algorithms which will assist a radiographer in the interpretation of radiographic inspection results. The system learns an intrinsic feature representation that models the underlying composition and structure of x-rays from limited and unlabeled x-ray images. Next, the system employs the resulting representation to build a classifier that automatically detects any defects and characterizes the type, geometry and size, for a radiographer to assess and make a final decision. We hypothesize that the proposed AI-based automation will dramatically improve upon existing systems that help the radiographer’s interpretation of inspection results for non-destructive testing.