Non-destructive evaluation (NDE) of additively manufactured (AM) parts is important to understand the impact of various process parameters and qualifying the built part. X-ray computed tomography (XCT) has played a critical role for rapid NDE and characterization of additively manufactured (AM) parts. XCT involves taking X-ray images of the part at different angles and computationally processing the images to obtain a 3D volumetric image. While the X-ray source has a polychromatic energy spectrum, common reconstruction methods makes the simplifying assumption that X-ray is monochromatic, and hence, the relation between the X-ray absorption and the object thickness is linear. In practice, however, the lower energy photons are absorbed easier than the higher energy photons, which hardens the X-ray spectrum as it passes through the object. This effect is called beam hardening which causes artifacts in reconstructed XCT images. This phenomenon is more prominent for high density materials such as in metals which are primarily used in metal additive manufacturing applications such as printing turbine blades for aviation industry. Beam hardening artifacts complicate the analysis of XCT images and adversely impact the process of detection of defects, such as pores and cracks, which are key to quality of the parts being printed.

 

In this work, we propose a novel deep learning method to remove beam hardening artifacts from the XCT images using available CAD models for the parts that are to be manufactured. In particular, we propose to use CAD models of the parts, introduce typical defects and simulate XCT measurements. Moreover, we use a physics-based model to simulate beam hardening artifact in the X-ray CT images. These simulated measurements, with and without beam hardening artifacts, were processed with common X-ray CT reconstruction algorithms. We then train the proposed deep learning method, on the pairs of simulated XCT reconstruction volumes, with and without beam hardening, to learn a fast, non-linear mapping function. The deep learning method will learn how to remove the beam hardening from the reconstructed volumes that have the artifacts. Once the mapping is learned, the proposed deep learning method can process new volumes on input that it hasn’t seen before, and remove beam hardening artifacts. In addition, the proposed deep learning method will reconstructs a 3D volume in a 2.5D scheme where each slice is reconstructed from multiple inputs slices of the input. This is in contrast with most deep learning approaches, for any application, that either work with 2D images or are too expensive when processing 3D volumes. The proposed approach allows high quality 3D reconstruction much faster than state-of-the-art methods. Through several case studies, the benefits of the proposed technique in comparison with common 2D/3D methods will be highlighted.


Beam Hardening Artifact Reduction in X-Ray CT Reconstruction of 3D Printed Metal Parts Leveraging Deep Learning and CAD Models

Division / Sponsors: 
Aerospace, Bioengineering, Applied Mechanics, Design Engineering Division, Advanced Energy Systems, Manufacturing Engineering Division, Noise Control and Acoustics, Nondestructive Evaluation, Diagnosis, and Prognosis (NDPD), Nuclear Engineering Division, Process Industries Division, Fluids Engineering, Heat Transfer, Materials, Microelectromechanical Systems, Safety Engineering and Risk Analysis, Computers & Information in Engineering


Paper Proceedings: https://asmedigitalcollection.asme.org/IMECE/IMECE2020/volume/84478

IMECE2020 - Virtual Conference

Supporting the need among researchers worldwide for additional online and multimedia content, ASME now offers the ASME Conference Video Collection. This platform will collect and provide long-term access to conference presentation videos, encompassing all technology areas within ASME's technical conference and events program. About 30,000-40,000 engineers across all mechanical engineering and related disciplines participate in ASME conferences each year. This includes authors of approximately 6,000 to 8,000 technical papers published by ASME, the majority of whom also submit video presentations for our virtual conferences. The video collection will be available as an add-on to existing product packages or as a separate subscription.