DIGITAL RADIOGRAPHY INSPECTION TOOL FOR LARGE ADDITIVELY MANUFACTURED METALLIC COMPONENTS
Open Access
- Author:
- Stoner, Brant Edward
- Graduate Program:
- Mechanical Engineering
- Degree:
- Master of Science
- Document Type:
- Master Thesis
- Date of Defense:
- March 19, 2018
- Committee Members:
- Sanjay B Joshi, Thesis Advisor/Co-Advisor
Timothy William Simpson, Committee Member
Guhaprasanna Manogharan, Committee Member - Keywords:
- Nondestructive inspection
Additive manufacturing
directed energy deposition
multi-meter components
qualification
digital radiography - Abstract:
- Through layerwise deposition, additive manufacturing (AM) allows previously unmanufacturable geometries such as organic looking structures and complex internal geometries to be produced. Utilizing these advantages, performance improvements can be gained through techniques such as topology optimization, lattice structures, and part consolidation. Continued improvement of the AM process has shifted the focus of AM to end-use components. Qualification is a key step in enabling AM as an end-use manufacturing method, and the industry has turned to Computed Tomography (CT) as a tool in this process. Although well suited for complex components a few centimeters in size, CT struggles as AM components exceed a meter in length due to limited source energy, restrictive inspection volumes, and excessive data generation. This work proposes Digital Radiography (DR) as a non-destructive inspection method capable of accurately inspecting AM components approaching and exceeding a meter in length. To both expedite the inspection process and make DR a feasible inspection technique, an application, SMART DR was developed to optimize component orientation during the inspection process and provide the probability of detection for flaw sizes of interest prior to manufacturing. By providing this information, SMART DR allows for an inspection plan to be developed or the necessary design changes to be made to ensure that a component can be accurately inspected. SMART DR utilizes a ray trace algorithm for orientation optimization and experimental data along with theoretical relationships to determine a flaw’s probability of detection. A parallelized back projection algorithm utilizes a virtual representation of the user’s DR system and a STL representation of the component in question. By projecting the STL facets to the detector and back tracing the X-rays from the detector pixels to the source, the radiographic thickness of the component can be calculated efficiently. To determine the optimal orientation for inspecting the component, the back-projection algorithm is coupled with a genetic optimization scheme to minimize the radiographic thickness of the component. The probability of detection metric is based off of the contrast-to-noise ratio and normalized image unsharpness. Due to limited information on high energy DR in the literature, an experimental plan was created to aid in determining the contrast to noise ratio for source energy between 0.450 Mev to 12 Mev and thicknesses from 25.4 mm to 203.3 mm (1 in. to 8 in.). Square Ti-6Al-4V blocks 76.2 mm (3 in.) wide with cylindrical artificial flaws were used during the experiment. The contrast-to-noise ratio was then measured across each flaw based on ASTM standard E2597 while the normalized image unsharpness was determined based on ASTM standard E2698. Combining these two metrics allow for determining a flaw’s probability of detection for a wide range of flaw sizes, DR systems, and component thicknesses. Validation was conducted on a 316-stainless steel fin produced using laser based directed energy deposition process. A 0.450 Mev DR system with a 2.5mm spot size and a GE DXR250U-W digital detector array were used during validation. SMART DR produced an optimal orientation that was 40mm thick and predicted that a 0.75 mm diameter flaw with length 0.400mm, .800mm, or 1.60 mm had probability of detection of 91%, 20.7% and 21.7%, respectively. This aligned closely with the smallest detectable flaws on the radiographs of the 316-stainless steel fin, which were 0.508 mm and 0.764 mm. Initial validation results showed that SMART DR is capable of optimizing the DR process and providing accurate probability of detection values within the bounds of the data.