Metal AM Research

Analyzing 3D-printed aluminum samples from a Laser Powder Bed Fusion (LPBF) process to develop techniques for defect reduction and detection.

  • 1. Top Surface Image

    Showing the top surface of an aluminum 3D-printed sample using Laser Powder Bed Fusion.

  • 2. Surface Roughness

    Heat map showing variations in height.

  • 3. Calculating Surface Roughness

    Analyzing the image using Fiji (ImageJ2) yields a height value for each pixel.

  • 4. Internal Microstructure

    Showing crack and pore defects in the cross-section of the aluminum sample. Image taken with a microscope.

  • 5. Deep Learning Classification

    By training Fiji’s (ImageJ2) deep learning algorithms, the defects can be segregated into binary classes.

  • 6. Threshold Classification

    Thresholding is another way to transform an image into binary classes. It works by filtering based on the intensity value of each pixel.

  • 7. Results Example

    Results demonstrate that surface roughness is proportional to the number of internal defects. This means that by looking at the top surface alone, the number of internal defects can be accurately predicted!