Expert Survey to Understand and Optimize Part Orientation in Direct Metal Laser Sintering

Hoelzle, David
Peng, Hao
Ghasri-Khouzani, Morteza
Gong, Shan
Attardo, Ross
Ostiguy, Pierre
Aboud Gatrell, Bernice
Budzinski, Joseph
Tomonto, Charles
Neidig, Joel
Journal Title
Journal ISSN
Volume Title
University of Texas at Austin

The additive manufacturing (AM) process Direct Metal Laser Sintering (DMLS) uses a layer-by-layer workflow to build complex architecture metal structures in low-volumes. The primary process failure mechanism is a thermal stress driven thermal distortion that yields out-of-tolerance manufacture or complete process failure. However, DMLS design experts have developed heuristic rules to optimize the part orientation and support structure to reduce the likelihood of failure. We believe that experts innately attempt to minimize the design metrics of support volume (V), support-to-part surface area (A), maximal cross-sectional area of the slicing planes (X), parallelism of part faces with the recoater blade (P), and part height (H); however, it is unclear what relative weighting of each metric the expert uses. This manuscript details an interactive expert survey, the statistical analysis of the survey responses, and the synthesis of an automatic algorithm for part orientation based on survey data. We received responses from 18 experts and 151 total part orientation responses. The median survey respondent had greater than four years of DMLS experience. Our analysis shows that the expert attempts to minimize metric V the most, metric X the second most, and metric H the third most; experts put essentially no weight on metrics A and P. The manuscript concludes with two orientation design studies where the expert survey responses are used in a least squares minimization algorithm to automatically orient the part for DMLS manufacture. As a comparison set, novice users were instructed to orient the parts for best DMLS printing success without using the tool and required multiple attempts to successfully print the test parts. The automatically oriented parts failed on our first iteration of the code. The manuscript concludes with our proposed modifications to the code to improve results.