Restricted (Penn State Only)
Sakpal, Aniket Sunil
Graduate Program:
Industrial Engineering
Master of Science
Document Type:
Master Thesis
Date of Defense:
April 07, 2017
Committee Members:
  • Hui Yang, Thesis Advisor
  • Russell Richard Barton, Committee Member
  • Janis P Terpenny, Committee Member
  • Additive Manufacturing
  • Defects
  • Simulation
  • Multifractals
Additive manufacturing is often a complex undertaking involving interaction between an energy source, feedstock, and substrate. Many independent process variables such as travel speed, feedstock flow pattern, laser power, chamber gas, etc. contribute significantly to the overall build quality of the component. Any fluctuation in these process variables affects the quality of the final part inducing defects and results in non-conformance to dimensional tolerance, microstructure, and properties. A wide variety of sensor data and analysis have been used for process monitoring and quality control of the additive manufacturing processes. Sensor feedback enables to form a correlation between the variation in process parameters contributing to the changes in the final build quality. This thesis presents new methods of multifractals and uncorrelated multilinear principal component analysis of additive manufacturing images for process machine modeling and defect characterization in the additive manufacturing based components. First, we simulated the classical defects encountered in additive manufacturing on a set of in-situ images of AM process. Second, we applied the proposed methodologies to the in-situ images and extracted the parametric features correlating to the defects and irregularities induced on the images. Third, we checked the statistical significance of the parametric features from the images of defects to the images of control quality. In the end, we compared both the methodologies for successful detection and characterization of defects/discontinuities emerged in the additive manufacturing processes.