Browsing by Subject "correlation coefficient"

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    Powder Features Affecting Structural and Mechanical Properties of Additively Manufactured Inconel 718: A Machine Learning Analysis
    (University of Texas at Austin, 2021) Hossain, M.S.; Silva, D.F.; Vinel, A.; Liu, J.; Shamsaei, N.
    The aim of this paper is to select important Inconel 718 powder properties that can have significant effect on the structural and mechanical properties of Laser-Beam Powder Bed Fusion manufactured specimens. The dataset used was provided by NASA and contains powder rheological, morphological, and chemical composition properties. The output variables considered are melt pool depth, high cycle fatigue life, porosity volume fraction and porosity size. Initially, Pearson correlation coefficient matrix is used to reduce the number of predictor features. Several statistical and machine learning algorithms including stepwise regression, LASSO, and random forest regression are used to identify the powder properties that have the strongest impact on the selected outputs. The variables identified using the different statistical and machine learning techniques are similar, which increases the confidence of the findings.