Page 51 - MSAM-2-1
P. 51

Materials Science in Additive Manufacturing                           Data imputation strategies of PBF Ti64




                                        Max  Min  373  11  251  40  460  35  5  730  34  200  20  90  15  7000  25  100  35  14.5  0  479  403.5  1407  1333  0  119  17.12  65  0


                                     GINN imputation  SD  67.23  31.88  72.23  -4  1.22  100.86  34.86  14.60  501.83  6.53  2.43  225  38.65  262  34.32  420  140.24  154.94  25.66  6.53








                                        N  Mean  85.32  401  104.18  401  114.57  401  1.30  401  109.81  401  45.25  401  39.72  401  580.02  401  96.99  401  6.76  401  389.81  401  358.81  401  1167.01  401  1078.48  401  96.59  401  3.02  401
                                        Max  Min  373  11  251  40  460  35  5  −4  730  34  200  20  90  15  7000  25  100  35  14.5  0  479  225  403.5  262  1407  420  1333  0  119  17.12  65  0


                                     MICE imputation  SD  67.56  50.74  74.70  1.70  96.52  34.85  15.82  520.05  6.46  2.53  36.34  28.90  218.33  269.83  22.62  6.46
                                  Table 2. Statistical summaries of observed and imputed datasets for kNN, MICE, and GINN for variables with missing values







                                        N  Mean  92.93  401  102.90  401  115.82  401  1.86  401  108.81  401  45.32  401  41.30  401  618.34  401  96.67  401  5.84  401  372.65  401  356.86  401  1073.71  401  916.49  401  95.03  401  3.34  401
                                        Max  373  251  460  5  730  200  90  7000  100  14.5  479  403.5  1407  1333  119  65

                                        Min  11  40  35  −4  34  20  15  25  35  0  225  262  420  0  17.12  0
                                     kNN imputation  SD  65.14  29.34  73.92  0.96  97.57  34.85  7.22  514.52  6.51  1.84  27.84  12.76  107.53  125.32  20.84  6.51 N is the number of datapoints; maximum number of datapoints is 401 in this study; number of available datapoints for observed dataset is always smaller than 401 and may not be the same.




                                        Mean  89.07  122.09  114.32  0.28  106.03  45.27  62.89  638.93  96.94  6.54  393.97  372.27  1168.89  1040.13  89.08  3.06


                                        N  401  401  401  401  401  401  401  401  401  401  401  401  401  401  401  401
                                        Max  373  251  460  5  730  200  90  7000  100  14.5  479  403.5  1407  1333  119  65


                                        Min  11  40  35  −4  34  20  15  25  35  0  225  262  420  0  17.12  0 kNN: k-nearest neighbor, GINN: Graph imputation neural network, MICE: Multivariate imputation by chained equations
                                     Observed  SD  68.05  43.88  56.43  1.56  133.86  34.89  12.51  531.09  6.92  3.05  46.64  26.84  147.47  189.93  32.28  6.92



                                        Mean  89.20  117.73  106.42  0.57  125.56  45.30  60.01  613.23  96.67  6.31  385.05  366.31  1175.21  1049.56  75.64  3.33


                                        N  364  78  377  107  194  400  72  342  350  82  40  47  82  71  31  350




                                           Energy density (J/mm 3 )  Exposure duration (µs)  Hatch spacing (µm)  Laser focus (mm)  Layer thickness (µm)  Point distance (µm)  Scan speed (mm/s)  Microhardness (HV)  Macrohardness (HV)  Ultimate tensile strength (MPa)  Yield strength (MPa)  Young’s modulus (GPa)




                                     Features         Laser spot (µm)  Density (%)  Elongation (%)  Porosity (%)




            Volume 2 Issue 1 (2023)                         10                       https://doi.org/10.36922/msam.50
   46   47   48   49   50   51   52   53   54   55   56