Page 49 - MSAM-2-1

Materials Science in Additive Manufacturing Data imputation strategies of PBF Ti64

Figure 7. Graphs comparing the distributions of the observed and imputed values of exposure duration, laser focus, point distance, and Young’s modulus
for graph imputation neural network-imputed dataset, where RF is relative frequency and P is probability.

Figure 8. Graphs comparing the distributions of the observed and imputed values of exposure duration, laser focus, point distance, and Young’s modulus
for the k-nearest neighbor-imputed dataset, where RF is relative frequency and P is probability.

For the imputed values of Young’s modulus, there of values imputed. To improve the distribution, more
appears to be a peak at around 120 GPa for all three varied data with a larger range of Young’s modulus values
imputed datasets. On further investigation of the imputed have to be obtained to allow the imputation algorithms to
datasets, it was found that most of the imputed values be more robust.
close to 120 GPa correspond to sets of observed data that The complete visualization plots for all incomplete
have similar values for the other parameters. Hence, it is
not unreasonable for the range of values for the imputed variables for all three imputed datasets can be found in the
Supplementary File.
Young’s modulus values to be around 120 GPa. There is also
limited observed data for Young’s modulus (92% missing Occasionally, imputing more than 50% of the values
values), which could have contributed to the limited range may be required depending on the specific dataset and

Volume 2 Issue 1 (2023) 8 https://doi.org/10.36922/msam.50