Characterization of Defect Distribution in an Additively Manufactured AlSi10Mg as a Function of Processing Parameters and Correlations with Extreme Value Statistics

Fatigue failure in additively manufactured components commonly initiates at critical defects that combine large size with proximity to the surface. Therefore, understanding the relation of extreme large sized defects with process parameters is essential for structural integrity. This study explores the correlation between the process parameters and statistically derived extreme defect sizes in Selective Laser Melting using Extreme Value Theory (EVT). AlSi10Mg samples are fabricated with varying bulk process parameters, while maintaining constant contouring conditions to generate different defect distributions and are characterized using computed tomography. Defect size is quantified via (Formula presented.) along the bearing plane for analysis. EVT is applied using the block maxima approach and modeled using the Gumbel distribution to estimate the 99th percentile maximum defect sizes (x_0.99) representing the statistical extremes. Correlations between x_0.99 and process parameter indicators such as modified volume energy density and melt pool width are established using quadratic regression and Gaussian process regression. Although based on a limited dataset, the results demonstrate a viable framework for linking process parameters to statistically derived extreme defect sizes in additively manufactured components.