Adaptive Data Analysis for Fatigue Studies of Additively Manufactured 17-4PH Stainless Steel

Author(s):

Natasha C Bradley, Michael D Coatney, Todd C Henry, and Mulugeta A Haile

Abstract:

In this report, principal component analysis (PCA) is used to investigate quasi-static and tension–tension fatigue stress and strain data sets of modified ASTM E8 subsized dog-bone-shaped samples manufactured by a Markforged Metal X additive manufacturing printer. These large data sets were made using the strain field data created using digital image correlation (DIC) imaging technology. Using PCA unsupervised learning algorithms, we investigated if these machine learning techniques can provide various trends and patterns of these materials that DIC and other traditional material characterization techniques cannot show us. Although a simple PCA of images did not show a trend or unique pattern within the data, it did, however, show that reducing the data size and processing via DIC does not cause a large reduction in the DIC strain output.