Deformation and Failure of Additively Manufactured Ductile Metals with Internal Porosity
Miers, John Carter
MetadataShow full item record
Interrupted in-situ tensile tests in a lab-based x-ray computed tomography machine were used to investigate the evolution of the strain field around internal defects. Digital volume correlation was utilized to directly determine local strain levels within the additively manufactured components in the vicinity of porosity defects. Effects of porosity on strain localization and eventual failure of the samples were evaluated. The influence of defect characteristics on the localization of strain was investigated. A non-local field analytical model for approximating the localization of strain based on defect geometry and loading direction was formulated. Correlation of the local characteristics and the non-local field with the local axial strain was evaluated. The non-local field, at both 1-to-1 voxel resolution and at the down sampled measurement point locations, was found to be the most correlated with the localization of strain followed by pore volume. Shallow neural network models were utilized to predict the local strain magnitude from the local characteristics in the undeformed frame and the magnitude of the tensile load applied to the component. Models were found to be accurate at predicting local strain magnitude before failure. With greater than 92% accuracy in predicting the variation in the measured strain field. Synthetic porosity alterations were applied to each sample to measure how small changes in porosity characteristics would affect the accumulation and distribution of strain in components. Volumetric alterations were found to have the greatest effect on magnitude regionally, but translations of the porosity were found to have a greater effect on the intervoid ligament location and shape and the overall likely path of ductile rupture. The direct measurements of strain field evolution in the present study established understanding regarding how internal defect structure characteristics influence the evolution of the local strain fields for additively manufactured components. Early onset of failure was found to be associated with the availability of neighboring porosity in the vicinity of large defects that allowed for rapid progression of the fracture path. This high-fidelity characterization and the associated phenomenological observations have bearing for supporting validation of numerical modeling frameworks for describing failure in these materials. The correlations established between local-characteristics and local-strain provide valuable information to designers to understand the hyperplane of specifying porosity tolerances. In addition, these correlations can improve the implemention of locally changing material properties and adaptation of constitutive relationships in numerical modelling frameworks of highly porous AM materials. The results of the non-local field analysis allow for the identification of strain localization hotspots and thus identifies areas that are more critical to model with finite element techniques. Utilization of the high-fidelity characterization and non-local fields via machine learning provides a procedural framework for the use of machine learning in component assessment and qualification. Additionally, predictions on altered data allowed for the generation of valuable engineering knowledge for understanding local porous material behavior.