Importance of microstructure modeling for additively manufactured metal post-process simulations

Highlights

• Influence of 3D SLM microstructure induced anisotropy on milling and laser peening.

• SLM microstructure inhomogeneity reveals expanded temperature/plastic strain fields.

• Shear and ductile damage progression rates in milling vary due to SLM inhomogeneity.

• Asymmetry in residual stress clearly evident in LSP from SLM-induced inhomogeneity.

Abstract

This work investigates the significance of microstructure-level modeling by simulating the material response of an inhomogeneous selective laser melted (SLM) Inconel 625 specimen subjected to two different post-process operations, namely micro-milling and laser shock peening (LSP). A physics-based thermal finite element simulation is executed to obtain the SLM thermal history from which a 3-dimensional inhomogeneous microstructure representative volume element (RVE) is generated via the Dynamic Kinetic Monte Carlo predictive model. A Johnson–Cook plasticity definition coupled with Hall–Petch strengthening is used to define unique yield surfaces for individual grains based on their major diameters. Micro-milling and LSP simulations are subsequently executed with and without considering an inhomogeneous microstructure RVE in attempt to elucidate differences in the plastic strain, temperature, induced stress magnitude and distribution, as well as differences that arise during material removal for the micro-milling only. The micro-milling simulations reveal a greater volumetric distribution of plastic strain and temperature for the inhomogeneous case, although the homogeneous case with isotropic assumption reveals greater heat dissipation at the tool-workpiece interface with 27% greater contact pressure and 39% greater frictional shear stress. Examining the ductile and shear damage progression at a specific time increment reveals that the inhomogeneous model has a slightly lower damage propensity in comparison the homogeneous case, despite having identical damage models and boundary conditions. Variation in the SLM process-dependent yield surfaces, for grains at different locations, results in spatial variations of the computed stress triaxiality, which influences the material removal, as well as the stress concentrations developed near the tool-workpiece interface. Thus, a process-structure-property relationship is captured with the microstructure modeling. This work is the first to illuminate the importance of capturing SLM-induced anisotropy, considering the additively manufactured grain structure subject to micro-milling and LSP post-processes.

Introduction

The understanding of microstructure-specific material response, such as distribution of residual stress (RS) in a manufactured component for example, is an important factor towards determining the mechanical performance it can deliver. While experimental, analytical, and numerical methods exist to characterize RS, their results and usefulness can vary significantly depending on (1) the degree of volumetric “averaging” in experimentally measured RS, and (2) the accuracy of material constitutive behavior representation in analytical and numerical models. A rigorous consideration of microstructure in predictive models may provide for better estimation of the state of stress during and after metal post-process operations. In this work, the influence of an inhomogeneous microstructure, induced by selective laser melting (SLM) type additive manufacturing, is investigated in conjunction with micro-milling and laser shock peening (LSP) post processes. These processes are studied considering that their respective strain rates differ by approximately two orders of magnitude, 10${}^4$ and 10${}^6$ s${}^{-1}$, respectively (Wu, Zhang, 2014, Zhou, Li, He, He, Nie, Chen, Lai, An, 2013). To distinguish the technical approach and findings of the demonstrated work, prior experimental, analytical and numerical attempts to characterize the microstructure-specific material response induced by micro-milling and laser shock peening are first discussed.

Various numerical techniques to model microstructure-specific response can be found in literature. Typically, microstructure modeling in finite element methods is achieved by either heterogeneous representative volume element (RVE) replication of the different phases (Chuzhoy, DeVor, Kapoor, Bammann, 2002, Ljustina, Larsson, Fagerström, 2014), or inhomogeneous RVE replicating the individual grains based on morphology, as well as porosity and other defects (Chen, Wang, Kysar, Yao, 2007, Sunny, Gleason, Mathews, Malik, 2021, Wang, Kysar, Yao, 2008). Capturing the microstructure-specific response in orthogonal cutting/turning models has been found to offer improved insight towards the distribution of RS in the machined component, whilst also providing better prediction of cutting forces and chip morphology, as summarized in Table 1. Capturing the microstructure-specific response in laser shock peening (LSP) models has revealed anisotropic material response with asymmetry in the RS distribution, as summarized in Table 2.

Contrary to the studies discussed in Table 2, researchers whose focus was not influenced by microstructure resorted to FE models for LSP that assumed material isotropy and homogeneity. They obtained RS distributions exhibiting perfect symmetry and axisymmetry (Brockman, Braisted, Olson, Tenaglia, Clauer, Langer, Shepard, 2012, Correa, Peral, Porro, Díaz, de Lara, García-Beltrán, Ocaña, 2015, Warren, Guo, Chen, 2008). Simulations by Warren et al. revealed significant difference when compared with X-ray diffraction (XRD) measurements of surface RS (Warren et al., 2008). Disregarding the plausible anisotropic response from the material considering the underlying microstructure, they attributed the discrepancy to: (1) averaging of the RS in the XRD technique, (2) inability to locate the exact LSP shot center for XRD measurement, and (3) numerical errors. In addition, Brockman et al. stated that XRD measurements could be misleading as the resolution may not necessarily allow complete details of the RS field to be captured (Brockman et al., 2012).

The importance of microstructure consideration in numerical models, especially in cases involving inhomogeneity and/or heterogeneity is evident from the literature, however, none of the models considered microstructure arising from an SLM build. Milling simulations in the literature do not highlight the importance of the ductile and shear damage parameters, nor do they discuss how these damage parameters are influenced during the cut. XRD measurements with limited spatial resolution, are subject to averaging, and hence cannot comprehensively capture the nuances in RS distribution. In addition, it should be noted that the studies in Tables 1 and 2 that do feature microstructure modeling (although not from SLM) relied on optical micrographs or EBSD images, and thus were based on 2D representations of the grain structure. While a 2D approach may sufficiently capture the in-plane material response at lower computational cost, assumptions have to be made regarding the out-of-plane behavior. Hence, a need for 3D predictive modeling of the SLM microstructure and its influence on post-processing operations, such as micro-milling or LSP, is recognized.

This work therefore seeks to elucidate the degree to which 3D microstructure consideration in finite element models influences the post-process material response. To achieve this, an inhomogeneous microstructure, produced by SLM, is subjected to micro-milling and LSP simulations. In the case of micro-milling, the influences of microstructure modeling on volumetric plastic strain, temperature distributions, induced stress, and material removal (considering the effects of ductile and shear damage parameters) are examined, while for LSP, the influences of microstructure modeling on the RS magnitude and distribution are examined. Accordingly, Section 2 describes an experimentally calibrated, physics-based thermal FE simulation from which the SLM thermal history is obtained. The thermal history is then used for 3D prediction and reconstruction of a Monte Carlo-based (Johnson, Rodgers, Underwood, Madison, Ford, Whetten, Dagel, Bishop, 2018, Rodgers, Bishop, Madison, 2018) inhomogeneous microstructure RVE. Micro-milling and LSP simulations that are executed both with and without the RVE to capture inhomogeneous and anisotropic behavior are subsequently discussed in Section 2 and 3, along with a description of the material plasticity model and damage criteria. Trends in plastic strain, temperature, residual stress and material removal from simulations with and without microstructure are presented in Section 5. Finally, insights from both simulations highlighting the importance of microstructure modeling are summarized in Section 6.