<< Volver atrás

Tesis:

Computational Modeling of Powder Bed Fusion Manufacturing of Metals

  • Autor: ELAHI, Seyed Mohammad

  • Título: Computational Modeling of Powder Bed Fusion Manufacturing of Metals

  • Fecha: 2022

  • Materia:

  • Escuela: E.T.S. DE INGENIEROS INDUSTRIALES

  • Departamentos: INGENIERIA MECANICA

  • Acceso electrónico: https://oa.upm.es/72184/

  • Director/a 1º: ROMERO OLLEROS, Ignacio
  • Director/a 2º: TOURRET, Damien

  • Resumen: Fusion-based additive manufacturing of metal is particularly challenging to simulate due to the broad range of length/time scales of multiphysics phenomena to bridge. Integrated Computational Materials Engineering (ICME) offers a promising route to accelerate the design of new materials, optimize manufacturing processes, and thus improve material properties and performance. ICME heavily relies on linking multiple modelling techniques relevant to distinct length/time scales and/or different physics. While a wide range of physics-based and predictive models exist, the efficient linking of these models remains challenging. In the context of additive manufacturing, interdependent phenomena across scales include: multicomponent alloys with complex phase diagrams, multiple phase transformations and temperature-dependent properties; complex thermal exchanges in the vicinity of the melt pool; and the development of nonequilibrium grain microstructures within this melt pool. A multiscale modeling strategy (i.e. micro and macro) has been developed to simulate powder-bed fusion of metallic alloys, which combines: (1) Temperaturedependent alloy properties and phase diagrams calculated by CalPhaD, (2) macroscale thermal and thermo-mechanical simulations of the material addition and fusion using finite elements (FE), (3) Phase-field (PF) simulations of the melt pool’s solidification at the microscopic scale, and (4) Simulations of the microscopic solidification of the melt pool using cellular automaton (CA). Realistic processing parameters are used in the methodology to simulate selective laser melting (SLM) of an Inconel 718 alloy. In the first step, three-dimensional thermal and thermo-mechanical frameworks were developed – using an in-house finite element code (IRIS) and a material library (MUESLI) implemented in C++ programming language – by considering these four factors: (1) Representing the thermal input from the scanning laser, (2) A material model to specify the temperature-dependent material property in the cyclic heating and cooling environment, (3) A physics-based method to explain the layer build-up process, and (4) A laser material interaction model to account for transient thermo-mechanical phenomena. Then, the thermal model was coupled with CalPhaD and PF methods to predict the formation of microstructures at the scale of the entire melt pool. In this part, the effect of temperature-dependent properties and the necessity of taking into account differences in properties between the powder bed and the dense material have been discussed. We perform an appropriately-converged PF solidification simulation at the scale of the entire melt pool using a two-dimensional longitudinal slice of the thermal field calculated through FE simulations. This calculation has over one billion grid points, yet is performed on a single cluster node with eight graphics processing units (GPUs). These microscale simulations, with a level of detail down to individual dendrites, offer new insight into the selection of grain texture via polycrystalline growth competition under realistic SLM conditions. Finally, comparisons between phase-field (PF) and cellular automaton (CA) simulations of polycrystalline growth in a two-dimensional melt pool under additive manufacturing (powder-bed fusion) relevant conditions have been performed. The generated grain structures from local (point-by-point) measurements and grain orientation distributions that were averaged over various simulations have been compared. We investigated how the melt pool aspect ratio and the CA spatial discretization level affected the selected grain texture. Our simulations demonstrate that only PF simulations are capable of capturing the fine-grained microscopic features related to transient growth conditions and solid-liquid interface stability, such as the initial planar growth stage before its cellular/dendritic destabilization or the early elimination of unfavorably oriented grains due to neighbor grain sidebranching. CA grid refinement can only partially address the resulting disagreement between PF and CA predictions. In addition, with some variability on the CA grid and melt pool shape, overall grain distributions averaged throughout the entire melt pools of several simulations appear to lead to a noticeably better agreement between PF and CA. This research offers a helpful step in that direction by quantitatively comparing both approaches at process-relevant length and time scales, even though more effort is still needed, especially to identify the appropriate selection of CA spatial discretization and its link to characteristic microstructural length scales.