This thesis discusses a model for simulation of the Powder Bed Fusion (PBF) process of metallic powder with the capability to become amorphous. The temperature field in the PBF process is predicted by a three-dimensional thermal finite element model in three dimensions using a layer-by-layer approach, meaning that the scanning strategy of the moving laser spot is consolidated into a single heat source acting on the entire layer momentarily. This temporal reduction enables simulations of large domains and many layers while it becomes less computational demanding compared to a detailed transient model that incorporates a scanning sequence. Predictions of the amorphous and crys- talline phase fractions are performed with a phase model coupled to the temperature field simulation. The phase model is based on differential scanning calorimetry measure- ments and optimized to fit continuous heating transformation into a crystalline state of an amorphous sample. The simulations are performed on the commercial available glass forming alloy AMZ4.

Bulk Metallic Glass (BMG) have an amorphous structure and possesses desirable me- chanical, magnetic and corrosion properties such as high yield stress, low magnetic losses and high corrosion resistance. Glass forming alloy has the potential to become amorphous provided that the solidification rate is rapid enough to avoid crystallization. However, traditional manufacturing techniques, such as casting, limits the cooling rates and size of components due to limited heat conduction in the bulk. With Additive Manufacturing (AM) on the other hand, it is possible to produce BMG’s as the melt pool is very small and solidification can be achieved very rapid to bypass crystallization. Yet, crystals may form by devitrification (crystal formation upon heating of the amorphous phase) because of thermal cycling in previous layers. Simulation of the process will aid the understanding of glass formation during AM and the development of process parameters to control the level of devitrification. 

In this thesis, finite element (FE) simulations of additive manufacturing (AM) and physically based material modeling are presented. AM is a process where the component is built layer-wise. The material undergoes repeated heating and cooling cycles when layers are added, which may result in undesired deformation and residual stress in the built component. The choice of process parameters and scan strategy affect the resulting residual stress. Simulations can be used to support the experimental determination of process parameters and scan strategy. AM processes often comprise many added layers, and the passes are lengthy relative to their thicknesses and widths. This makes the FE simulations computationally expensive, with many elements and time steps. In this work, AM processes have been simulated with the FE-method using a lumping technique. This technique allows fewer time steps and a coarser mesh. Thermal behavior, deformation, and residual stresses have been simulated and compared with experiments. The simulations show that, by using the lumping technique, the computational effort can be reduced significantly with retained accuracy for the resulting temperature and deformations. The residual stresses become somewhat smaller. Alloy 625 is a nickel-based superalloy used in high-temperature applications owing to the hightemperature strength. The material is difficult to manufacture by conventional machining owing to excessive tool wear and low material removal rates. Thus alloy 625 is a material appropriate for the AM technology with its near-net shape potential. An existing, physically based flow stress model has been further developed to fit the mechanisms typical for alloy 625. This model gives an accurate mechanical behavior and capture viscoplasticity, creep, and relaxation. The physically based model has been calibrated versus compression tests and validated with a stress relaxation test performed in a Gleeble 3800 machine. The predicted relaxation was in good agreement with the measured relaxation. The usage of this kind of material model is expected to improve the prediction of the material behavior during the AM process and, thereby, the overall prediction of the AM process.

The work presented in this thesis aims to develop a modelling approach to predict crystalline phase evolution in bulk metallic glass during additive manufacturing with laser-based powder bed fusion (PBF-LB). Metallic glasses are non-crystalline metallic materials that generally possess exceptional properties because of its amorphous struc-ture. Manufacturing of metallic glass is possible by rapid cooling of a liquid metal alloy. The required cooling rates to avoid crystallisation generally limits traditional manufac-turing techniques to small/thin samples. The desirable properties of metallic glasses motivate manufacturing of larger samples. PBF-LB is one promising method by which bulk metallic glass potentially can be produced without size limitation. Cooling rates in this process are generally several orders of magnitude higher than critical cooling rates to bypass crystallisation in glass forming alloys. Crystalline structures may still evolve within the solid material because of thermal cycling during the manufacturing process. Numerical simulation can assist development of process for bulk metallic glass formation by predicting the phase evolution. Simulations can also help to increase the understand-ing of where and when crystalline structures develop with respect to process parameters and scanning strategy. Simulation of bulk metallic glass formation during PBF-LB is a challenge. The thermodynamic and kinetic properties of the material and the large variations in time and length scales in the process makes accurate simulations difficult. This thesis address these challenges by developing a modelling approach for simulation of the temperature history and crystalline phase evolution. The objective is to assist the development of process parameters for bulk metallic glass formation. The approach includes finite element modelling to compute the temperature history in the heat affected zone. The modelling includes approximations of the energy input and approaches to sim-ulate the large variations in time and length scales associated with PBF-LB. Computed temperature histories acts as input in calculations of the crystalline phase evolution in the metallic glass. The phase transformation modelling approach includes a modified isothermal model and classical nucleation and growth theory. The result is a coupled thermal and phase transformation model that can predict the trend in crystalline phase evolution in a bulk metallic glass with respect to the process parameters. The predictions show very good agreement to experimental estimates of the crystalline volume fraction. Comparison of simulations makes it possible to evaluate the process parameters in terms of crystalline size distribution. The model is a powerful tool that help the development and fine tuning of process parameters to produce bulk metallic glass.

This thesis presents finite element (FE) simulations of additive manufacturing (AM) and physically based material modeling of alloy 625 and alloy 718. In recent years, there has been an increasing interest in AM and there has been a dramatic increase in publications in the field. AM can be beneficial compared to conventional manufacturing methods in many applications. The method offers short component lead times and large design freedom with the possibility to create complex components. Alloy 625 and alloy 718 are nickel-based superalloys used in high-temperature applications owing to their high-temperature strength. The materials are difficult to manufacture by conventional machining due to rapid tool wear and low material removal rates. Thus, the alloys are appropriate for the AM technology with its near-net shape potential.Owing to the rapid heating and solidification in the AM process, residual stresses are induced in the component. This is a well-known problem and causes distortion of the samples when removing them from the build plate. The residual stresses may also deteriorate the fatigue properties. It is important for the manufacturer to understand how the choice of process parameters and scanning strategy affect the residual stresses to minimize those and improve the quality of the components. Simulation can be used as a tool while developing the process parameters and support the experimental efforts. FEM is generally the preferred method for simulation of deformations and residual stresses in AM. The simulation technique used when modeling AM has its origin from welding simulations that was performed already since the beginning of 1970. However, it is not possible in practice to simulate an AM process in the traditional way due to a large number of elements and time increments to be calculated. This is especially true for the laser-based powder bed fusion (PBF-LB) process where the process of a full-scale part may comprise many thousands of added layers, and the passes are lengthy relative to their thicknesses and widths.The aim of this thesis work is to develop FE simulation techniques that reduce the computational effort when modeling residual stresses in AM processes to enable simu-lation of full-scale parts. This has been done with thermo-mechanical FE-models using different lumping techniques e.g., lumping of layers and lumping of hatches. Lumping of layers and hatches means that several physical layers, or several physical hatches, are merged and added in one modeled layer or hatch respectively. Lumping allows fewer time steps and a coarser mesh which reduces the computational effort. An existing mechanism based flow stress model has been developed to fit the mechanisms typical for alloy 625 and alloy 718 and implemented in the FE model. Also, synchrotron X-ray diffraction was performed to measure the residual stress for comparison with the models. The stress was extracted from the diffraction data using the full Debye ring fitting method.In this work, using the lumping techniques described above, it was possible to simu-late AM processes with up to physical 1500 layers. For different process parameter sets and scan strategies, thermal behavior, deformation and residual stresses have been mod-eled and compared with experiments. Using the lumping of layer technique resulted in modeled residual stresses showing the same trend as measured stresses from synchrotron X-ray diffraction for two different process parameter sets. Utilizing lumping of hatches, the resulting deflection in a part was modeled successfully for different scanning strate-gies. In the modeling, the larger deflection was seen for the samples printed with the scanning direction parallel to the long-side which was also shown experimentally.The results in this work shows that the presented lumping approaches are promising when it comes to modeling of the deformations and residual stresses in AM. Using lumping approaches, it is also possible to simulate different scanning strategies for processes of larger parts. The description of the mechanical behavior of the material is improved, using the mechanism based material model, compared to when the flow stress was modeled with tabulated data, since it takes mechanisms as viscoplasticity and stress relaxation into account. The mechanism based model includes microstructural information as grain size and solutes and can thus more easily be combined with a microstructure model. The combination of the mechanism based material model and the use of lumping techniques is thus an advance in the development of predictive models of the AM process.