Page 38 - MSAM-4-3

Materials Science in Additive Manufacturing Numerical simulation of plasma WAAM for Ti-6Al-4V

Table 5. Heat source parameters determined by measurement and simulation
Process Measured values (mm) Calculated values (mm) Relative error (%)
2bexp dexp φexp 2bsim dsim φsim 2 b d φ
e r e r e r
Preheating 6.34 1.56 4.06 8.00 3.60 2.22 26.18 42.31 -45.32
Single bead 8.88 2.68 3.31 8.80 2.4 3.67 -0.90 -10.45 10.88
Single bead+pre-heating 9.45 3.12 3.03 9.8 4.0 2.45 3.70 28.21 -19.14

Table 6. Computational time comparison between different 4.2. Thermal boundary conditions
models
In this study, the thermal boundary conditions, namely,
Model Adaptive Parallelization Computational emissivity ε, convective heat transfer coefficient h , and
c
mesh time (h) contact heat transfer coefficient a, were assumed to be
refinement temperature-independent. This simplification is adopted
Calibration Yes Yes ~ 1 to reduce both experimental effort and computational
pre-heating complexity. However, it introduces inherent limitations
Calibration single Yes Yes ~ 4 that can affect the accuracy of the thermal model. In
bead practice, these parameters exhibit strong temperature
Calibration single No Yes ~ 2 dependence and are affected by the dynamic nature of
bead+pre-heating the heat transfer during the WAAM process. Accurately
characterizing their variation over a broad temperature
calibration techniques for heat source modeling in WAAM. range requires detailed, process-representative
To improve heat source accuracy, in-process measurements, measurements, which are often challenging to obtain
such as high-speed camera monitoring of the weld pool experimentally. By assuming these parameters as
shape, can be incorporated into the calibration procedure. temperature-independent, this simplification can result
34
A study by Guo et al. introduced a convolutional neural in localized discrepancies in the predicted temperature
43
network-based method to identify heat source parameters fields, particularly in regions with steep thermal
from the cross-sectional profile of the weld zone, providing a gradients or changing surface conditions. Given the wide
novel approach to heat source calibration. Ilani and Banad temperature range inherent in welding and the plasma-
44
presented a data-driven computational approach using based WAAM process, incorporating temperature-
Goldak’s semi-ellipsoidal heat source model to predict melt dependent boundary conditions becomes essential for
pool geometry in metal AM. enhancing model accuracy. Future work should consider
integrating such formulations. As demonstrated by
This comparative discussion underscores that while Tröger et al., the inclusion of temperature-dependent
47
reduced-order, semi-analytical, and geometric models boundary conditions significantly improves the reliability
provide advantages in computational efficiency or of numerical predictions in WAAM simulations.
design utility, the present study contributes a robust,
experimentally validated framework capable of supporting 4.3. Multi-layer welding
in-depth analysis of process-induced thermal and While the present study focuses on single-track deposition
structural phenomena in WAAM.
in plasma-based WAAM, understanding how these
Structural welding simulations are invaluable for conditions evolve in multi-layer builds is essential. In multi-
predicting and optimizing geometric deviations and layer WAAM processes, previously deposited layers are
accumulated residual stresses before experimentation. subsequently remelted and reheated during the deposition
Depending on the complexity of the models, simulation of each new layer, resulting in complex thermal cycles that
times can vary significantly, ranging from a few minutes to strongly influence both the material’s microstructure and
several days. A comparison of the computational times macrostructure. The macrostructure and microstructure
8,45
used for the models is presented in Table 6. Simulations are key determinants of the mechanical properties of
were performed on an eight-core 3.30 GHz i75820K WAAM-produced components. Understanding these
48
processor with 32 GB of installed random access memory. remelting effects is crucial when extrapolating single-bead
All models have been computed using parallelization, that observations to multi-track, multi-layer components. The
is, the computation of a simulation is done in parallel using evolution of the melt pool geometry, thermal gradients,
multiple central processing unit cores. 46 and cooling rates across successive layers plays a decisive

Volume 4 Issue 3 (2025) 11 doi: 10.36922/MSAM025140021