Sie können Operatoren mit Ihrer Suchanfrage kombinieren, um diese noch präziser einzugrenzen. Klicken Sie auf den Suchoperator, um eine Erklärung seiner Funktionsweise anzuzeigen.
Findet Dokumente, in denen beide Begriffe in beliebiger Reihenfolge innerhalb von maximal n Worten zueinander stehen. Empfehlung: Wählen Sie zwischen 15 und 30 als maximale Wortanzahl (z.B. NEAR(hybrid, antrieb, 20)).
Findet Dokumente, in denen der Begriff in Wortvarianten vorkommt, wobei diese VOR, HINTER oder VOR und HINTER dem Suchbegriff anschließen können (z.B., leichtbau*, *leichtbau, *leichtbau*).
This study presents an experimental analysis and computational fluid dynamics (CFD) model for studying single-track formation in the Laser Powder Bed Fusion (LPBF) process, implemented in OpenFOAM. The model incorporates key physical phenomena, including Marangoni convection, recoil pressure, and phase change, to accurately capture melt pool dynamics. A stochastic powder size distribution, statistically characterized from experimental observations, is used as the initial condition to enhance the fidelity of the simulations. AlSi10Mg single-track deposits were performed using a Concept Laser M2 LPBF system, varying laser powers, scan velocities, and spot sizes. Cross-sectional microscopy and porosity analyses revealed transitions from lack-of-fusion to keyhole-induced porosity as a function of volumetric energy density (VED). Numerical results showed good qualitative and quantitative agreement with experimental observations, with relative errors in melt pool dimensions below 12\(\%\). Simulations further revealed transient flow patterns and melt pool instabilities difficult to capture experimentally. The findings demonstrate that the combination of experimental observations and simulation tools can be used to construct map process windows, predict defects, and support process optimization in LPBF, understanding the multiphysics phenomena at different building conditions, which is particularly important for high-conductivity aluminum alloys.
Saúl Piedra, Arturo Gómez-Ortega, Christian Félix-Martínez and James Pérez-Barrera contributed equally to this work.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
1 Introduction
The industrial adoption of Metal Additive Manufacturing (MAM) processes faces challenges in ensuring repeatability and defect-free parts [1], particularly in Powder Bed Fusion (PBF), due to complex physical processes and small-scale defects that can be present at the end of the fabrication process. In this regard, research has focused on understanding the Process-Structure-Property-Performance (PSPP) relationships [2] to optimize process parameters and improve reliability, although this is not easy to achieve since some technologies cover more than a hundred parameters whose effect on performance can be important [3]. While experimental methods are costly and time-consuming, modeling and simulation offer a promising alternative tool to understand the relevant physical phenomena and process parameters during manufacturing, as it can give guidance to experimental efforts in order to streamline manufacturing and reduce manufacturing defects or implement control algorithms during fabrication. However, simulating these processes involves precise modeling of the multiphysics that occur at the different length and time scales [4].
In general, the simulation of the MAM process can be divided according to different degrees of approximation, namely, residual stress estimation [5], microstructure evolution [6], powder spreading and packing [7], powder trajectories determination in material addition processes, and transport phenomena inside and around the melt pool [8]. Regarding the latter, for the Laser Powder Bed Fusion (LPBF) process, the spreading of the metallic powder, alongside the heat transfer and fluid dynamics in the melt pool, determine the appearance of defects such as porosities, lack of fusion or keyholing [9], which can result in stress concentrators, ultimately compromising the mechanical integrity of the fabricated part. In this sense, it is a big challenge to simulate the LPBF at the mesoscale, firstly due to the random nature or the spreading and particle size of the metallic powder, and secondly because these types of simulation involve solving the coupled equations of mass conservation, balance of momentum, and energy conservation in order to understand the dynamics of the melt pool and the processes of metal fusion, material addition and solidification. Additionally, the model must consider the features of the energy source provided by the laser, which is mathematically represented as a source term in the energy conservation equation, and has been modeled using different approaches [10‐12].
Anzeige
Simulation of all the physical phenomena with an impact on the final properties of the printed part is not easy, however, several efforts have been made in this regard since several papers about modeling and simulation of the LPBF process have been published. King et al. [13] summarized the studies regarding the “effective medium model”, in which the powder layer is considered as a continuous solid with varying density in order to take the interstitial space between powder particles into account. They proved their model using single tracks of 316L stainless steel and observed that this model can reproduce some features of the printed tracks. In their review, Markl and Körner [14] detailed some important features of the published papers regarding LPBF simulation up to that point. Noteworthy points of importance extracted from their work are, in general: 1) powder particles are modeled as perfect spheres, almost always having uniform size and distribution, and 2) the heat source in considered as a surface phenomenon since the penetration of the laser is small. Additionally, many of the revised papers use common metals or alloys such as Ti6Al4V and stainless steel to test the numerical models.
More recently, Cook and Murphy [15] reviewed the physics modeling and simulation inside and around the melt pool. Particularly, they focused on the different models used to realistically model the dynamics of the liquid metal, including the mushy zone existing at the liquid-solid interface and the surface tension effects to accurately reproduce the geometry of the melt pool. Similarly, most of the reviewed papers used perfect spheres as powder particles, usually using a single particle size or a simple size distribution. Interestingly, the authors concluded that two important weaknesses present in almost all the reviewed papers lie in the lack of a mesh sensitivity analysis, and poor validation work reported, mainly because in-situ observations are hard to achieve.
In another, more recent review, De Leon [4] reviewed the physics and simulation of the LPBF process spaning from the mesoscale (melt pool scale) up to part scale. They point out that key parameters in this manufacturing process are, among others, the laser spot size, power and velocity, type of alloy, powder size distribution, layer thickness and temperature of the base plate. Several of these parameters are normally combined to calculate the volumetric energy density, which is an important parameter affecting the performance of fabricated parts and must be considered to try to understand the multi physics involved in the whole process.
Aluminum alloys are materials of high value in the automotive, energy and medical industries, to name a few, due to its high-strength and features for ligthweighting compared to other metals. In the last decade, the fabrication of Al alloys in additive manufacturing technologies has become an important challenge, mainly due to its high thermal conductivity and highly reflective properties. Maamoun et al. [16, 17] fabricated Al6061 and AlSi10Mg test cubes using LPBF and studied the effect of the volumetric energy density on the surface roughness, relative density, internal porosity formation, dimensional accuracy, microstructure and mechanical properties of the as-built samples. They found the formation of microcracks in the Al6061 samples and attributed their appearance to the scaning velocity and chemical composition of the alloy, since low Si content in the Al6061 increases the coefficient of thermal expansion, resulting in hot crack formation. In this sense, the use of simulation analyses can aid in understanding the physical mechanisms and discriminate the process parameters of importance for the additive manufacturing of aluminum alloys. In their work, Bunaziv et al. [18] developed a CFD model adjustable to simulate both Laser Beam Welding (LBW) and Laser Powder Bed Fusion (LPBF). Their model uses ray tracing for the laser-matter interaction and the Volume-Of-Fluid (VOF) method to solve the moving interface between the metal and the processing atmosphere. For the LPBF case, the authors used a powder consisting of spheres with average size of 50 \(\mu\)m and random distribution to achieve 50% porosity of the bed. The numerical model was used to generate printability maps for both processes, mainly searching for the appearance of defects such as lack of porosity, keyholing and balling. These maps were compared with experimental results and previously reported observations, finding good qualitative agreement, however, the lack of experimental results due to its high time consumption and effort required, resulted in quantitative differences between the simulated and experimental results.
Anzeige
As it can be seen, the modeling and simulation of AM processes, particularly, LPBF of metals, although growing rapidly, it is still in development, specially regarding the use of high thermal conductivity and high entropy alloys. In this paper, we present the computational modeling and simulation of single track constructions of a high conductivity alloy (AlSi10Mg) during LPBF process, considering a powder bed having a particle size distribution obtained from experimental measurements. The particles are localized randomly on top of a substrate in order to approximate the spreading due to a roller or rake.
2 Mathematical and numerical models
The mathematical model governing the LPBF process is formulated based on the fundamental principles of mass conservation, momentum balance, and energy conservation. In addition, an advection equation for the volume fraction is incorporated to accurately track the interphase between the materials in the computational domain, namely inert gas, metallic powder, and molten metal. To simplify the model, the liquid metal and the inert gas are treated as incompressible fluids. As a result, the mass conservation equation reduces to:
where \({\textbf {u}}\) is the velocity vector field.
The balance of momentum equations must account for the spatial dependence of the thermophysical properties of the phases within the computational domain. Consequently, these equations can be expressed as [19]:
where \(\rho\) is the density of the fluids (liquid metal and gas), t is time, p is the scalar pressure field, and \(\varvec{\tau }\) is the stress tensor. The term \(F_g\) represents the force due to gravitational acceleration (natural convection effect), while \(S_m\) is a momentum sink term that accounts for metal solidification. Finally, \(F_s\) represents the forces due to surface tension and recoil pressure. This source term is expressed as:
The first term on the right-hand side of Eq. 3 represents the forces resulting from surface tension and recoil pressure, \(\kappa\) and \(\hat{n}\) being the curvature of the interphase and normal vector, respectively. The second term accounts for the force induced by the Marangoni effect, which arises from variations in surface tension (\(\sigma\)) due to temperature gradients within the liquid metal. The surface tension of the liquid metal was modeled as a linear function of temperature. Specifically, the surface tension was defined as
where \(\sigma _0 = 0.9~\mathrm {N/m}\) is a representative value of the liquid-phase surface tension of AlSi10Mg and \(T_{\textrm{ref}} = 870.15~\textrm{K}\) corresponds to the liquidus temperature of the alloy. The temperature coefficient of surface tension was set to \(\frac{d\sigma }{dT} = -0.5 \times 10^{-4}~\mathrm {N/(m\cdot K)}\), consistent with reported values for aluminum alloys.
This effect, along with natural convection, plays a dominant role in the dynamics of the melt pool. Therefore, incorporating these phenomena into the model is essential to achieve accurate predictions of the flow behavior within the melt pool. These terms are multiplied by the gradient of the volume fraction, \(\alpha\), which is a scalar variable used to track the interphase. Finally, the energy conservation equation in terms of temperature (T) can be expressed as
where \(c_p\) is the specific heat capacity, k is the thermal conductivity, q is the source term that represents the heat added by the laser, and \(S_h\) quantifies the latent heat due to metal phase-change (solidification and melting). In the present work, a spatially surface Gaussian distribution is used to model the heat source q [20]:
where Q is the laser power, \(\eta\) is the laser absorption rate of the metal, \(\omega\) is the beam radius, \(v_{laser}\) is the scanning velocity (in the z direction), bx and bz are the offsets of the heat source reference frame in the x and z-axis, respectively.
The mathematical model can solve the conservation equations for the entire computational domain using the one-fluid approach. However, this formulation requires a method to track the spatial and temporal evolution of the interface between fluids. In the literature, numerous methods have been developed to address interface tracking in two-phase flow problems [21]. In particular, immersed boundary methods (IBM) have gained significant popularity in recent years [22], as they enable direct and highly accurate resolution of two-phase flows. Some well-known IBM approaches include the level-set, Volume-Of-Fluid (VOF), shock-capturing, ghost-fluid, and front-tracking. The primary distinction between these methods lies in their numerical strategies for tracking and identifying the interface.
In this study, simulations were performed using the open-source software OpenFOAM [23], which also implements the VOF method, incorporating immersed boundary techniques for handling complex geometries and moving interfaces. This approach involves solving an advection equation for the volume fraction and applying a dedicated interface-capturing algorithm at each time step to accurately track the phase boundary within the computational cells. The governing equation for the volume fraction (\(\alpha\)) is expressed as:
The system of partial differential equations is numerically solved using the finite-volume method implemented in OpenFOAM-v10 libraries. The solver was adapted from the beamWeldFoam presented by Flint et al. [24], which was developed based on the pimpleFoam solver. The convective terms of the Navier-Stokes equations were discretized using a second-order upwind scheme, and the diffusive terms with a central difference scheme. The resulting linear systems from pressure, velocity, and temperature were solved with a Generalized Geometric-Algebraic MultiGrid solver. In addition, a Gauss-Seidel method is used to solve the linear system resulting from the volume fraction advection equation.
Figure 1 depicts the computational domain used for numerical simulations, which consists of an aluminum powder bed and an aluminum substrate. Additionally, an inert gas is present at the top of the computational domain, serving as a controlled atmosphere during the LPBF process. The substrate dimensions are \(800 \times 600 \times 1000 \,\mu\)m\(^3\) in the x, y and z directions, respectively. The domain was dicretized using a Cartesian mesh with 192 \(\times\) 160 \(\times\) 480 cells in the x, y and z directions, respectively.
Fig. 1
Computational domain including the powder distribution
Finally, the metal powder used in this study consists of an AlSi10Mg alloy, and argon was used as inert gas controlled atmosphere for the LBPF process. The material properties of this aluminum alloy and Argon are presented in Table 1.
Table 1
Physical properties of AlSi10Mg powder and liquid metal
Property
Symbol
Value
Density
\(\rho\)
2.58 g/cm\(^3\)
Thermal Conductivity
k
150 W/m\(\cdot\)K
Specific Heat Capacity
\(c_p\)
880 J/kg\(\cdot\)K
Latent Heat of Fusion
\(L_f\)
396 kJ/kg
Solidus Temperature
\(T_m\)
830.15 K
Liquidus Temperature
\(T_b\)
870.15 K
Thermal Expansion Coefficient
\(\alpha\)
23 \(\times 10^{-6}\) K\(^{-1}\)
Radiation emissivity
\(\epsilon\)
0.4
Viscosity (liquid metal)
\(\mu\)
1.2 mPa\(\cdot\)s
Surface Tension (liquid metal)
\(\sigma\)
0.9 N/m
3 Modeling the stochastic distribution of the powder bed
To initialize the computational simulations, it is essential to model the particle size distribution of the powder bed and define a computational domain that incorporates such a distribution. An experimental characterization of the aluminum powder particle size was conducted to determine its distribution. Figure 2a presents the particle size distribution obtained using a Mastersizer 3000 particle analyzer, resulting in a distribution range characterized by \(D_{10} = 11.5~\mu\)m, \(D_{50} = 28.0~\mu\)m, and \(D_{90} = 45.7~\mu\)m. Here, \(D_{10}\) indicates that 10% of the particles are smaller than 11.5 \(\mu\)m; \(D_{50}\) signifies that 50% of the particles are smaller than 28.0 \(\mu\)m, while the remaining 50% are larger; and \(D_{90}\) denotes that 90% of the particles are larger than 45.7 \(\mu\)m. This particle size distribution is typical for the powder bed fusion process.
Additionally, Fig. 2b displays the morphological characteristics of the powder, obtained using a field emission scanning electron microscope (FE-SEM) Jeol JMS-7200F. The image reveals that the powder predominantly consists of spherical particles with only a few small satellite particles.
Fig. 2
a Powder size distribution and b microscopy of a powder sample
Once the particle size distribution was experimentally characterized, the 3DRSP toolbox in MATLAB was used to incorporate this powder bed and initialize the CFD simulations [25]. This toolbox enables the generation of stochastic powder distributions in a three-dimensional space. Figure 1 illustrates the aluminum powder distribution obtained within the computational domain. As observed, the distribution is non-uniform, with powder packing strongly influenced by particle size variations. The use of this tool facilitated the initialization of the simulations presented in the following section, thereby completing a workflow for implementing LPBF process simulations at the mesoscopic scale.
4 Specimen configuration and fabrication process
To validate the numerical implementation, an experimental study of single tracks fabricated by the LPBF process was carried out on a Concept Laser M2 machine. The influence of key printing parameters (laser spot diameter, scan velocity, and laser power) was analyzed to assess their effect on track morphology and melt pool characteristics through the volumetric energy density (J/mm\(^3\)), calculated as
where P is the laser power (J/s), v is the scanning velocity (mm/s), th is the layer thickness (mm) and d is the spot diameter (mm).
A complete factorial design was implemented, as summarized in Table 2, to systematically assess the individual and interactive effects of the parameters on the morphology of the track. A longitudinal and cross-sectional analysis was performed using optical (OM) and scanning electron microscopy (SEM) to evaluate the continuity, width and depth of the tracks, as well as the presence of defects such as keyholing and lack of fusion. Figure 3 shows the distribution of the aluminum tracks on the fabrication plate. For each set of printing parameters, five repetitions of tracks were produced, each with a nominal length of 30 mm. It is important to highlight that not all tracks achieved proper consolidation, primarily due to a lack of fusion.
Table 2
Fabrication parameters and their values used for the factorial design
Level/factor
Spot diameter (\(\mu\)m)
Power (W)
Scan velocity (mm/s)
1
50
200
350
2
100
285
700
3
370
1050
4
1400
5
1750
6
2100
Fig. 3
Tracks distribution on the fabrication plate, five repetitions of tracks with the same printing parameters were produced
To perform microscopic and porosity analyses, the melt tracks were sectioned along the plane perpendicular to the melting direction (i.e., the track cross-section). The surfaces were prepared through successive grinding using SiC papers ranging from 180 to 2000 grit, followed by polishing with a 1 \(\mu\)m diamond paste. To reveal the geometric features (width, height, and depth) as well as potential defects, the samples were etched with Keller’s reagent. Observations were carried out using a digital optical microscope (Olympus DSX510).
Figures 4a-c show the tracks’ cross sections manufactured with low VED (70.47 J/mm\(^3\)), medium VED (457.14 J/mm\(^3\)), and high VED (845.71 J/mm\(^3\)), respectively. It is well-known that the volumetric energy density is typically used to describe the average applied energy per unit volume of material during a powder bed fusion process, which involves laser power, scan velocity, hatch distance, and layer thickness. All of these processing parameters are significant factors in the LPBF process, but complicated interaction effects can be expected if they are varied simultaneously in an experiment. In Fig. 4a, the track’s cross-section of the low VED is observed, where the geometric characteristics are minimal; that is, low height, medium depth and width, which is related to the parameters of this VED. In Figures 4b and c, it is observed that as the VED increases, the height of the tracks grows as well as the depth, to the point that the highest energy density generates keyhole defects such as porosity due to excess energy. In this case (Fig. 4c, the defect (pore) with spherical geometry can be attributed to gas bubbles generated when high laser energy is applied to the melt pool and vaporizes the liquid metal.
Fig. 4
Tracks cross-sections built with different printing parameters
The tracks’ cross-sections shown in Fig. 4d-f were manufactured keeping constant laser power (370 W) and spot size (100 \(\mu\)m), while the scan velocities were 350 mm/s, 750 mm/s, and 1050 mm/s, respectively. In Fig. 4a-c it was observed that the VED has a greater effect on depth, that is, as the VED increased, the height and the depth increased, depth having the greatest increase. On the other hand, for Fig. 4d-f, it can be seen that both the height and depth of the tracks decreased as the scan velocity increased.
Although the LPBF process offers a great advantage in the manufacture of complex parts, it is affected by many factors, one of these factors is the VED, which governs the effective energy used in the manufacturing process; but likewise, as observed in Fig. 4d-f, the variation of individual parameters also has an important effect on the appearance or not of defects, as well as in the geometric characteristics of the tracks.
Figure 5 presents an experimental process map of the VED as a function of laser power, which is divided into three regions: low laser power (200 W), medium laser power (285 W), and high laser power (370 W). As can be observed, higher VED values generally result in tracks with greater height. However, an excessively high VED, induced by elevated laser power, promotes the keyhole phenomenon, leading to the formation of large pores in the deeper regions of the melt pool. In contrast, low VED values contribute to porosity near the substrate surface as a result of insufficient fusion and inadequate powder bed packing. This experimental map demonstrates the importance of simultaneously tuning laser power and VED to avoid the extremes of Lack of fusion and keyhole porosity.
Fig. 5
Experimental process map for LPBF of AlSi10Mg built using a ConceptLaser M2 machine
Figure 6 presents the quantitative effect of VED on the geometrical characteristics of single-melt tracks, using two different laser spot sizes: 50 \(\mu\)m and 100 \(\mu\)m. In particular, in Fig. 6a, the melt region increases with VED for both laser spot sizes. Tracks produced with the smaller 50 \(\mu\)m laser spot consistently exhibit larger melt regions than those formed with the 100 \(\mu\)m spot at comparable VED. This difference stems from the higher peak power density associated with the smaller beam diameter, which concentrates the laser energy into a narrower area. As a result, localized heating is intensified, leading to steeper thermal gradients, deeper penetration, and a more pronounced melt region. In contrast, a larger spot size distributes the same energy over a wider area, resulting in lower local temperatures and more superficial melting. The fitted linear trends show a stronger sensitivity of the melt region growth with VED for the 50 \(\mu\)m spot, suggesting that the depth or volume of the melt pool can be modulated more effectively through energy input when using smaller beams.
In Fig. 6b, a similar positive correlation is observed between the melt track width and the VED for both laser spot sizes. However, in contrast to the trend in the melt region, the 100 \(\mu\)m laser spot produces consistently wider tracks than the 50 \(\mu\)m spot. This outcome aligns with expectations, as the broader beam induces a larger lateral heat-affected zone. The slope of the width-VED relationship is also steeper for the 100 \(\mu\)m spot, indicating that lateral melt pool expansion is more sensitive to increases in energy input under this condition. This behavior reflects the dominant role of lateral thermal diffusion and melt spreading in governing track width, especially under conduction-mode melting conditions typical of LPBF.
Fig. 6
Geometrical features of the tracks as functions of VED: a Melt region and b Width. Dots: experimental data. Dashed lines: linear fit
To validate the numerical model, the results were compared both qualitatively and quantitatively with the experimental observations. The selected experimental cases are presented in Table 3. Figure 7 shows a comparison between the tracks topologies obtained from the simulations and the corresponding SEM observations for the three cases listed in Table 3. As seen, there is strong qualitative agreement between the simulation and the experimental results. The yellow, red, and black dashed squares highlight regions where similar undulations are observed on the track surface in both the numerical and experimental results. Additionally, the track width obtained from the simulations is compared in Table 3 with the experimental observations, showing a good quantitative agreement with relative errors less than 11\(\%\)
Table 3
Tracks measured for CFD model validation
ID
Laser
Spot
Scan
Experimental
Simulated
power
diameter
velocity
width
width
(W)
(\(\mu\)m)
(mm/s)
(\(\mu\)m)
(\(\mu\)m)
1
370
50
350
349
363
2
370
100
700
267
291
3
370
50
1750
145
162
Fig. 7
Track topology comparison for case 3. Left panel: simulation, right panel: experimental SEM observation
The cross-section obtained from optical microscopy for the track fabricated with a volumetric energy density (VED) of 108.57 J/mm\(^3\) is shown in Fig. 8a). Figure 8b) presents the corresponding numerical results, where the upper panel shows the final solidified track profile after the laser has moved away from the region, allowing a direct comparison of the track height, while the lower panel shows the melt region formed during laser-material interaction. The comparison of the final track geometry indicates that the simulated track height (34.5 \(\mu\)m) is in reasonable agreement with the experimental measurement (45.1 \(\mu\)m), capturing the overall shape of the solidified track. The remaining discrepancy is attributed to the strong sensitivity of track height to powder-bed stochasticity, local particle arrangement, and melt pool collapse during solidification. In contrast, the melt region formed during laser interaction (lower panel in Fig. 8b) shows a closer correspondence with the experimentally observed melt penetration depth, with a deviation of approximately 12%.
Fig. 8
a Optical microscopy cross-sectional image; b Top panel: Final solidified cross-section track profile after the laser has passed, Bottom panel: XY cross-section of the melt region (VED=108.57 J/mm\(^3\))
In addition to melt pool geometry, the capability of the numerical model to predict defect formation was also evaluated. In particular, the occurrence of keyhole-mode melting at high volumetric energy density was evaluated by directly comparing numerical results with experimental cross-sectional observations obtained under the same processing conditions. Figure 9 compares the simulated cross-section of a track with the corresponding optical microscopy observation for a specimen fabricated at VED = 845 J/mm\(^3\). The simulation predicts the formation of a deep and narrow vapor cavity, characteristic of keyhole melting, which is consistent with the keyhole-induced pore observed experimentally. This qualitative agreement demonstrates that the numerical model not only reproduces melt pool dimensions but also captures the transition to keyhole formation at large VED values.
These qualitative and quantitative comparisons confirm that the computational simulations accurately reproduce the formation of aluminum tracks in different printing parameters.
Fig. 9
Comparison between numerical simulation and experimental cross-section illustrating keyhole formation at VED = 845 J/mm\(^3\))
5.3 Power and Scan velocity effects on the tracks topology
Once the model is validated, the impact of laser power and scan velocity on the formation of melt tracks is examined. The primary objective is to investigate how variations in energy input affect melt pool width and surface topology, thereby revealing transitions between the conduction and keyhole regimes.
Figure 10 presents a comparison of simulated melt track morphologies and thermal fields for two laser power levels for a constant scanning velocity of 700 mm/s and laser spot size of 100 \(\mu\)m. The upper panels depict top views of the temperature field and surface morphology during laser scanning, while the lower panels show longitudinal cross-sections of the melt pool region, highlighting the geometric characteristics of the molten zone.
For the 200 W case (left column), the melt pool assumes a relatively shallow and semi-elliptical profile, with a limited penetration depth and lateral spread. The top view reveals a track with regularly spaced ripples along the scan direction. These ripples are indicative of melt pool surface instabilities, which arise from the interaction of Marangoni convection, recoil pressure fluctuations, and the cyclic nature of melting and solidification at lower power levels. The energy input is insufficient to maintain a stable melt pool, leading to recurrent collapses of the molten surface and incomplete coalescence of adjacent powder particles. In contrast, the 370 W case yields a substantially deeper and wider melt pool with a keyhole-shaped geometry, characteristic of deep-penetration melting. The top view shows a much smoother and more continuous surface morphology. The higher laser power leads to increased energy absorption, enhancing melt pool stability through more efficient convection and prolonged melting.
Fig. 10
Simulation results of single tracks deposited using a laser spot diameter of 100 \(\mu\)m and scan velocity of 700 mm/s, for two laser powers: left panels: \(P=\)200 W and right panels: \(P=\)370 W. In the top panels, the color map represents the temperature scalar field, whereas the red region in the bottom panels indicates the melted region
Figure 11 highlights how the scan velocity affects both surface topology and melt pool geometry by keeping the spot size and laser power constant, with values of 100 \(\mu\)m and 370 W, respectively. The top images present isometric views of the melt tracks with temperature fields, while the bottom images display cross-sections of the melt pool regions, where the red volume denotes the molten material.
At 1050 mm/s (left panels of Fig. 11), the melt pool exhibits a deep and well-developed keyhole shape, with significant penetration into the substrate and lateral fusion with surrounding powder particles. The isometric view shows a relatively smooth and continuous melt track with moderate ripples and signs of internal flow disturbances. These characteristics suggest that the energy input per unit length is sufficient to maintain a stable melt pool, resulting in effective melting, good powder wetting, and reduced porosity risk.
Fig. 11
Simulation results of single tracks deposited using a laser spot diameter of 100 \(\mu\)m and power of 370 W for two different scan velocities: left panels, \(v=\)1050 mm/s and right panels, \(v=\)2100 mm/s. In the top panels, the color map represents the temperature scalar field, whereas the red region in the bottom panels indicates the melted region
In contrast, at 2100 mm/s (right panels of Fig. 11), the energy input is reduced as the laser moves faster, leading to a shallower and more parabolic melt pool with relative symmetry in the cross-sectional profile. This shape is typical of a conduction-dominated regime, where the reduced residence time limits keyhole formation and favors radial heat conduction. However, despite the more regular melt pool shape, the surface track exhibits stronger ripples, as revealed in the isometric view, due to higher cooling rates, rapid solidification, and dynamic surface instabilities. The surface reveals a segmented and undulated morphology, with localized bulging and necking of the track, which are typical precursors to ball formation. The reduction in volumetric energy density leads to a shallower, less stable melt pool and incomplete wetting of the surrounding powder particles. Under these conditions, surface tension forces begin to dominate over melt pool inertia, triggering capillary instabilities and promoting the breakup of the melt track into discrete beads. These fluctuations, while not leading to full balling, pose a risk to track continuity and interlayer adhesion in subsequent passes. Overall, the results demonstrate that increasing scan velocity transitions the melt pool behavior from keyhole-driven and asymmetric shaping to a symmetric conduction-driven mechanism, but at the cost of increased surface instability. This highlights the trade-off between geometric regularity in the melt pool and surface quality in high-speed LPBF processing.
5.4 Fluid flow and mass-heat transfer analysis
The most valuable insights provided by the simulations are the evolution of the melt pool and the mechanisms responsible for defect formation. This section analyzes the flow behavior during both conduction and keyhole melting regimes. Figure 12 illustrates the flow and heat transfer within the melt pool for a low VED (169.14 J/mm\(^3\)) case at two distinct time instants during the laser scan, showing the evolution of melt pool dynamics. The top row (a–b) displays the velocity field magnitude and direction, the middle row (c–d) shows the temperature field, and the bottom row (e–f) presents the volume fraction field, where the formation and distribution of process-induced porosity are evident.
Fig. 12
Velocity field (a–b), temperature field (c–d), and volume fraction (e–f) at two time instants under low VED conditions (VED=169.14 J/mm\(^3\)).
At the first instant (left column: a, c, e), the velocity field (a) shows strong localized recirculation zones and high-velocity jets directed upward (ahead of the laser beam) and backward (behind the power source), primarily governed by surface tension forces. This flow destabilizes the melt pool surface and contributes to spatter and ripples formation. The corresponding temperature field (c) shows a highly localized hot spot beneath the laser source, surrounded by steep thermal gradients, indicating that the thermal input is concentrated but may be insufficient to stabilize the entire melt pool volume. The volume fraction field (e) reveals the presence of small entrapped gas voids and lack-of-fusion porosity, especially beneath the powder bed and at the melt pool boundary. These pores result from insufficient melting and irregular melt pool collapse, common under low VED conditions.
At the second time instant (right column: b, d, f), the melt pool has moved forward with the laser scan. The velocity field (b) continues to exhibit strong recirculation zones, though the flow appears more elongated and directed toward the trailing edge. The persistence of high-velocity gradients reflects ongoing melt pool dynamics, which can sustain the micropores formation near the substrate. In the corresponding temperature field (d), the thermal front advances but remains highly localized, reinforcing the limited heat penetration and narrow melt region typical of low VED processing. The volume fraction field (f) at this stage highlights the growth and transport of pores. The observed porosity is attributed to incomplete melting, trapped gas in partially fused powder, and flow instabilities that prevent uniform solidification.
In contrast, Fig. 13 shows the evolution of melt pool behavior in a longitudinal cross-section of a laser scan track under high VED conditions. Again, the top row (a–b) displays the velocity magnitude and vector field at two distinct time instants, the middle row (c–d) illustrates the temperature distribution, and the bottom row (e–f) shows the volume fraction field, representing the distribution of phases and highlighting pore formation in the melt track.
Fig. 13
a) Velocity field (a–b), temperature field (c–d), and volume fraction (e–f) at two time instants under high VED conditions (VED=422.85 J/mm\(^3\))
At the first time instant (left column: a, c, e), the velocity field (a) exhibits highly dynamic melt pool behavior with intense recoil pressure-driven flow and Marangoni convection. The flow structure includes upward and rearward jets, which are typical of keyhole-mode melting, where the vapor recoil pressure pushes molten material outward. The velocity magnitudes show strong convective mixing within the melt pool. The corresponding temperature distribution (c) reveals a deep, localized hot region, surrounded by sharp thermal gradients. The melt pool is well-developed, extending both vertically and horizontally, indicative of a keyhole-shaped profile. While such conditions are often associated with improved melting and deeper penetration, the simulation results reveal the onset of keyhole-induced macroporosity. The volume fraction fields (e, f) show the formation of large, irregular voids at the bottom and trailing edge of the melt pool. These macropores originate from unstable keyhole collapse, where the vapor cavity becomes highly elongated and unstable due to excessive energy input, however, it appears that, after the first elongated pore is formed, spherical pores are formed almost periodically.
At the second time instant (right column: b, d, f), the velocity field (b) remains highly active, with strong flow now shifted toward the trailing edge of the melt pool. This change reflects the oscillatory nature of the keyhole and ongoing interactions between vapor pressure, melt recoil, and surface tension. The vortex formation suggests a localized collapse of the vapor cavity, a key mechanism for the formation of larger internal pores almost periodically. The temperature field (d) shows that the melted region remains sharply defined but displaced as the laser progresses. In the volume fraction field (f), the extent of porosity is more pronounced than in the earlier instant, with the appearance of large, irregular macropores in the melt track. Unlike lack-of-fusion defects, these pores originate from within a fully melted region and represent a critical form of process-induced porosity in LPBF. The results highlight the role of momentum and energy transport mechanisms in defect formation and emphasize the need for an optimized energy input to balance fusion efficiency and porosity control, especially for high-conductivity metal alloys as AlSi10Mg.
6 Conclusions
This study presents a combined computational and experimental investigation of melt pool dynamics and defect formation during laser powder bed fusion (LPBF) of AlSi10Mg, with a focus on fluid mechanics and porosity formation. A high-fidelity CFD model was implemented and validated using experimental data, enabling a detailed investigation into the effects of process parameters, especially the volumetric energy density (VED), on melt pool behavior and defect formation. The main findings from both simulation and experimental efforts are summarized below:
Experimental results confirmed that VED governs track morphology. Increasing the VED leads to deeper and taller melt pools, with a transition from shallow, poorly fused tracks to deeper penetrations.
A strong influence of laser spot size on the melt pool geometry width was found experimentally. Tracks fabricated with a 50 \(\mu\)m spot exhibited deeper and more concentrated melt regions than those with a 100 \(\mu\)m spot for the same VED, due to higher power density. In contrast, the track width was wider for the larger spot size, as expected from the increased lateral heat diffusion. In addition, it was found that the melt region depth is not directly influenced by the spot size.
On the computational side, the implemented CFD model in OpenFOAM accurately captured the fluid flow, heat transfer, and morphology evolution in the melt pool. The model incorporated realistic powder bed distributions and physical phenomena such as Marangoni convection and recoil pressure. Simulated velocity and temperature fields revealed the transition from conduction-mode melting to keyhole-mode melting with increasing VED.
The numerical results exhibited good agreement with experimental observations in terms of melt pool geometry, track morphology and width. Relative errors in quantitative measurements were below 11\(\%\), confirming the model’s reliability and predictive power.
High scan speeds promote conduction-dominated melting, causing surface rippling and balling precursors. Low laser power necessitates precise VED control to avoid discontinuities, while excessive power destabilizes keyholes despite deeper penetration.
The melt pool behavior is strongly dependent on VED. At low VED, incomplete fusion occurs due to insufficient energy input, leading to the formation of micropores. At high VED, excessive energy input promotes deep keyhole formation, which becomes unstable and results in macroporosity due to keyhole collapse.
The velocity field analysis reveals that the melt pool dynamics are governed by surface tension, Marangoni convection, and recoil pressure. In the low-VED regime, surface tension effects dominate, resulting in weaker convective flow and reduced melt pool penetration. In contrast, at high VED, strong Marangoni-driven circulation and recoil pressure enhance the melt depth but also contribute to keyhole instability.
The simulation framework provides access to detailed, time-resolved information on melt pool geometry, velocity fields, temperature gradients, and gas entrapment phenomena, features that are extremely challenging to capture experimentally during the process. This enables not only the identification of melt pool instabilities but also, on the future, the systematic exploration of the process parameter space to delineate safe operating windows.
Acknowledgements
Saúl Piedra and Arturo Gómez-Ortega acknowledge the support given by SECIHTI (Mexico) through the “Investigadoras e Investigadores por México” program.
Declarations
Competing interests
The authors have no conflicts of interest to declare that are relevant to the content of this article.
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Bayat M, Klingaa CG, Mohanty S, De Baere D, Thorborg J, Tiedje NS, Hattel JH (2020) Part-scale thermo-mechanical modelling of distortions in laser powder bed fusion – analysis of the sequential flash heating method with experimental validation. Addit Manuf 36:101508. https://doi.org/10.1016/j.addma.2020.101508CrossRef
2.
Hashemi SM, Parvizi S, Baghbanijavid H, Tan ATL, Nematollahi M, Ramazani A, Fang NX, Elahinia M (2022) Computational modelling of process–structure–property–performance relationships in metal additive manufacturing: a review. Int Mater Rev 67(1):1–46. https://doi.org/10.1080/09506608.2020.1868889CrossRef
De Leon E, Riensche A, Bevans BD, Billings C, Siddique Z, Liu Y (2025) A review of modeling, simulation, and process qualification of additively manufactured metal components via the laser powder bed fusion method. J Manuf Mater Process 9:33. https://doi.org/10.3390/jmmp9010022CrossRef
5.
Megahed M, Mindt H-W, N’Dri N, Duan H, Desmaison O (2016) Metal additive-manufacturing process and residual stress modeling. Integr Mater Manuf Innov 5:61–93. https://doi.org/10.1186/s40192-016-0047-2CrossRef
6.
Wei HL, Mukherjee T, Zhang W, Zuback JS, Knapp GL, De A, DeBroy T (2021) Mechanistic models for additive manufacturing for metallic components. Prog Mater Sci 116:100703. https://doi.org/10.1016/j.pmatsci.2020.100703
7.
King WE, Anderson AT, Ferencz RM, Hodge NE, Kamath C, Khairallah SA, Rubenchik AM (2015) Laser powder bed fusion additive manufacturing of metals: physics, computational, and materials challenges. Appl Phys Rev 2(4):041304. https://doi.org/10.1063/1.4937809CrossRef
Sanaei N, Fatemi A (2021) Defects in additive manufactured metals and their effect on fatigue performance: A state-of-the-art review. Progess Mater Sci 117:100724. https://doi.org/10.1016/j.pmatsci.2020.100724CrossRef
Unni AK, Vasudevan M (2021) Determination of heat source model for simulating full penetration laser welding of 316 ln stainless steel by computational fluid dynamics. Mater Today Proc 5:4465–4471. https://doi.org/10.1016/j.matpr.2020.12.842. 2nd International Conference on Aspects of Materials Science and Engineering (ICAMSE 2021)
12.
Nascimento EJG, Santos Magalhães E, Santos Paes LE (2023) A literature review in heat source thermal modeling applied to welding and similar processes. Int J Adv Manuf Technol 126:2917–2957. https://doi.org/10.1007/s00170-023-11253-zCrossRef
13.
King W, Anderson AT, Ferencz RM, Hodge NE, Kamath C, Khairallah SA (2014) Overview of modelling and simulation of metal powder bed fusion process at Lawrence Livermore national laboratory. Mater Sci Technol 31:957–968. https://doi.org/10.1179/1743284714Y.0000000728CrossRef
Mamoun AH, Xue YF, Elbestawi MA, Veldhuis SC (2018) Effect of Selective laser melting process parameters on the quality of Al alloy parts: powder characterization, density, surface roughness, and dimensional accuracy. Materials 11:2343. https://doi.org/10.3390/ma11122343CrossRef
17.
Mamoun AH, Xue YF, Elbestawi MA, Veldhuis SC (2018) The effect of selective laser melting process parameters on the microstructure and mechanical properties of Al6061 and AlSi10Mg. Materials 12:12. https://doi.org/10.3390/ma12010012CrossRef
18.
Bunaziv I, Hovig EW, Brizuela OEG, Zhang K, Ma X, Ren X, Eriksson M, Skjetne P (2024) CFD modeling for predicting imperfections in laser welding and additive manufacturing of aluminum alloys. J Laser Appl 36:032010. https://doi.org/10.2351/7.0001401CrossRef
19.
Flint TF, Anderson MJ, Akrivos V, Roy MJ, Francis JA, Vasileiou A, Smith MC (2022) A fundamental analysis of factors affecting chemical homogeneity in the laser powder bed fusion process. Int J Heat Mass Transf 194:122985. https://doi.org/10.1016/j.ijheatmasstransfer.2022.122985CrossRef
20.
Flint TF, Dutilleul T, Kyffin W (2023) A fundamental investigation into the role of beam focal point, and beam divergence, on thermo-capillary stability and evolution in electron beam welding applications. Int J Heat Mass Transf 212:124262. https://doi.org/10.1016/j.ijheatmasstransfer.2023.124262CrossRef
21.
Prosperetti A, Tryggvason G (2007) Computational methods for multiphase flow. Cambridge PressCrossRef
22.
Tryggvason G, Scardovelli R, Zaleski S (2011) Direct numerical simulations of gas-liquid multiphase flows. Cambridge Press
Flint TF, Parivendhan G, Ivankovic A, Smith MC (2022) beamWeldFoam: numerical simulation of high energy density fusion and vapourisation-inducing processes. SoftwareX 18:101065. https://doi.org/10.1016/j.softx.2022.101065CrossRef
Die im Laufe eines Jahres in der „adhäsion“ veröffentlichten Marktübersichten helfen Anwendern verschiedenster Branchen, sich einen gezielten Überblick über Lieferantenangebote zu verschaffen.
