Computational Investigation of Synchronized ... - ETH E-Collection

Available online at www.sciencedirect.com

ScienceDirect Physics Procedia 83 (2016) 899 – 908

9th International Conference on Photonic Technologies - LANE 2016

Computational Investigation of Synchronized Multibeam Strategies for the Selective Laser Melting Process Thorsten Heeling୹,*, Konrad Wegener୽ ࢟Institute of Machine Tools and Manufacturing, ETH Zurich, Technoparkstr. 1, 8005 Zurich, Switzerland ࢣInstitute of Machine Tools and Manufacturing, ETH Zurich, Leonhardstr. 21, 8092 Zurich, Switzerland

Abstract The selective laser melting process features a nearly incomparable freedom of design. But its potential is still limited due to remaining porosity, cracking, distortion, low build-up rates and a limited range of materials. While there is some progress in process control and multiple parallel scan fields to tackle these issues, the potential of synchronized multibeam strategies has not yet been investigated. The presented synchronized multibeam approach is characterized by two widely overlapping scan fields fed by two independent laser sources that can be controlled to work in a synchronized manner with or without a defined offset. This allows a selective manipulation of the local temperature field and thus of melt pool dynamics, the temperature gradients and cooling rates, which are all influencing the processes’ porosity, cracking and distortion behavior. Therefore the influences of these strategies on the melt pool dimensions and dynamics as well as the temperature gradients are investigated in this work. © 2016 The line Authors. by Elsevier B.V. This is an open access article under the CC BY-NC-ND license Copyright will Published appear here. (http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer-review under responsibility of the Bayerisches Laserzentrum GmbH. Peer-review under responsibility of the Bayerisches Laserzentrum GmbH

Keywords: Additive Manufacturing, Selective Laser Melting, Multibeam Strategies, Simulation, Melt Pool Dynamics

1. Introduction Additive manufacturing offers the chance to design functional optimized parts with only little restrictions and thus is of rapidly increasing importance to various industry sectors. Selective laser melting (SLM) as one of these additive manufacturing technologies especially fulfills the needs of industry since it enables the manufacturing of metal parts of various materials. Stainless steels (e.g. Lima et al. 2014), nickel-based superalloys (e.g. Cloots et al. 2016), titanium alloys (e.g. Zhang et al. 2014), aluminum alloys (e.g. Buchbinder et al. 2011) and more alloys have

* Corresponding author. Tel.: +41-44-633-44-33 . E-mail address:[email protected]

1875-3892 © 2016 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer-review under responsibility of the Bayerisches Laserzentrum GmbH doi:10.1016/j.phpro.2016.08.094

900

Thorsten Heeling and Konrad Wegener / Physics Procedia 83 (2016) 899 – 908

already been processed to an equally high density as casted parts but of previously unreachable complexity and freedom of design. But for now there are several issues which limit the fields of application. These issues can basically be summarized to the criteria of surface quality, robustness and productivity. In terms of robustness SLM tends to result in porous parts, interlayer bonding errors due to insufficient melting and hot and cold cracking as well as distortion due to high temperature gradients and fast cooling rates (Val Belle et al. 2013). 1.1. Simulation of Selective Laser Melting To overcome these drawbacks several simulation models have been established to get a more detailed understanding of the process dynamics of the selective laser melting process and therefore to derive optimization strategies from these calculations. The focus of simulation based research in SLM is split into two main approaches (Schoinochoritis et al. 2015). One is focusing on residual stresses and therefore usually utilize finite element methods. Recent models tend to construct the layerwise irradiation from single and multi-line simulations and take them as one loading per layer. As a consequence of a sequence of punctual, local oder layerwise thermal loads the residual stresses are calculated for the whole part (Papadakis et al. 2014). But there are some models which simulate track after track to predict the part’s residual stresses as well (Pal et al. 2015) or just a single track as a first indicator (Hussein et al. 2013). The second approach takes a deeper look into melt pool dynamics in consequence of surface tension, Marangoni effect and evaporation but neglecting the thermomechanical stresses because these fluid dynamic related effects got a more microscopic scale and thus need more detailed meshes for an appropriate simulation. Simple models of this approach elaborate volume heat sources and temperature dependent material properties but neglect the melt pool dynamics (Roberts et al. 2009 & Verhaeghe et al. 2009). More detailed models like the one of Dai & Gu (2015) are considering most of the previously mentioned physical effects but consider the power bed as a continuum. Furthermore a very detailed model has recently been published by Khairallah et al. (2016) modeling every single powder particle and thus resulting in high calculation times but showing the importance of the fluid dynamic related aspects regarding the melt pool geometry and therefore the process robustness (also see King et al. 2015). In this study an approach of a homogeneous powder bed and a representation of various physical effects are used to significantly reduce the calculation times but to still enable a detailed view into process dynamics by considering multiple physical effects in the melt pool. 1.2. Multibeam Processing Today multibeam strategies for the selective laser melting process are usually considered to be parallel strategies, meaning multiple beams are used to independently melt on separate scan fields with a small overlap to improve productivity (Wiesner & Schwarze 2014) but the use of multiple beams to improve robustness, part quality and the range of usable materials is rarely investigated. Wilkes et al. (2011) used a second, static high power laser in addition to the melting laser to heat the building chamber and therefore to reduce temperature gradients as well as the residual stresses. Weingarten et al. (2015) used just one beam but scan the part two times to dry the aluminum powder bed in the process previous to melting to reduce the amount of hydrogen porosity and Abe et al. (2001) used a second moving beam in a simple test to selectively reduce the cooling rates and thus influence the microstructure whith promising results. Therefore this study focuses on the simulation of synchronized multibeam strategies, which could be established using two indepent lasers and beam deflection units, to evaluate the influence of these strategies regarding the temperature field and melt pool dynamics prior to implementation.

Thorsten Heeling and Konrad Wegener / Physics Procedia 83 (2016) 899 – 908

Nomenclature A cp dpp hpl n n* Q t T V į(ĭ) ț Ȝ ȡ ȡpl

area specific heat capacity diameter of powder particle powder layer heigth inward normal outward normal amount of heat time step index temperature volume surface indicator function surface curvature optical thickness density relative density of powder layer

2. Modeling The computational model elaborates a temperature field calculation that is interconnected with a melt pool dynamics calculation using a weak coupling. The model consists of cubic elements which can be categorized regarding their state, meaning powder, solid, liquid or gaseous. The effects of temperature dependent material properties, thermal conduction, convection and radiation, buoyancy and capillary forces, the Marangoni effect as well as evaporation and recoil pressure are considered to ensure a detailed representation of the selective laser melting process. For the evaluation of the strategies’ influences simulations of single tracks are performed in which a single track is molten right next to a previously solidified line and therefore imitating a multi track experiment. 2.1. Powder Bed The initial configuration of the powder bed is realized as homogeneous partly filled elements, whereas the filling degree can change due to melting and consolidation of the elements as shown in Fig. 1. Therefore the material properties are weighted by the filling degree. This weighting of material properties is used for all properties but the thermal conductivity because studies of Sih & Barlow (2004) show that the conductivity of the powder bed is far less due to small contact areas. The homogeneous approach allows a good level of detail for comparably low calculation times since the element sizes can be set to higher values than for the representation of every single powder particle. The temperature dependency of material properties is represented using piecewise interpolation for the material properties of 316L using the information of the IAEA (2008). The melting, solidifying as well as the evaporation are realized using a separated consideration of the heat of fusion and evaporation which apply when the solidus, liquidus and evaporation temperature respectively are reached. The heat of fusion is distributed over the temperature range between solidus and liquidus temperature while the consideration of evaporation is set to the pressure dependend evaporation temperature. The values used for the simulation of 316L are listed in Table 1.

901

902

Thorsten Heeling and Konrad Wegener / Physics Procedia 83 (2016) 899 – 908

Table 1. 316L material parameters. Parameter

25 °C

1400 °C

1450 °C

2800 °C

Specific Heat Capacity [J/kg K]

450

700

707

900

Thermal Conductivity [W/m K]

13.3

33.78

18.1

22.2

Surface Tension [N/m]

1.76

0.41

Dynamic Viscosity [Pa s]

0.0059

0.0014

Heat of Fusion [J/kg]

270000

Heat of Evaporation [J/kg]

7450000

Hemispherical Reflectance

0.64

2.2. Temperature Field The temperature field calculation uses a finite differences scheme considering the energy input due to laser absorption, the energy exchange due to conduction and melt advection as well as the energy loss due to convection, radiation and evaporation:

T t 1

Tt 

Qabs  Qcond  Qconv  Qrad  Qadv  Qevap

U ˜V ˜ cp

(1)

The absorption model elaborates the physical equation proposed by Gusarov et al. (2009) in which the optical thickness is calculated as follows.

O

3 U pl hpl ˜ ˜ 2 1  U pl d pp

(2)

Numerical adjustments at the optical thickness are made to allow a more detailed representation of the different areas in and around the melt pool by considering the current configuration of every element stack for the calculation of the values of powder layer heigth and relative powder layer density as illustrated in Fig.1.

903

Thorsten Heeling and Konrad Wegener / Physics Procedia 83 (2016) 899 – 908

Fig. 1. Illustration of the determination of layer heigths for the calculation of optical density. Material is illustrated gray.

2.3. Meltelt Pool Dynamics For the calculation of melt flow a combined level set volume of fluid (CLSVoF) method is elaborated in combination with a pressure implicit splitting of operators (PISO) solver in accordance with the publications of Son (2005), Son & Hur (2002) and Issa (1986) and adjusted to match the challenges of multiple not completely filled elements which result from the initial configuration until the melt consolidates. In the CLSVoF method a level set method is used to identify the melt pool boundaries while the volume of fluid method is elaborated to guarantee the conservation of mass while advecting material from element to element. The PISO solver calculates the melt flow by solving the momentum equation in a splitwise general transportation approach using a double correction scheme. The forces considered in the momentum equation cover the buoyancy Fbou and capillary forces Fcap as well as the forces due to the Marangoni effect Fmar and recoil pressure Frec and are listed in equations (3) to (6). The surface reconstruction is established in a piecewise linear manner for the calculation of normals.

F bou

g ˜ U ˜V

F cap

n * ˜V ˜ N ˜ A ˜ G )

F mar

>I  n … n ˜ ’@˜V ˜ A ˜G )

F rec

n ˜ p rec ˜ A ˜ G )

(3) (4) (5) (6)

3. Synchronized Multibeam Strategies 3.1. Layerwise synchronized Layerwise synchronized strategies are layerwise synchronizations in which one beam is used to melt the material as usual while the second beam is used to heat the whole part to establish a homogeneous temperature field and therefore aiming for a reduction of residual stresses by lowering temperature gradients. The strategy is illustrated in Fig. 2 in comparison to the spatially synchronized strategies. Furthermore the melting can be improved due to the

904

Thorsten Heeling and Konrad Wegener / Physics Procedia 83 (2016) 899 – 908

additional heat input and the cooling rates can be manipulated partwise and therefore the microstructure evolution can be influenced. To guarantee a quasi homogeneous temperature field the second beam should be characterized by a larger spot size and much higher scan speeds than the first beam so that the part is scanned multiple times while the first beam melts the defined cross section. This strategy can comparibly easy be implemented since the second beam does not need to hit the exact same spot as the melting laser does. For the computational evaluation the layerwise synchronized strategy will be represented as higher starting temperatures in the simulation model. 3.2. Spatially synchronized The spatially synchronized strategies are characterized by a defined offset of both beams to each other. Therefore these strategies are defined by a synchronization of the beam movements and require an accurate calibration all over the base plate to guarantee the defined offsets. Using two beams with a defined offset the intensity profile can be adjusted to fit various needs. While an offset in scanning direction can be used for drying and presintering or the adjustment of cooling rates and temperature gradients in the vicinity of the melt pool tail, an offset in lateral direction can be used to influence the melt pool cross sections to be wider and therefore to increase the connection to previous layers and possibly increasing the energy utilization for melting by overlapping the outermost regions of the Gaussian profiles so that these result in melting as well. The second laser can also be used to heat up stripes aside the main scanning line to induce closing stresses. For the simulation of these strategies two beams of same size are placed in defined offsets to each other and being moved along the scanning vector. The use of two beams of different sizes is possible as well but is not considered in this work.

Fig. 2. Illustration of layerwise and spatially synchronized scanning strategies.

4. Results and Discussion The simulations are conducted for the stainless steel 316L using the same parameter sets for all simulations. In case of the spatially synchronized strategies the offsets and power distributions between the two lasers are varied but the overall input power is kept constant. For layerwise synchronized strategies the starting temperature is increased, imitating the heating of a second, much faster beam. The parameters are listed in Table 2.

Thorsten Heeling and Konrad Wegener / Physics Procedia 83 (2016) 899 – 908

Table 2. Simulation parameter overview. Parameter

Value

Unit

Scan speed

1150

[mm/s]

Overall power

200

[W]

Focus diameter

90

[μm]

Layer thicknes

30

[μm]

4.1. Melt Pool Dimensions The knowledge of melt pool dimensions is of crucial importance since the overlap from one line to the next and from one layer to another has to be wide enough to prevent defects which increase porosity. Furthermore insufficient remelting increases the chance for balling during the cooling of the melt pool tail. Fig. 3 shows the simulated remelting depth and width for spatially synchronized strategies with a beam offset of 0.25 to 1 times the beam diameter.

Fig. 3. Simulation results regarding remelted depth and width of the melt pool for both spatially and layerwise synchronized strategies. Layerwise synchroniezed simulations are represented by the lines of higher starting temperatures.

It shows that the remelted depths are continuously decreasing for increasing beam offsets. For offsets in scanning direction it has to be related to a longer interaction time of the same energy amount, meaning the input power is distributed over a wider area and therefore the maximum intensity decreases and the time to conduct energy away from the melt pool increases as well. The remelting depth is strongly influenced by this effect since the maximum depth is related to the recoil pressure which increases for higher intensities. It has to be pointed out that the remelting depth at a beam offset of 0.25 times the beam diameter is larger than the one of the reference single beam. This needs a closer look into the melt pool dynamics because it cannot be explained by intensities or interaction times alone. The strong decrease of the remelting depth in case of the lateral beam offset is a result of the decreasing energy density. But because the energy is distributed over a wider area perpendicular to the scanning direction the remelting width increases with increasing lateral beam offsets. Furthermore the geometry changes significantly from a single bell like melt pool at low offsets to a flat and wide melt pool at offsets of about half the beam diameter to a double peak melt pool at higher offsets. The results of the simulation for different starting temperatures of the layerwise synchronized strategies are illustrated in Fig. 3 as single beam simulations at higher ambient temperatures. Since the starting temperature is higher, less energy is needed to melt the material and thus the remelting depth and width increase with growing starting temperatures, meaning a more powerful heating of the whole part by the second beam.

905

906

Thorsten Heeling and Konrad Wegener / Physics Procedia 83 (2016) 899 – 908

4.2. Melt Pool Dynamics The melt pool dynamics of layerwise synchronized strategies are not much different to those using a single beam if the short time of overlapping is neglected. Naturally the higher starting temperatures influence the melt pool dimensions due to the fact that less energy is needed to reach the melting point. But the influence of the melt pool dynamics is small since it ist strongly driven by the recoil pressure. Compared to that the use of a defined offset strongly influences the melt pool dynamics in comparison to a single beam. This is mostly due to a change in the area in which evaporation takes place. As Fig. 4 illustrates for a small offset in scanning direction the evaporation related vortex in the front of the melt pool is enlarged by the second beam’s evaporation of material. This second evaporation area prevents the first vortex to close right after the first beam but pushes the melt deeper down so that the heat is more effectively transported to the unmolten material. If the offset is large enough so that both evaporation areas do not overlap a wave like surface shape is generated in which a maximum of melt is surrounded by a minimum in the front of the melt pool and one to the back of it. The maximum is generated by the two colliding upward streams of both evaporation related vortices. The same effects occur for lateral offsets so that the melt pool cross sections are influenced as previously shown.

Fig. 4. Illustration of the melt flows in longitudinal section for a) the 200W single beam and two 100W/100W laser beams with offsets of b) 0.25 times and c) 1 times the focus diameter.

Thorsten Heeling and Konrad Wegener / Physics Procedia 83 (2016) 899 – 908

4.3. Temperature Gradients Temperature gradients are a good indicator for resulting stresses because they are directly influencing these by the temperature gradient mechanism (Munsch 2013). The plotted temperature gradients are taken as the component perpendicular to scanning direction of the solid uppermost elements of the previous layer. As Fig. 5 shows the temperature gradients can significantly be influenced by the multibeam strategies. Even in case of small beam offsets the maximum lateral temperature gradients in the upper part of the previous layers can be reduced by about 10% by using a 150W/50W or a 50W/150W split. For increasing beam offsets the usability of the parameter sets has

Fig. 5. Simulation results regarding the temperature gradients perpendicular to scanning direction in the upper part of the previous layer.

to be examined since the remelted depth decreases which might result in higher porosity. But for a beam offset of half a beam diameter the gradients can be reduced by 15% and for an offset of a single beam diameter by about 25%. While the 150W/50W and 50W/150W split are more effective for low beam offsets, the 100W/100W split leads to lower temperature gradients at higher beam offsets. This is because the influence of the 50W beam decreases for higher offsets and the 150W beam dictates the melt pool behavior and therefore the temperature field. For a partwise heating with temporal synchronization the temperature gradients are decreased at higher surrounding temperatures as expected. For a surrounding temperature of 400°C the reduction of the lateral temperature gradient is of about 10%. But the reduction for the investigated temperature range is not as high as for the simulated beam offsets. 5. Conclusion The simulation results show the influence of different scanning strategies onto the melt pool dimensions and dynamics as well as the temperature gradients in the previous layer using two independent but synchronizable beams for the selective laser melting process. Both strategies show great potential for the optimization of this process. While all strategies tend to reduce the temperature gradients of about 10% to 20% and therefore are assumed to reduce the residual stresses, there are some remarkable potentials for the use of small beam offsets in scanning direction to increase productivity. As shown, a defined positioning of the beams to each other influences the evaporation related vortices and can therefore support the downward streams in the melt pool front so that the remelted depth increases. Due to these stronger downward streams the beams can be moved faster because the necessary depth can be reached using less energy. The faster scan speed would directly result in an increase of the process productivity. The lateral offset shows no direct advantage over the the other strategies since with increasing lateral offsets the energy density decreases so that the remelted depth will no more be large enough to guarantee dense parts. Furtheron these results have to be experimentally validated to support the simulation results and to deduct further optimized parameter sets. An investigation of strategies using two different focus diameters is also necessary because these most probably increase the positive influence on the temperature gradients.

907

908

Thorsten Heeling and Konrad Wegener / Physics Procedia 83 (2016) 899 – 908

Acknowledgments The authors gratefully acknowledge the financial support given by the Bosch Research Foundation (former Hans L. Merkle Program for Excellency in Science and Technology). References Abe, F., Osakada, K., Shiomi, M., Uematsu, K., Matsumoto, M., 2001. The manufacturing of hard tools from metallic powders by selective laser melting. Journal of Materials Processing Technology 111, 210-213. Buchbinder, D., Schleifenbaum, H., Heidrich, S., Meiners, W., Bültman, J., 2011. High Power Selective Laser Melting (HP SLM) of Aluminum Parts. Physics Procedia 12, 271-278. Cloots, M., Kunze, K., Uggowitzer, P.J., Wegener, K., 2016. Microstructural characteristics of the nickel-based alloy IN738LC and the cobaltbased alloy Mar-M509 produced by selective laser melting. Materials Science & Engineering A 658, 68-75. Dai, D., Gu, D., 2015. Tailored reinforcement/matrix interface and thermodynamic mechanism during selective laser melting composites. Materials Science and Technology, DOI: 10.1179/1743284715Y.0000000012. Gusarov, A.V., Yadroitsev, I., Bertrand, P., Smurov, M., 2009. Model of Radiation and Heat Transfer in Laser-Powder Interaction Zone at Selective Laser Melting. Journal of Heat Transfer 131, 072101. Hussein, A., Hao, L., Yan, C., Everson, R., 2013. Finite element simulation oft he temperature and stress fields in single layers built withoutsupport in selective laser melting. Materials and Design 52, 638-647. International Atomic Energy Agency, 2008. Thermophysical Properties of Materials For Nuclear Engineering: A Tutorial and Collection of Data. Issa, R.I., 1986. Solution of the implicitly discretized fluid flow equations by operator-splitting. Journal of Computational Physics 62, 40-65. Khairallah, S.A., Anderson, A.T., Rubenchik, A., King, W.E., 2016. Laser powder-bed fusion additive manufacturing: Physics of complex melt flow and formation mechanisms of pores, spatter, and denudation zones. Acta Materialia 108, 36-45. King, W.E., Anderson, A.T., Ferencz, R.M., Hodge, N.E., Kamath, C., Khairallah, S.A., Rubenchik, A.M.,2015. Laser powder bed fusion additive manufacturing of metals; physics, computational, and materials challenges. Applied Physics Review 2, 041304. Lima, M.S., Sankarè, S., 2014. Microstructure and mechanical behavior of laser additive manufactured AISI 316 stainless steel stringers. Materials and Design 55, 526-532. Munsch, M., 2013. Reduzierung von Eigenspannungen und Verzug in der laseradditiven Fertigung. Dissertation Technische Universität Hamburg-Harburg. Pal, D., Patil, N., Zeng, K., Teng, C., Stucker, B., 2015. An Efficient Multi-Scale Simulation Architecture for the Prediction of Performance Metrics of Parts Fabricated Using Additive Manufacturing. Metallurgical and Materials Transactions A 46(9), 3852-3863. Papadakis, L., Loizou, A., Risse, J., Bremen, S., Schrage, J,. 2014. A computational reduction model for appraising structural effects in selective laser melting manufacturing. Virtual and Physical Prototyping 9(1), 17-25. Roberts, I.A., Wang, C.J., Esterlein, R., Stanford, M., Mynors, D.J., 2009. A three-dimensional finite element analysis of the temperature field during laser melting of metal powders in additive layer manufacturing. International Journal of Machine Tools & Manufacture 49, 916-923. Schoinochoritis, B., Chantzis, D., Salonitis, K., 2015. Simulation of metallic powder bed additive manufacturing processes with the finite element method: A critical review. Proceedings of the Institution of Mechanical Engineers, Part B: Journal of Engineering Manufacture, 0954405414567522. Sih, S.S., Barlow, J.W., 2004. The Prediction of the Emissivity and Thermal Conductivity of Powder Beds. Particulate Science and Technology 22, 427-440. Son, G., Hur, N., 2002. A coupled level set and volume-of-fluid method for the buoyancy-driven motion of fluid particles. Numerical Heat Transfer, Part B 42, 523-542. Son, G., 2005. A level set method for incompressible two-fluid flows with immersed solid boundaries. Numerical Heat Transfer, Part B 47, 473489. Verhaege, F., Craeghs, T., Heulens, J., Pandelaers, 2009. A pragmatic model for selective laser melting with evaporation. Acta Materialia 57, 6006-6012. Van Belle, L., Vansteenkiste, G., Boyer, J.-C., 2013. Investigation of residual stresses induced during the selective laser melting process. Key Engineering Materials 554-557, 1828-1834. Weingarten. C., Buchbinder, D., Pirch, N., Meiners, W., Wissenbach, K., Poprawe, R., 2015. Formation and reduction of hydrogen porosity during Selective Laser Melting of AlSi10Mg. Jorunal of Materials Processing Technology 221, 112-120. Wiesner, A., Schwarze, D., 2014. Multi-Laser Selective Laser Melting. 8th International Conference on Photonic Technologies LANE 2014. Industrial Paper. Wilkes, J., Hagedorn, Y.-C., Meiners. W., Wissenbach, K., 2011. Additive manufacturing of ZrO2-Al2O3 ceramic components by selective laser melting. Rapid Prototyping Journal 19(1), 51-57. Zhang, S., Wei, Q., Cheng, L., Li, S., Shi, Y., 2014. Effects of scan line spacing on pore characteristics and mechanical properties of porous Ti6Al4V implants fabricated by selective laser melting. Materials and Design 63, 185-193.