Scaling Effects in Laser-Based Additive Manufacturing Processes

Scaling Effects in Laser-Based Additive Manufacturing Processes Andrew J. Birnbaum and Jack L. Beut Department of Mecanical Engineering Carnegie Mellon University Pittsburg, Pa 15213 James W. Sears Advanced Materials Processing Center Sout Dakota Scool of Mines and Tecnology Rapid City, SD 5771 Abstract Termal modeling is used to address te control of melt pool size in laser-based additive powder fusion processes under steady state conditions. Tese processes use localized melting of metal powder to add features to metallic components during manufacture or repair. Te problem of process size scale is considered, wit te aim of applying knowledge developed at one size scale (e.g. te LENS TM process, using a 5 W laser) to similar processes operating at larger scales (e.g. a 3 kw system under development at Sout Dakota Scool of Mines and Tecnology). Results presented erein provide engineers wit a means for easily predicting melt pool size for bot of tese processes over te full range of process variables. Results also demonstrate ow process size scale affects te sensitivity of melt pool size to minor canges in process parameters. Tese issues are addressed via a process map approac developed by te autors and co-workers. Tis approac collapses results from a large number of simulations over te full range of practical process variables onto plots process engineers can easily use. Introduction Recently, te use of laser-based powder fusion processes for component repair and additive manufacturing applications as received significant attention in te aerospace industry. In te case of additive manufacturing, significant cost savings are possible in te fabrication of some components if laser-based deposition is used to add small features to larger parts manufactured by traditional processes. For suc applications, process scaling as emerged as a critical issue. Many industrial additive manufacturing applications demand te use of large-scale deposition processes; yet significant process development as taken place on small-scale processes. Tis includes an extensive researc effort over te past eigt years at Sandia National Laboratories to develop te LENS TM process (Griffit et al., [1]). Most LENS TM process development researc as been performed using a 5 W Nd:YAG laser. In contrast, AeroMet, wic manufactures components for te aerospace industry, uses an 18 kw CO 2 laser. Tere is currently no fundamental understanding of ow to apply knowledge of smallscale systems to analogous large-scale systems. Te result is tat wenever a new laser-based manufacturing system is developed at a different size scale, processing engineers must nearly start from scratc, performing large numbers of experiments to caracterize teir specific process. Te researc described in tis paper attempts to address tis issue as it relates to steadystate melt pool size control. Work described erein builds directly on modeling work by Vasinonta et al. [2, 3, 4] wic developed easy-to-use process maps, allowing te prediction of steady-state melt pool size in tin walled structures and bulky parts for any practical combination 151

of LENS TM process variables. A brief overview of te process map approac to understanding laser-based freeform fabrication processes is given by Beut and Klingbeil [5] and a complete presentation of te process map approac for controlling steady-state melt pool size and residual stress is given by Vasinonta [6]. Most recently, process maps of cooling rates and termal gradients at te melt pool boundary ave been developed wit te goal of predicting microstructure (Bonta and Klingbeil, [7]). Wit te exception of a preliminary consideration of process scaling provided in reference [7], results from tis earlier work are tailored for application to te LENS TM or oter similarly sized processes. Te work described erein establises an approac for predicting melt pool size for processes operating at different scales. It builds on work first reported by Birnbaum et al. [8] wic presented a preliminary analysis of process scaling limited to te consideration of tinwalled structures. In tis paper, a more complete consideration of te deposition of tin-walled structures is given, followed by an analogous consideration of te deposition of bulky features. All of te results presented erein relate to te prediction of steady-state melt pool size. No information is given on te rate of cange of melt pool size (from one steady-state value to anoter) if process variables are altered. Modeling of transient canges in melt pool size is addressed by Birnbaum et al. [8] and Aggarangsi et al. [9]. Altoug te approac taken is applicable to te laser fusion of any material, researc described in tis paper will specifically address te deposition of stainless steels. Numerical Models and te Process Map Approac Numerical Models: Te issues addressed in tis researc are considered wit reference to te part geometries sown in Figure 1. Te first geometry represents a tin-walled structure deposited onto a comparatively large base plate tat acts as a eat sink. Te second geometry represents a bulky structure also deposited onto a large base plate. For bot geometries, termal models are of a concentrated eat source moving across te top of te structure and do not model te effects of material addition. Te absorbed laser power is designated as αq, were α is te fraction of laser power from te source tat is absorbed by te structure. Te models used in tis paper are analogous to tose developed by Vasinonta [6]. For te tin-walled geometry, in comparing wit experiments and in determining ranges of absorbed laser powers, a value of α =.35 is used. For te bulky part geometry, a value of α =.7 is used. Predictions from numerical models assuming tese values of α ave sown good agreement wit melt pool sizes measured via termal imaging using te LENS TM process (Vasinonta et al., [2, 6]). In bot types of models, te successive deposition of layers is not modeled, but te preeating effects of te deposition of prior layers can be approximated via te specification of an elevated uniform temperature in te part and base plate, designated as T base, wic exists before te laser begins its travel across te top of te part. In all cases considered in tis paper, te part is tall enoug suc tat any increases in eigt will not cange te results. Te issue of sufficient part eigts to acieve tis condition is also addressed by Vasinonta et al. [2, 3] and Vasinonta [6]. Similarly, in tis study, melt pool size results are taken wen te eat source is sufficiently far from te vertical free edges suc tat results are independent of te distance from te edges. Te modeling of canges in melt pool size as a free edge is approaced is addressed by Aggarangsi, Beut and Griffit [1]. In 152

te process scaling simulations, processes using large values of laser power were performed wit models aving dimensions scaled up to ensure tat te conditions described above were satisfied. αq V Tin Wall αq V x l x z z L Bulky Part W L t Base plate Base plate Figure 1 Deposited Geometries wit a Base Plate Figure 2 sows typical finite element meses used for te scaling effects analyses. For te 2-D analysis of a tin-walled structure, boundary conditions are termal insulation on te vertical free edges and te top edge, wile a constant temperature is enforced at te bottom edge, simulating te effects of te base plate. Te models use four-node quadrilateral bi-linear elements provided by te ABAQUS finite element package. Models contain approximately fifteen tousand elements. However, traveling in te direction of te eat source (left to rigt), tere are two separate mes densities. Te first tird of te model as fairly coarse resolution, wile te remaining portion is of significantly finer resolution. Tis approac was taken to reduce analysis time wit te caveat tat fine resolution is only required were melt pool lengts are extracted from te model. Element lengts in te region were melt pool lengts are extracted are small enoug tat tere are always at least ten elements witin te melt pool. Mes resolution also increases as te top edge of te model is approaced. Constant power is applied to individual nodes for a time interval equal to te element edge lengt divided by V. Te mes used for te bulky part simulations is analogous to tat used to model tin walls, except an axisymmetric condition is applied about te axis parallel to te direction of laser travel (designated as te z axis in Fig. 2). Use of an axisymmetric model allows a drastic reduction in computation time compared to an analogous 3-D model. Te axisymmetric model actually simulates te movement of a eat source troug te center of a large solid (modeling double te volume of te actual geometry). Tus, te applied power used in te simulations was twice tat suggested by a value of α =.7. Termal properties of AISI 34 stainless steel are used as inputs to te models (Dobranic and Dykuizen, [11]). A solidus temperature of 1672 K, a liquidus temperature of 1727 K, a latent eat of fusion of 2.65 x 1 5 J/kg, and a constant density of 7652 kg/m 3 are specified. Below a temperature of 15 K temperature dependent termal conductivity, k and specific eat, c are given by te following linear equations: 153

k = 8.116 +.1618(T) (W/mK ) (1) c = 465.4 +.1336(T) (J/kgK). Above 15 K, bot termal conductivity and specific eat are eld constant at te 15 K value. t q = Absorbed Power, αq x V x z z q = L Constant Temperature T = T base q = z Axisymmetric about z Axis Constant Temperature T = T base Absorbed Power, αq V q = r Figure 2 Finite Element Meses wit Boundary Conditions for Tin Wall and Bulky Part Simulations Process Map Approac: Te process scaling researc described erein builds upon previously developed process map concepts. A process map for melt pool lengt for a tin-walled structure traversed by a concentrated laser eat source as been developed by Vasinonta et al. [2, 3]. As suggested by te Rosental [12] solution for a point eat source moving across a (2-D) alfspace, a process map for melt pool lengt is represented troug tree dimensionless variables: te normalized melt pool lengt (l ), te normalized substrate eigt ( ) and te normalized melting temperature ( T m ) wic are defined as follows: l l =, ρcv = and ρcv T m T T = m base. (2) αq πkt In eq. (2), ρ, c and k are te density, specific eat and termal conductivity, respectively. If termal properties are temperature-independent and latent eat effects are not modeled, results from te analysis of a concentrated eat source moving over a tin-walled structure of finite eigt,, can be represented as a single surface plotted on tree coordinate axes of l, and 154

T m. Tis forms te basis of a process map approac for analyzing te laser deposition of tinwalled structures. A process map for deposition of tin-walled structures of 34 stainless steel via te LENS TM process can be constructed using results from temperature-dependent termal simulations including latent eat effects if te following procedures are followed: 1. Properties at 1 K are used in te normalizations. 2. For cases involving a cange in preeat, a linear cange in termal conductivity wit preeat temperature (in deg. C) is assumed, given by k = 24.3 +.13(T base -3) W/(mK). 3. For predicting steady-state melt pool lengts resulting from a cange in process variables, wall tickness is assumed to scale proportionally wit melt pool lengt. Te melt pool lengt/wall tickness scaling is assumed to be unaffected by velocity. Te tird assumption is necessary because te wall tickness, t, is included in te normalized variable T m used in te process map. Tis requires tat some assumption be made regarding te relationsip between melt pool lengt and wall tickness. It also means, owever, tat witin te limits of assumption #3 and given a value of t from a single experiment, te process map can be used to predict not only melt pool lengt as a function of process variables, but also wall tickness. For process variables of laser velocity, laser power wall eigt and preeat temperatures of interest for te LENS process, termal simulation results normalized using te rules above rougly fall on a single surface plotted on tree coordinate axes of l, and T m. Te variability of results due to temperature-dependent properties is +/-6.5% or less. An analogous approac as been taken to construct process maps for te deposition of bulky parts [6]. For suc structures, te dimensionless variables are: d d =, ρcv T T = and T = m base, (3) m αq ρcv ρcv πk 4k were melt pool depts are considered instead of melt pool lengts and te nondimensionalization for T m no longer includes a tickness, t. It sould be noted tat te T m definition cited above differs by a factor of two from te bulky part T m definition used in te study of microstructure in reference [7]. Normalization procedures needed to collapse termal simulation results for te LENS TM process onto a single surface in 3-D nondimensional variable space are: 1. Properties of SS34 at 11 K are used in te normalizations. 2. For cases involving a cange in preeat, a linear cange in termal conductivity wit preeat temperature (in deg. C) is assumed, given by k = 25.9 +.13(T base -3) W/(mK). If tese rules are followed, te variability in melt pool dept results is also witin +/-6.5%. Targeted Manufacturing Process: A laser processing facility is currently under development witin te Advanced Materials Processing Center at te Sout Dakota Scool of Mines and Tecnology for use in not only net sape manufacturing but also welding, micro-macining, surface treatment and oter 155

applications. Te system consists of a 3 kw Nd:YAG laser wit a robotic positioning system, dual powder feeders and geometric, temperature and position sensing capabilities. It as been tested for net sape manufacturing applications troug te building of a series of tin-walled structures deposited using laser powers from 45 W to 9 W. Laser velocities of interest range from 1 to 2 mm/s. Because te power range of te laser at te AMP Center is significantly larger tan tat for te 5 W LENS TM system, te development of tis new facility offers a unique opportunity for testing te applicability of a process map approac on multiple process size scales. Process Scaling for Tin-Walled Structures Process Maps for Multiple Process Scales: In order to analyze te effects of process scaling, laser powers of 123 W to 27 W (αq from 43 W to 945 W assuming a value of α =.35) were divided into two power ranges. Te ig power range is from 43 W to 27 W (αq from 15 W to 945 W). Te low power range is based on te LENS process and as powers ranging from 123 W to 471 W (αq from 43 W to 165 W). As in earlier process map work, a single experimental value of te wall tickness, t, is needed to predict values of l and t as a function of process variables. A prediction of melt pool lengt, l, for te experiment aving a known value of t yields a value of l/t tat is used in all subsequent predictions. Also, as in earlier work, prediction of melt pool lengt and wall tickness using te process map must be done iteratively. For example, a larger value of αq results in a new (smaller) value of T m. Tat smaller value of T m results in a new (larger) value of l, and, given a value of l/t, a larger value of t. Te larger value of t yields a sligtly larger value of T m. Tis again leads to new values of l and t. Tese calculation steps are repeated until l and t stop canging significantly. Results ave been extracted from tin wall numerical models for a T m range of.36 to 2.7. Iterative calculation of melt pool lengts and ticknesses using V = 5.93 to 9.31 mm/s for LENS TM and V = 1 to 2 mm/s for te AMP process yields ranges of T m for te two power ranges tat are witin tis range. Figures 3 and 4 provide plots of l vs. T m over te full range of.36 T m 2.7 applicable to te LENS TM and AMP Center processes. Results are for tall walls ( large). Data plotted in te low range of powers (large values of T m ) reproduces existing process map data for LENS TM for V = 7.62 mm/s. Data for smaller values of T m is new and relates to power ranges and velocities appropriate for te AMP Center process. Figure 3 gives results for a value of T base = 33 K and Figure 4 gives results for an upper bound value of T base = 673 K. In bot cases, results are given for te upper and lower bounds of V = 1 mm/s and V = 2 mm/s for te ig power ranges (applicable to te AMP Center process). Results for V = 15 mm/s (not sown) fall between te results for te upper and lower values of velocity. Te variability in results plotted in Figs. 3 and 4 is confined to +/-3.5% in te AMP power range if te following procedures for applying te process map are followed: 1. Properties at a normalization temperature T norm = 889 K are used in te normalizations. 2. For cases involving preeating, a linear cange in termal conductivity wit a preeat temperature (in deg. C) is assumed, given by k = 22.5 +.58(T base -3) W/(mK). 156

3. For predicting steady-state melt pool lengts resulting from a cange in process variables, wall tickness is assumed to scale proportionally wit melt pool lengt. In oter words, te process map approac can be applied over multiple process size scales by simply canging te normalization temperature in step 1 and te terms in te equation in step 2 wit canges in power range. 14 12 1 Rosental LENS, V=7.62 mm/s AMP, V=1 mm/s AMP, V=2 mm/s 14 12 1 Rosental LENS, V=7.62mm/s AMP, V=1 mm/s AMP, V=2 mm/s l 8 l = / ρcv 6 l 8 l = / ρcv 6 4 4 2 2.5 1 1.5 2 2.5 3 3.5 Tm Tbase Tm = α Q/πkt Figure 3 Comparison of Normalized Tin Wall Numerical Predictions wit te Rosental Solution over te Full Range of T m (T base = 33 K)..5 1 1.5 2 2.5 3 3.5 Tm Tbase Tm = α Q / πkt Figure 4 Comparison of Normalized Tin Wall Numerical Predictions wit te Rosental Solution over te Full Range of T m (T base = 673 K). Application of te Results: Figure 5 sows a plot of predicted and measured wall ticknesses vs. αq. Measured values are from te AMP Center process for V = 2 mm/s and are sown as large data points. Predictions as a function of αq and for V = 2 mm/s are sown as plotted lines wit small data points. Two sets of predictions are presented. Te top line represents process map predictions using an experimental value of t from te LENS process for αq = 15 W and V = 7.62 mm/s to obtain a value of l/t = 1.5. Altoug te trends in te experiments are captured by tese predictions, te predicted values are larger tan te experimental values. Tis difference can be explained by te use of a value of l/t determined from experiments at a significantly lower velocity tan was used in te AMP experiments. Tis ratio will, in fact, increase wit an increase in velocity. Te second set of predictions (te lower line) was generated from te same process map results (Figure 3), but wit a value of t from te AMP process for αq = 21 W and V = 14 mm/s used to obtain a value of l/t = 1.67. Tese predictions agree quite well wit te available experimental data. Furtermore, te predictions for larger powers could be a useful tool in reducing te number of experiments needed to caracterize te AMP process. Process Robustness: Process robustness as been assessed by determining te sensitivity of melt pool size to canges in power and velocity for te two power ranges considered in tis researc. Figure 6 summarizes te tin wall process robustness results. Te Rosental solution and appropriate normalization rules for eac power range and a base temperature of T base = 33 K were used to generate te plotted data. 157

Tickness, t (mm) To quantify te sensitivity of melt pool size to canges in power, low, middle and ig power values witin te LENS TM and AMP process power ranges were used as initial power levels. Te initial LENS TM and AMP powers were αq = 43, 14 and 165 W and αq = 15, 548 and 945 W, respectively. Te percent cange in melt pool lengt due to power increases of up to 2% were ten determined. To measure te effect of laser velocity, power cange calculations were performed at tree different speeds for eac power range: 5.93, 7.62 and 9.31 mm/s for LENS TM, and 1 15 and 2 mm/s for te AMP process. 4.5 4 3.5 3 2.5 2 1.5 1 l/t, LENS.5 l/t, AMP Experimental 2 4 6 8 1 αq (Watts) Figure 5 Comparison of Experimental and Predicted Ticknesses as a Function of αq (α =.35) % Cange in Melt Pool Lengt 5 1 15 2 Te plot in Figure 6 presents tis information as two curves defining te upper and lower bounds of results due to power canges for bot processes over all operating velocities. Te LENS TM process operating at αq = 43 W (minimum power, maximumt m ) and te lowest velocity (V = 5.93 mm/s) yields te upper bound curve, wile te AMP process operating at αq = 945 W (maximum power, minimumt m ) and te igest velocity (V = 2 mm/s) yields te lower bound curve. Tere is a maximum 6.6% difference between te AMP and LENS TM results indicating tat process robustness wit respect to power canges is nearly independent of process size scale. Interestingly, tis convergence of results for all power ranges does not occur unless increases in wall tickness due to increases in melt pool size are accounted for. Anoter interesting feature of te results plotted in Fig. 6 is te nearly linear relationsip between melt pool lengt and percent cange in power. Tis is despite l vs. T m plots tat are igly nonlinear (see Figs. 3 and 4). Tis is due to te relatively small canges in T m tat are involved in generating te power cange data in Fig. 6, wic render l vs. T m beavior approximately linear. Sensitivities to velocity canges were obtained in a manner analogous to te determination of power sensitivity. Melt pool lengts as a function of canges in velocity were calculated for tree initial velocities and at tree operating powers witin eac power range. Te velocity cange results sown in Figure 6 are te upper and lower bounds of te curves for bot 15 1 5-5 -1 LENS, Low Power and Velocity AMP, Hig Power and Velocity LENS, Low Power and Velocity AMP, Hig Power and Velocity % Cange in Power or Velocity Figure 6 Process Robustness for Tin Walls for Canges in Power or Velocity 158

te LENS TM and AMP ranges. Te LENS TM results operating at low power and low velocity yield te lower bound in sensitivity, wile te AMP process operating at te igest power and ig velocity yields te upper bound. Tere is a maximum 14% difference between AMP and LENS TM results, indicating tat process robustness wit respect to canges in velocity is somewat dependent on process size. Te data plotted in Fig. 6 also indicates tat te magnitudes of te slopes of te power cange curves are approximately tree times tose of te velocity cange curves. Tus, for te consistent construction of tin-walled structures, it is more important to minimize fluctuations in power tan fluctuations in velocity to maintain a desired melt pool geometry. Conversely, if a cange in melt pool size is needed, canging laser power is more effective tan canging laser velocity. Process Scaling for Bulky Structures Process Maps for Multiple Process Scales: Steps analogous to tose taken to analyze tin-walled structures are followed ere, except tat for bulky structures, a value of α =.7 is used to determine T m ranges. Te use of a value of α =.7 is based on agreement between numerical model predictions and measurements of melt pool widt for te LENS TM process. It is believed tat in te building of bulky parts, power absorption by te powder streams acts to increase te ability of te substrate to absorb laser power by rougly a factor of 2. Experiments involving laser glazing of bulky structures (passing of a laser over te surface witout powder deposition) ave not sown tis effect. Tis assumed value of α only affects te ranges of absorbed laser powers analyzed. Results are presented in terms of αq, and any oter value of α can be used wen applying results to predict melt pool sizes. In order to analyze te effects of process scaling, laser powers of 123 W to 27 W (αq from 86 W to 189 W assuming a value of α =.7) were divided into two power ranges. Te upper range, based on te AMP process, is from 43 W to 27 W (αq from 3 W to 189 W). Te lower range is based on te LENS process and as powers ranging from 123 W to 471 W (αq from 86 W to 33 W). Figures 7 and 8 provide plots of d vs. T m ( large) over te full range of.59 T m 5.82 applicable to te LENS TM and AMP Center processes. Figure 7 gives results for a value of T base = 33 K and Figure 8 gives results for T base = 673 K. In bot cases, results for te AMP process (wic is new data) are given for te upper and lower bounds of V = 1 mm/s and V = 2 mm/s. Properties at 1 K are used in te normalizations. For te cases of T base > 3 C, te dependence of conductivity on preeat temperature T base is taken as: k = 24.3 +.13(T base -3) (W/(mK). (4) As wit te tin-walled geometry, normalization by properties at a lower temperature as allowed results from larger-scale processes to be collapsed onto a single curve. Combining te data of Figs. 7 and 8 reveals tat variability of results in te AMP power range due to latent eat and termal property temperature dependence can be confined to +/-3%. 159

2.5 2 Rosental LENS, V=7.62 mm/s AMP, V=1 mm/s AMP, V=2 mm/s 2.5 2 Rosental LENS, V=7.62 mm/s AMP, V=1 mm/s AMP, V=2 mm/s 1.5 d d = / ρcv 1 1.5 d d = / ρcv 1.5.5 1 2 3 4 5 6 T T T m base m = αq ρ c V π k 4 k Figure 7 Comparison of Normalized Bulky Part Numerical Predictions to te Rosental Solution over te Full Range of T m, using a Pre-Heat Temperature T base = 33K. 1 2 3 4 5 6 Tm Tbase Tm = αq ρ c V π k 4 k Figure 8 Comparison of Normalized Bulky Part Numerical Predictions to te Rosental Solution over te Full Range of T m using a Pre-Heat Temperature T base = 673 K. Process Robustness: Figures 9 and 1 present plots of melt pool dept canges due to power canges (Fig. 9) and velocity canges (Fig. 1) analogous to te single plot for tin walled structures provided in Fig. 6. Analogous to te tin-wall case, te 3-D Rosental solution wit a base temperature of T base = 33 K as been used as a basis for all calculations. Low, mid-level and ig power values are all twice tose used to generate Fig. 6, consistent wit a value of α =.7. Low, mid-level and ig velocities for eac process are uncanged. Te plot in Figure 9 sows sensitivity to canges in power as a set of six curves, eac corresponding to one value of initial power. Te curve is drawn troug te data for te midlevel velocity for eac case (V = 7.62 mm/s for te LENS TM power range and V = 15 mm/s for te AMP power range). Values for te oter two velocities are plotted as data points only. Unlike te results for tin walls, tere is a substantial dependence of results on initial power level and velocity. However, te plot in Figure 9 indicates tat te smaller-scale LENS TM process is generally more sensitive to power fluctuations. Overall, sensitivity decreases wit an increase in initial power (decreasing initial T m ). Increases in operating velocity for a specific initial power value result in decreased sensitivity to canges in power. As in te results for tin walls, curves are nearly linear, due to te relatively small canges in T m involved. Sensitivities to velocities are caracterized in Fig. 1. Melt pool depts as a function of canges in velocity were recorded at te tree different velocities and at tree operating powers witin eac range. Te display of data is analogous to tat for Fig. 9. Te plot of Fig. 1 sows tat altoug tere is significant variability of results wit power and velocity, te AMP process is more sensitive to fluctuations in velocity tan te LENS TM process. Overall, increases in operating power for a specific velocity level result in increased sensitivity to canges in velocity. 16

In contrast to te tin wall case, te results in Figures 9 and 1 suggest tat, for te manufacture of bulky components, te sensitivity of laser-based additive manufacturing processes to canges in eiter laser power or laser velocity are significantly dependent on process size scale. Te larger-scale AMP process is less sensitive to (more robust wit respect to) fluctuations in power. Te smaller-scale LENS TM process is more robust wit respect to fluctuations in velocity. Results also suggest tat wen working exclusively in te smaller-scale LENS TM range, it is more important to control fluctuations in power rater tan velocity. Wen operating in te larger AMP range, control of fluctuations in velocity and power are rougly equally important. % Increase in Melt Pool Dept 18 16 14 12 1 8 6 4 2 LENS, Low Power LENS, Mid Power LENS, Hig Power AMP, Low Power AMP, Mid Power AMP, Hig Power 5 1 15 2 25 % Increase in Power Decreasing LENS AMP T m % Increase in Melt Pool Dept -2-4 -6-8 -1-12 -14 Decreasing T m LENS, Low Velocity LENS, Mid Velocity LENS, Hig Velocity AMP, Low Velocity AMP, Mid Velocity AMP, Hig Velocity 5 1 15 2 25 % Increase in Velocity LENS AMP Figure 9 Process Robustness for Bulky Parts for Canges in Laser Power Figure 1 Process Robustness for Bulky Parts for Canges in Laser Velocity Summary and Conclusions In tis study, process map approaces previously developed for application to te LENS TM process ave been extended to predict steady-state melt pool sizes for larger-scale processes. Tis allows te easy prediction of melt pool size for any combination of laser powers, laser velocities and part preeat temperatures. Process scaling predictions of wall ticknesses ave been made for te full power range of a large-scale process currently under development and predictions ave compared well to ticknesses measured to date. A study of te sensitivity of melt pool size to canges in process variables suggests tat, due to canges in wall tickness, process robustness is not a strong function of process size for te deposition of tin walls. In contrast tere is a large variation in process robustness wit process size for te deposition of bulky features. Acknowledgements Modeling researc at Carnegie Mellon is supported by te National Science Foundation Division of Design, Manufacture and Industrial Innovation, troug te Materials Processing and Manufacturing Program, award number DMI-227. Manufacturing researc at te AMP 161

Center is supported by te U.S. Army Researc Laboratory and te U.S. Army Office under grant number DAAD 19-2-2-1. Te autors would like to tank Dave Alexander and Ralp Anderson of Pratt & Witney for teir insigts and effort in guiding te industrial applications of tis researc. References 1. Griffit, M.L., Keicer, D.M., Atwood, C.L., Romero, J.A., Smugeresky, J.E., Harwell, L.D. and Greene, D.L., 1996, Freeform Fabrication of Metallic Components Using Laser Engineered Net Saping (LENS TM ) Solid Freeform Fabrication Proceedings, Austin, August 1996, pp. 125-132. 2. Vasinonta, A., Beut, J. L. and Griffit, M. L., 1999, Process Maps for Laser Deposition of Tin-Walled Structures, Solid Freeform Fabrication Proceedings, Austin, August 1999, pp. 383-391. 3. Vasinonta, A., Beut, J. L. and Griffit, M. L., 21, A Process Map for Consistent Build Conditions in te Solid Freeform Fabrication of Tin-Walled Structures, Journal of Manufacturing Science and Engineering. Vol. 123, pp. 615-622. 4. Vasinonta, A., Beut, J.L., and Ong, R., 21, Melt Pool Size Control in Tin-Walled and Bulky Parts via Process Maps, Solid Freeform Fabrication Proceedings, Austin, August 21, pp. 432-44. 5. Beut, J.L. and Klingbeil, N.W., 21, "Te Role of Process Variables in Laser-Based Direct Metal Solid Freeform Fabrication," JOM, September 21, pp. 36-39. 6. Vasinonta, A., 22, Process Maps for Melt Pool Size and Residual Stress in Laser-based Solid Freeform Fabrication, P.D. Tesis, Carnegie Mellon University, May 22. 7. Bonta, S. and Klingbeil, N.W., 23, Termal Process Maps for Controlling Microstructure in Laser-Based Solid Freeform Fabrication, Solid Freeform Fabrication Proceedings, Austin, August 23, pp. 219-226. 8. Birnbaum, Aggarangsi and Beut, 23, Process Scaling and Transient Melt Pool Size Control in Laser-Based Additive Manufacturing Processes, Solid Freeform Fabrication Proceedings, Austin, August 23, pp. 328-339. 9. Aggarangsi, P., Beut, J.L., and Gill, D.D., 24, Transient Canges in Melt Pool Size in Laser Additive Manufacturing Processes, Solid Freeform Fabrication Proceedings, Austin, August 24 (in te current proceedings). 1. Aggarangsi, P., Beut, J.L., and Griffit, M.L., 23, Melt Pool Size and Stress Control for Laser-Based Deposition Near a Free Edge, Solid Freeform Fabrication Proceedings, Austin, August 23, pp. 196-27. 11. Dobranic, D. and Dykuizen, R.C., 1998, Scoping Termal Calculation of te LENS TM Process, Sandia National Laboratories Internal Report. 12. Rosental, D., 1946, Te Teory of Moving Sources of Heat and Its Application to Metal Treatments, Transactions of ASME, Vol. 68, 1946, pp. 849-866. 162