Effect of Substrate Thickness and Preheating on the Distortion of Laser Deposited Ti–6Al–4V

The additive manufacturing (AM) process known as directed-energy-deposition (DED) [1] fabricates (or repairs) metal parts through layer-by-layer deposition onto a pre-existing part or substrate. In a popular DED configuration, the powder is directed through gas-fed nozzles at a melt pool created by a high-intensity laser beam on the surface of a substrate. DED is generally able to deposit material at a lower heat input compared to arc-based processes, and at a faster rate than powder bed fusion processes. These advantages make DED an effective process for repair and coating of surfaces.

Primary objectives of an AM repair are to produce a fully dense, metallurgical bond to the substrate that is free of microstructural defects and that results in minimal distortion of the component. The strategy to select processing parameters that is generally employed to achieve these objectives is based on single-track [2,3] and multitrack [4–6] builds. Practically, however, parameters are typically selected first to achieve a fully dense build, then adjustments are made to reduce distortion, and these adjustments can impact the microstructural quality. As a result, the process of developing parameters that meet quality objectives while minimizing distortion is, at its core, an optimization problem.

Typical methods to reduce part distortion are to modify substrate fixturing (i.e., mechanical constraints) [7,8] and to control thermal gradients throughout the part by either heating near the laser-interaction zone [9,10] or heating the bulk substrate [11,12].

There are three out-of-plane distortion modes observed in AM and welding processes: angular distortion, buckling, and longitudinal bending. Angular distortion and buckling, both products of transverse shrinkage, are less common in AM. Most common among AM processes is longitudinal bending, which is mainly attributed to the contraction of the molten material after deposition [13]. Heating along the top of the substrate causes thermal expansion. Cooler material below the melt pool restricts the expansion of the top surface, resulting in compressive stresses and plastic deformation near the weld region. As the top of the substrate begins to cool and contract, longitudinal bending is observed due to the inability of the plastically deformed material to recover.

Vasinonta et al. investigated the effects of preheating as a method of process optimization and distortion reduction through thermal model analysis of the deposition of stainless steel thin walls on small substrates [14]. The study found that uniform preheating of the substrate to a temperature of 400 °C led to a reduction in residual stress by 40%. Generally in DED repair, the repair part is either too large to be completely and uniformly preheated or the DED machine cannot support bulk heating within its processing chamber. Therefore, surface preheating is often employed as a method of distortion mitigation. An alternative approach was used by Klingbeil et al. to investigate the effect of substrate preheating on substrate warping during microcasting and welding of 308 stainless steel. The top surfaces of the substrates were manually preheated via welding torch to a temperature of 275 °C then allowed to cool to a temperature of 200 °C in an attempt to achieve a uniform temperature. Klingbeil et al. found that substrate preheating reduced the amount of substrate warping from a normalized curvature of 0.376 to 0.319 [12]. Additionally, Jendrzejewski et al. discovered through numerical calculations that base preheating reduced the amount of distortion of laser cladded stellite SF6 coating on X10Cr13 chromium steel and produced crack-free coatings when preheated to 500 °C [11].

In studying the impact of substrate preheating on distortion within powder-fed laser DED depositions of thin Ni60A walls on thin Q235 steel substrates, Zhang et al. found that heating the substrates prior to deposition reduced the amount of observed final distortion [15]. However, the study found that increasing the preheat temperature increased the amount of total accumulated distortion. Zhang et al. attributed these results to preheating leading to a higher deposition rate, which, in turn, results in an increased amount of thermal contraction during cooling.

Cao et al. investigated the effect that plate preheating had on the longitudinal bending distortion during wire-fed, electron beam DED of Ti–6Al–4V thin walls [16]. In this case, 15 mm thick substrates were preheated by scanning the electron beam, at half power, along the surface of the substrate. They found that preheating the substrate to 197.85±67 °C by this method decreased the amount of distortion by 3.5%; increasing the number of preheating scans of the electron beam, up to three times to 456.85±84 °C, decreased the amount of distortion up to 18%.

This study investigates the effect substrate surface preheating has on the distortion of laser deposited Ti–6Al–4V on substrates of the same composition. A key distinction between this and existing works is the in situ measurement of distortion and substrate temperature. Small square patches were built atop freely supported substrates, enabling real-time acquisition of distortion and temperature throughout the heating, deposition, and cooling processes. The effects of preheating and substrate thickness were analyzed and are reported here.

This study used an Optomec^® LENS^® MR-7, a DED machine employing a 500 W IPG fiber laser. The LENS machine injects metal powder through four argon-assisted (4 lpm of argon gas), radially symmetric nozzles. The four nozzles surround a central coaxial nozzle that directs the laser beam (Fig. 1). The substrate surface was placed 9.27 mm below the powder feeding nozzles. This distance resulted in a second moment laser beam diameter (D4σ) of 0.99 mm, measured by a PRIMES^® Focus Monitor system. The coaxial nozzle also directs 30 lpm of argon gas at the substrate to protect the focusing optics. The processing chamber of the LENS machine is filled with argon prior to processing and maintains an oxygen content below 15 ppm during processing.

The LENS machine was used to build 25.4 × 25.4 mm patches of Ti–6Al–4V onto the center of 76.2 × 50.8 mm substrates of the same composition to simulate a typical repair process. PREP^® powder, sieved to the −100/+325 mesh, producing powder diameters of 44–149 μm, was used. The substrates were freely supported within the fixture shown in Fig. 2, arranged for in situ distortion measurement with a laser displacement sensor (LDS) situated directly below the substrates. Within the fixture, a substrate was placed horizontally on top of four, level, upward-facing, cone-tipped set-screws. The substrates were then held in place by four downward-facing, cone-tipped set-screws directly above the upward-facing set-screws. Each set-screw was tightened to a torque of 0.7 N $·$ m (6.25 in⋅lb_f) to ensure consistency between depositions. The fixture was designed to hold the substrate in place while not significantly restricting deformation.

The same laser path, shown in Fig. 3, was used for all deposits, and nominal processing parameters are provided in Table 1. Three processing conditions were varied between depositions: substrate thickness, deposition thickness (set by adjusting the number of deposited layers), and initial substrate temperature (set by adjusting the preheat), as shown in Table 2. Substrates were preheated using a plate heater, which was a 63.5 × 50.8 × 12.7 mm block of steel with two cartridge heaters press-fit into 3/8 in holes drilled through the side of the steel block (Fig. 4). A thermocouple probe was fit through the center of the steel block with the tip of the probe flush with the bottom surface of the steel block and coincident with the top surface of the substrate. The thermocouple probe and the cartridge heaters were controlled by an Omega CNI8DH temperature controller, which was set to 400 °C for these experiments. The heating block was left on until the temperature of the substrate surface, measured by the thermocouple probe, reached the set temperature and was removed immediately before processing.

An additional set of experiments was conducted that involved preheating the substrate without performing deposition in order to assess the substrate distortion caused exclusively by surface preheating.

Temperature and displacement data were captured during LENS depositions using multiple thermocouples and an LDS. Type K thermocouples (Omega GG K 30) were laser welded to the surface of the substrate at locations on the top and bottom surfaces of the substrate (Fig. 5). Thermocouples measuring the top of the substrate were welded along the edges to avoid direct contact between the heater and the thermocouple. The thermocouples were placed at the same x–y position on the top and bottom surfaces in order to directly gather information related to the thermal gradient through the substrate thickness. An additional thermocouple was welded to the bottom surface 12.7 mm from the edge of the substrate to measure temperatures closer to the deposition (labeled as TC 3 in Fig. 5).

The displacement of the substrate was measured using a KEYENCE LK-031 LDS head operated with a KEYENCE LK-2001 Controller to produce measurement resolution of 1 μm and linear accuracy of ±1 μm. The LDS measured the displacement of the substrate at the direct center, directly beneath the repair patch. Temperature and displacement measurements were recorded at 100 Hz using National Instruments data acquisition hardware and LabVIEW code synchronized with a common clock, the x, y, z axis positions of the LENS system, and the laser on/off state.

Temperature measurements are shown in Figs. 6–9. Within each figure, distortion data (left axis) are overlaid with temperatures (right axis). Initiation of the deposition occurs at 0 s and the end of the deposition is indicated by a vertical, dashed line. Figures 6 and 7 show the temperature profile of the top and bottom surfaces of thin substrates, as measured approximately near the edge of the substrate (TC 1 and 2 in Fig. 5), during the deposition in each case. For deposition of three layers atop a thin substrate, the temperature difference between the top and bottom surfaces in the room temperature case (Fig. 6(a)) is negligible throughout the entire build process. However, in the preheated case (Fig. 6(b)), the top surface of the substrate is substantially hotter than the bottom surface immediately before (20.5 °C hotter) and during deposition (up to 31.7 °C). Temperature measurements from deposition of ten layers atop a thin substrate (Fig. 7) show a similar result. Additionally, the data show that for the room temperature case (Fig. 7(a)), the bottom surface of the substrate is hotter than the top surface (up to 23.1 °C), but when preheated (Fig. 7(b)), the top surface is hotter than the bottom surface (13.6 °C immediately before and up to 22.2 °C during deposition).

Temperature profiles collected from the same relative locations during deposition atop thick substrates (Figs. 8 and 9) contrast with those atop thin substrates (Figs. 6 and 7). For deposits atop thick, room-temperature substrates (Figs. 8(a) and 9(a)), temperature oscillations do not reach a steady-state envelope, rather the substrate temperatures increase with every deposited layer. With preheating (Figs. 8(b) and 9(b)), the temperature oscillations reach and maintain a near-constant envelope. Also, the top of the substrate is generally hotter than the bottom of the substrate throughout the deposition process. Similar to the response of the thin substrate, the top surface of the preheated, thick-substrate cases is hotter than the bottom surface before deposition due to the presence of the heater. However, soon after deposition begins, the temperature difference between the top and bottom surfaces is reduced.

The variation of the processing conditions resulted in differing amounts of substrate distortion. The effects substrate preheating had on distortion can be seen in Figs. 10 and 11. In Fig. 10, four graphs plot distortion against time for different substrate thicknesses (2.54 mm versus 12.7 mm) and deposition thicknesses (three layers versus ten layers). Note the differing time scales. Dashed, vertical lines show the beginning and end of deposition process, so that data to the right of the end of deposition represent distortion during the cool-down of the build. Negative distortion signifies the substrates bent downward toward the LDS. Figure 11 shows the incremental distortion per layer for each deposition (tabulated in Tables 3 and 4).

The data shown in Figs. 10 and 11 reveal three important qualities regarding substrate distortion. First, substrate preheating reduced the final amount of total distortion for thin substrates, reducing distortion by 13.4% and 3.2% in the three- and ten-layer cases, respectively. In contrast, substrate preheating increased the final amount of total distortion for thick substrates, increasing distortion by 178.3% and 44.5% in the three-layer and ten-layer cases, respectively. In fact, preheating thick substrates increased the amount of distortion for nearly all deposited layers (Figs. 11(c) and 11(d)). Only two deposited layers (layers 5 and 9) accumulation cause less distortion than the room temperature deposited layers. This slight decrease for those layers is likely due to the substrate slightly sliding within the fixture as the substrate bent due to thermal expansion.

Second, the data show that the amount of distortion observed when preheating the thin substrate is not linear throughout the process as additional layers are deposited. Comparing the thin, room-temperature to the thin, preheated deposits, there is a significant distortion reduction within the first layer of deposition (27.4%) in the preheated case, but the rate of distortion increases for subsequent layers, resulting in a small total distortion reduction in the final result (0.132 mm for the three-layer deposition and 0.061 mm for the ten-layer deposition).

Third, the data show that the preheated thick substrates experienced significant distortion during cooling, lasting up to an hour after the deposition process, itself, concludes. Upon cooling, the deposits continued to move, distorting an additional 101.38% and 52.72% of the total deposition distortion for the three-layer and ten-layer cases, respectively. This phenomenon was not observed with thin substrates—distortion appeared to stop within a minute following the completion of the deposition.

The results from the set of preheated experiments with no deposition are shown in Fig. 12. These experiments revealed that the thick substrate distorted upward during preheating, and returned to its original position during cool-down after the heater was removed. This effect was also observed within the thin substrate, however on a much smaller scale (0.03 mm as opposed to 0.09 mm in the thick substrates) and the thin substrate did not fully return to its original position.

The data show that distortion of Ti–6Al–4V DED parts is strongly dependent on the thickness of the substrate, the number of deposition layers, and the initial substrate temperature. While preheating thin substrates from the top surface reduced distortion, preheating of thick substrates actually increased distortion. Additionally, the amount of distortion accumulated is nonlinear with respect to the number of consecutively deposited layers, and significant distortion accumulates during cooling particularly for preheated, thick substrates, but is negligible for thin substrates. Most importantly, the data reveal that substrate preheating, a technique commonly used to reduce distortion during repair processes, in fact has the opposite effect on thicker Ti–6Al–4V substrates.

The temperature differential (T_top− T_bottom) is plotted in Fig. 13. This metric is based on in situ thermocouple measurements taken at an x–y location along the edge of the substrates (TC 1 and 2 in Fig. 5). Analysis of the temperature differential suggests that forced convection from argon gas flowing through coaxial and powder feeding nozzles plays an important role in explaining the distortion results. As evidence of this, thin, room-temperature substrates experience a negative temperature differential; that is, the bottom surface is predominately hotter than the top surface. Heigel et al. showed that forced convection caused by the argon gas flow through the coaxial nozzle is a significant quality that greatly impacts surface temperatures within the Optomec^® LENS^® system. Heigel et al. modeled the effect of forced convection on a flat substrate as an exponentially decaying function from the centerline of the coaxial nozzle though which argon gas is directed at the surface [17].

Thin, preheated substrates, on the other hand, show a positive temperature differences (Figs. 13(a) and 13(b)). This is attributed to an initial, large temperature difference between the top and bottom surfaces at the start of the deposition process. During the preheating cycle, the top surface is not being cooled by forced (and/or natural) convection or by radiation. When the heater is removed, and the top surface is subject to forced (and/or natural) convection and radiation, the top and bottom surface temperatures show an initial drop, but the laser process is initiated within 30 s, i.e., before the temperature differential decreases below 10 °C. The initiation of the deposition provides enough heat to the top surface of the thin substrate such that full equalization of temperatures does not occur at any time during the depositions.

Thick, room-temperature substrates exhibit a positive temperature difference (Figs. 13(c) and 13(d)). This result, contrasting with the thin, room-temperature substrates with otherwise identical deposition conditions, is simply due to the greater thermal mass and physical separation of the top and bottom surface measurements. At steady-state, the temperature profiles along the top surface of the sample become hemispherical far from the primary heat sources and sinks (laser and forced convection).

Processing of thick, preheated substrates results in a positive temperature difference for the ten-layer case (Fig. 13(d)). However, a negative temperature difference is observed for the three-layer case (Fig. 13(c)); this is due to the time delay between the end of the preheating cycle and the beginning of deposition, leading to a slightly different initial substrate temperature, which resulted in an initial, negative temperature differential as deposition commenced. The amount of time between removing the heater and beginning deposition for the three-layer case was ∼24 s longer than the ten-layer case, allowing the top surface of the three-layer case to cool to a temperature lower than the bottom surface before deposition.

It was found that preheating thin substrates reduces accumulated distortion on the first layer of deposition only, as shown in Figs. 11(a) and 11(b). During the deposition of subsequent layers, distortion accumulates slightly more rapidly on preheated, as opposed to room temperature, substrates. However, the distortion reduction on the first layer of preheated substrates exceeds the increased distortion on subsequent layers for both the three-layer and ten-layer deposits, and the final, total distortion of thin, preheated substrates was less than thin, room temperature substrates. For thin substrates, preheating the substrate prior to deposition reduces the thermal gradient near the melt pool on the first layer. However, examination of the top and bottom surface temperatures (Figs. 6 and 7) for subsequent layers shows that the bulk substrate temperature, approximated as the average of measured temperatures, is not significantly affected by preheating.

Of course, this explanation does not account for the distortion generated on thick substrates. The distortion of thicker substrates is strongly affected by the flexural rigidity of the thicker substrates. Any distortion resulting from large, brief thermal gradients, as introduced by the deposition of material on a room temperature substrate, is resisted by the stiffness of the thicker substrate.

Preheating thick substrates, as opposed to thin substrates, increases the distortion rate during deposition as seen in Figs. 11(c) and 11(d). One possible explanation for this phenomenon is that a thick, preheated substrate has a larger, local thermal gradient near the top of the substrate than the thin preheated substrate. The thermocouples on the bottom surface near the center of the thin, preheated substrates (TC 3 in Fig. 5) exhibit a peak temperature of 600 °C, whereas preheated thick substrates reached a peak temperature of just 350 °C (Fig. 14). This difference in peak temperatures on the bottom surface indicates that the thick, preheated substrates have a larger temperature differential throughout the thickness of the substrate thus resulting in more distortion. Another explanation could relate to the reduced modulus of elasticity and stiffness of Ti–6Al–4V at elevated temperatures. An increase in temperature would effectively lessen the stiffness and constraint against bending. Thermo-mechanical simulation, which can account for heating by the laser as well as cooling by the coaxial nozzle, is required to more fully understand these results.

Further, the temperature measurements of the bottom surface indicate that the increase in distortion rate of thick substrates caused by preheating may be a result of stress relaxation inherent in Ti–6Al–4V, as outlined in Denlinger et al. [18]. Substrates of materials that do not experience stress relaxation will exhibit a limited amount of distortion caused by compressive stresses generated directly beneath the deposition area as subsequent layers are deposited (Fig. 15(a)). But in Ref. [18], it was shown that when temperatures of Ti–6Al–4V parts exceed 690 °C, the approximate temperature at which they report stress relaxation occurs, instantaneous annealing and creep are observed, thus alleviating the plastic strain and stress. The temperature measurements of thick substrates suggest that temperatures reach and exceed 690 °C in a substantial volume within the substrate, thus relieving stresses and reducing constraint allowing compressive stresses directly below to cause greater distortion (Fig. 15(b)). The temperature difference between room temperature and preheated thick substrates as illustrated in Fig. 14 suggests that the stress relaxation isotherm is extended toward the substrate bottom for the first few layers of deposition indicating a reduced constraint. Again, thermo-mechanical simulations of the depositions could be used to support the relative importance of stress relaxation in the behavior of the samples.

The increase in distortion rate, however, does not, in and of itself, account for the fact that the thick, preheated substrates exhibit significantly more distortion after deposition than nonpreheated substrates. This postdeposition accumulation of distortion is likely an additional consequence of the stress relaxation that occurs at 690 °C in Ti–6Al–4V. When the substrates are preheated, the substrates will exhibit higher temperatures throughout their thickness. Figure 16 illustrates how these higher temperatures will result in an increase in the volume of the substrate encompassed by a notional stress relaxation isotherm. For thin substrates, it is anticipated that preheating will result in extending the isotherm such that it encompasses the top and bottom surfaces of the substrate (Fig. 16(a)), thus eliminating differential stress through the substrate thickness and decreasing angular distortion. In contrast, with thick substrates, preheating will result in similar extension of the isotherm, but not enough to encompass the bottom surface of the substrate (Fig. 16(b)) resulting in an added amount of distortion.

The experiment outlined in this paper on Ti–6Al–4V materials addresses the effects of substrate preheating, a commonly used method for substrate distortion mitigation, on distortion, and how the effects vary depending on deposition and substrate thickness. A custom surface heater was used to preheat the deposition surface of the substrate to a desired temperature. The out-of-plane distortion was measured during and after deposition using a laser displacement sensor. The data indicate that substrate preheating has inverse effects on distortion depending on the thickness of the substrate–decreasing distortion for deposits upon thin substrates and increasing distortion for thick substrates.

Substrate preheating significantly reduces the thermal gradient near the melt pool during the first layer of deposition, thus reducing the amount of distortion accumulated during that layer. Preheating thin substrates reduced distortion accumulated in the first layer by 27.4%. Subsequent depositions on the thin substrates, however, do not experience a reduction in incremental distortion. On the other hand, the deposition of the first layer on thick substrates is not appreciably affected by substrate preheating, likely due to the higher stiffness inherent in the thicker substrate; rather, preheating thick substrates results in an increased distortion rate during deposition, and the substrate continues to deform long after the deposition process has ended. Preheated thick substrates had distortion rates increasing by as much as 390% during deposition, and distortion continued after deposition up to an additional 101.38%. These outcomes are theorized to result from larger thermal gradients within the thick substrates and the stress relaxation inherent in Ti–6Al–4V.

The findings in this study suggest that the thickness of preheated Ti–6Al–4V substrates is a critical factor that influences not just the level of observed substrate distortion, but also the general trend of distortion. The behavior of the experimental depositions suggests that a stress relaxation isotherm extends farther through the depth of a substrate when preheated, which has a significant effect on the amount of distortion caused by compressive stresses. With thick substrates, the isotherm expands but does not reach the bottom of the substrates resulting in an increase of distortion rate and total accumulated distortion, whereas with thin substrates, the extension of the isotherm reaches the bottom of the substrates thus reducing the amount of total distortion.

This work was supported by the Air Force Research Laboratory through America Makes under agreement number FA8650-12-2-7230. The U.S. Government is authorized to reproduce and distribute reprints for Governmental purposes notwithstanding any copyright notation thereon. The views and conclusions contained herein are those of the authors and should not be interpreted as necessarily representing the official policies or endorsements, either expressed or implied, of Air Force Research Laboratory or the U.S. Government.

This work was supported by the Office of Naval Research, under Contract No. N00014-11-1-0668. Any opinions, findings and conclusions or recommendations expressed in this publication are those of the authors and do not necessarily reflect the views of the Office of Naval Research.

• Air Force Research Laboratory (Grant No. FA8650-12-2-7230).

• Office of Naval Research Global (Grant No. N00014-11-1-0668).