Abstract
Additive manufacturing (AM) is a manufacturing method that can build high-strength materials layer-by-layer to form complex geometries. Previous studies have reported large variations in the mechanical properties of materials made by this process. One of the key factors that may contribute to variations within and among parts made by this process is a difference in the material’s microstructural phase and composition. A continuous cooling transformation (CCT) diagram is a useful tool that can be used with a thermal model for microstructure design and manufacturing process control. However, traditional CCT diagrams are developed based on slow and monotonic cooling processes such as furnace cooling and air cooling, which are greatly different from the repetitive heating and cooling processes in AM. In this study, a new general methodology is presented to create CCT diagrams for materials fabricated by AM. We showed that the effect of the segmented duration within the critical temperature range, which induced precipitate formation, could be cumulative. As multiple cooling processes occurred in a short time, and the temperature drops at a high cooling rate, a constant average cooling rate was assumed when constructing the CCT diagram. Inconel 718 parts fabricated by selective laser melting were analyzed. The accumulated duration required for γ′/γ″ precipitate formation was found to shift from at least 432 s to about 83.38 s. The large difference in the duration (around 518%) highlights the importance of creating CCT diagrams specifically for AM materials. The key factor contributing to phase transformation was identified as the accumulated duration within the critical temperature range. The presented methodology demonstrated the capability of combining a thermal model and experimental observation to quantitatively predict phase transformation and could be used to design microstructures and control AM processes.
1 Introduction
The mechanical properties of materials fabricated by additive manufacturing (AM) are known to have large variations compared with those of traditionally fabricated materials. Li et al. reviewed the fatigue failure data of Ti–6Al–4V parts made by AM and found that the cycles to failure for wrought parts were around 1000 times longer than that for additively manufactured parts with no surface and heat treatments [1]. Shamsaei et al. [2] studied the tensile properties of materials made by AM and compared them with wrought counterparts, and they found that the ultimate stress was 29% larger for additively manufactured 316 stainless steel (SS) [3], the yield stress of DLD 316L SS parts was 202% higher [4], and the elongation to failure of IN-738 parts made by AM was 276% higher than that of wrought parts [4,5]. Lass et al. [6] conducted a series of experiments and reported that the precipitation formation and behavior were different for materials made by AM. One of the key factors that may contribute to variations in the mechanical properties within and among parts is a difference in the material’s microstructural phase and composition.
It is critical to predict the phase transformation for materials made by different manufacturing processes to ensure that fabricated parts meet product performance requirements. This is especially true for parts made using high-strength alloys for severe operating conditions. These materials can be hardened by fine and uniformly distributed precipitation. Acharya et al. reported that the hardness of Inconel 718 (IN718) parts ranged from 180.40 to 307.50 BHN and that the hardness of Nimonic 263 parts ranged from 165 to 378 BHN due to the formation of hardening precipitates, formation of coarsening precipitates, and over-aging through different heat treatments [7]. Therefore, a revised continuous cooling transformation (CCT) diagram is essential to obtain a precise estimation of the material’s performance to handle these variations.
Recently, there has been growing interest in the prediction of phase transformation. Time–temperature transformation (TTT) diagrams and CCT diagrams are widely used as a reference to predict phase transformation. These methods are often coupled with heat transfer models and compute phase transformation based on the temperature profile. CCT/TTT diagrams can be generated using software such as thermocalc2 and jmatpro [8], which are based on the classical Johnson–Mehl–Avrami–Kolmogorov (JMAK) formulation [9] and nucleation theory [10–12] developed with traditional manufacturing methods such as furnace cooling and air cooling. Makiewicz [13] and Kelly [14] applied these methods to develop a computational model that can predict the phase transformation for AM. Zhang et al. [15] developed a framework to link a thermal model to microstructure features. Several studies [16–18] compared post-heat treatment parts made by AM with traditional TTT diagrams and found that the boundary of precipitation was shifted left to a shorter duration for phase transformation. Promoppatum et al. [19] developed a thermal model, observed the microstructure, and compared it with a traditional CCT diagram for IN718 parts made by selective laser melting (SLM). However, only one sample design and one processing condition were used, and the critical condition that may lead to precipitation was not identified. The central issues for these studies are either based on traditional phase transformation mechanism or only have one simulation result that cannot be replicated for different processing conditions and microstructure design. It remains unclear whether CCT diagrams developed based on traditional methods can be used for materials made by AM. Therefore, a systematic methodology is required to create CCT diagrams based on AM.
Another stumbling block to reconstructing CCT diagrams for AM materials was the lack of efficient computational temperature models. To simulate an accurate temperature history for the cycling, the heating and cooling process of AM was computational expensive. Morville et al. took around 4 weeks to simulate five layers of the direct laser metal deposition sample with 35 mm long [20]. Khairallah and Anderson spent 100,000 h of CPU time to compute the three-dimensional (3D) mesoscopic model with a size of 1000 µm × 300 µm × 50 µm [21]. Constructing a CCT diagram is time-consuming since it is a combination of sample simulation and experimental results. Recently, increasing interest in the temperature profile has heightened the need for a computational tool suitable to predict the temperature of materials made by AM. Bandyopadhyay and Traxel have conducted a comprehensive literature review of metal AM modeling [22]. The analytical heat transfer models [23–27] make it possible to compute the AM temperature history more efficiently. Ou and Richard Liu developed a computational efficient approach that can adjust the computational time with acceptable accuracy required [26]. As a result, these efficient temperature models should be used to help identify the inconsistency between CCT diagram for traditional processing condition and that of AM.
In this study, IN718 samples made by SLM were analyzed. IN718 is a precipitation-strengthened Nickel-based superalloy. The alloy matrix is mainly based on a gamma (γ) matrix with a FCC crystal structure. Ni3Nb precipitate (γ″) is the primary precipitate that strengthens the materials, and Ni3(Al,Ti) precipitate (γ′) is the secondary precipitate [28–30]. Based on the study by Makiewicz [13], the critical temperature for γ′/γ″ precipitate formation is the range of 893–993 K. Therefore, this study focused on examining the γ′/γ″ microstructure and the correlation with the computed temperature profile for SLM.
The objective of this study was to provide a general methodology to create CCT diagrams for materials made by AM. The presented methodology combined a thermal model and experimental observation to quantitatively predict phase transformation. This can be a useful tool for microstructural design and AM process control.
2 Methods
The systematic methodology for creating CCT diagrams required both experimental analysis and temperature simulation. The procedure is presented in Fig. 1. First, the manufacturing parameter was selected, and the component geometry was designed. Second, samples were fabricated, and the region of interest that may lead to phase transformation was determined. Third, the microstructure, precipitate formation, and phase composition of the concerned area were analyzed. Fourth, the temperature profile was simulated, and the segmented duration within each temperature range was computed. Fifth, the accumulated duration within the critical temperature range was computed, and the boundary was created. Sixth, the experiment can be redesigned, and the loop was continued until there were enough data points to create an entire CCT diagram.
For the new proposed CCT diagram, it was hypothesized that the effect of the segmented duration within the critical temperature range, which induced precipitate formation, could be cumulative. In addition, as AM processes involve multiple heating and cooling processes in a short time, and the temperature drops at a high cooling rate (around 105–107 K/s [18]), the time (x-axis) measured in the CCT diagram was revised as the accumulated duration within the temperature range, and a constant average cooling rate was assumed.
2.1 Sample Design and Manufacturing Process Design.
In this study, IN718 powder was purchased from AMC Powders Co., Ltd. (Beijing, China), which had a particle size ranging from 15 to 53 μm. The Mlab cusing R SLM machine was used to fabricate the samples. To compare the microstructure deviations of the samples at different locations, nine samples with three different heights (25 µm, 2.5 mm, 5 mm) were designed, as listed in Table 1. All the samples were made with the same laser processing conditions presented in Table 2 using the bidirectional scanning mode for both odd and even layers (Fig. 2). Figure 3 shows the fabricated samples. The IN718 substrate was used to cool down the samples.
2.2 Microstructure Observation.
To examine the microstructures at the top and bottom areas of the samples, the as-fabricated samples were polished using carbide sandpaper with ANSI grits of 400, 600, 800, 1200, and 2000. The samples were etched after polishing, immersed into Kallings reagent for 1.5 min, and rinsed with water. Kallings etchant was composed of 50 ml of HCl, 50 ml of ethanol, and 2.5 g of CuCl2 based on the ASTM E407 standard.
Microstructure and phase composition analysis can be performed using Kurt’s method [13] to quantify precipitation. Both low-magnification micrographs and high-magnification scanning electron microscope (SEM) images were taken for different regions. In this study, the aim of high-magnification observation was the identification of nano-scale γ′/γ″ precipitates, which are critical for strengthening the nickel alloy. The potential region that may have γ′/γ″ precipitates can be determined with low-magnification observation, and the detailed γ″/γ′ precipitate composition can be analyzed by high-magnification observation and automated particle analysis [31].
First, samples with a single layer were examined in the middle of the top area. Figure 4 shows the microstructures after etching. The grains grew in the direction parallel to the scanning direction induced by the heating and cooling process. In high-magnification figures, equiaxed microstructures were observed; however, there was a lack of γ′/γ″ precipitates.
Second, the microstructures of the samples with 2.5 mm height were examined in the middle of both the top area (Fig. 5) and the bottom area (Fig. 6). Equiaxed microstructures were observed at the top of the samples; however, there was a lack of nano-scale precipitates. However, at the bottom of the sample, columnar dendritic microstructures were observed, and there were some γ′/γ″ precipitate precursors.
Third, the microstructures at the top and bottom areas of 5 mm height were examined (Figs. 7 and 8). The top area of the sample with 5 mm height was similar to that of samples with 25 μm and 2.5 mm height, which lacked precipitates. However, as shown in Fig. 8, some γ′/γ″ precipitates were found in columnar dendritic microstructures at the bottom of the samples. The experimental results indicated heterogeneous microstructures at different locations. This difference may be largely attributed to the variation in the temperature profile at different times of the heating and cooling processes.
2.3 Temperature Simulation.
Unlike traditional heat treatments, such as furnace cooling and air cooling, AM includes cyclic and repetitive heating and cooling processes with high heating and cooling rates. Zhang et al. reported that the cooling rate for SLM was around 1.4 × 106 K/s compared with the rate in a casting process, which was around 3.75 K/s [32]. In addition, as AM builds materials layer-by-layer, the bottom areas would be subjected to multiple heating and cooling processes, leading to microstructure differences at different positions.
In this study, an analytical model developed based on Carslaw and Jaeger [23] and a revised model [24–27] were utilized as shown in Eqs. (1)–(4):
The temperature profile and segmented duration within each temperature range for the sample with a single layer (height: 25 μm) are shown in Fig. 9. As the laser was moved from the edge to the point of interest (x, y, x), the peak temperature started increasing. However, the temperature did not remain constant. It oscillated as the relative distance between the point of interest (x, y, x) and the laser position (xi, yi, zi) changed. It took around 2.28 s to scan each layer. The critical temperature range (893–993 K) [13] for inducing the γ′/γ″ precipitate formation is marked in the figure.
For comparison, as the sample height was increased, the temperature profile of the top area was similar to that observed with single layer. However, the temperature profile of the bottom area was different (Fig. 10). The segmented duration within the critical temperature range for the bottom area was decreased for each scan when the laser scanned the top area. It is important to note that Fig. 10 only shows the temperature profile of one layer when the laser scanned the top area. The bottom area for 2.5 mm sample was subjected to multiple scanning cycles for 100 layers, which required 228 s. On the other hand, the bottom area for 5 mm sample was subjected to scanning cycles for 200 layers, which required 456 s. The accumulated duration within the critical temperature range that may induce precipitate formation was longer than that for the top area.
3 Results and Discussion
Equiaxed microstructures were observed at the top of all samples; however, there was a lack of γ′/γ″ precipitates. However, in the microstructure at the bottom of the sample with 2.5 mm height, there were some γ′/γ″ precipitate precursors. In addition, at the bottom of the sample with 5 mm height, there were some γ′/γ″ precipitates in the columnar dendritic microstructures. The formation of the equiaxed microstructures and columnar dendritic microstructures may be associated with the number of heating cycles and the cooling rate [18]. As the layers at the top were subjected to fewer scanning cycles, equiaxed microstructures were mainly observed. Following a series of heating and cooling cycles, the equiaxed microstructures were transformed into dendritic microstructures at the bottom.
Precipitate formation may also be associated with the temperature profile of each layer. In this study, the temperature profiles at different locations were computed. The critical temperature range (893–993 K) [13] for inducing γ′/γ″ precipitate formation was marked. As the sample height was increased, the segmented duration within the critical temperature range for the bottom area was decreased for each scan when laser scanned the top surface area, as shown in Fig. 9. However, as the bottom area was subjected to a longer processing time and more heating cycles, the accumulated duration within the critical temperature range for the bottom areas was longer (Table 3).
Sample height | Position | Accumulated duration within 893–993 K (s) | Total processing time (s) | Average cooling rate (K/s) | Average cooling rate at melting temperature (K/s) | Microstructure | γ′/γ″ precipitation |
---|---|---|---|---|---|---|---|
2.5 μm | Top | 0.37 | 2.28 | 1.406 × 105 | 4.717 × 105 | Equiaxed | Lack |
2.5 mm | Top | 0.372 | 2.28 | 1.403 × 105 | 3.669 × 105 | Equiaxed | Lack |
2.5 mm | Bottom | 39.615 | 228 | 1.444 × 105 | 7.383 × 106 | Dendritic | Lack but some precursors |
5 mm | Top | 0.377 | 2.28 | 1.399 × 105 | 2.452 × 105 | Equiaxed | Lack |
5 mm | Bottom | 83.38 | 456 | 1.407 × 105 | 1.144 × 107 | Dendritic | Yes |
Sample height | Position | Accumulated duration within 893–993 K (s) | Total processing time (s) | Average cooling rate (K/s) | Average cooling rate at melting temperature (K/s) | Microstructure | γ′/γ″ precipitation |
---|---|---|---|---|---|---|---|
2.5 μm | Top | 0.37 | 2.28 | 1.406 × 105 | 4.717 × 105 | Equiaxed | Lack |
2.5 mm | Top | 0.372 | 2.28 | 1.403 × 105 | 3.669 × 105 | Equiaxed | Lack |
2.5 mm | Bottom | 39.615 | 228 | 1.444 × 105 | 7.383 × 106 | Dendritic | Lack but some precursors |
5 mm | Top | 0.377 | 2.28 | 1.399 × 105 | 2.452 × 105 | Equiaxed | Lack |
5 mm | Bottom | 83.38 | 456 | 1.407 × 105 | 1.144 × 107 | Dendritic | Yes |
In addition to calculating the accumulated duration within the critical temperature range, the average cooling rates for the whole fabrication process and the average cooling rates at the melting temperature (1667 K) were computed. Previous studies [18,32–35] have indicated that the temperature gradient at the solid–liquid interface and solidification velocity may induce microstructure differences including both grain geometry variations and phase composition variations. The average cooling rates at the melting temperature ranged from 3.7 × 105 to 1.1 × 107 K/s according to the position, which may contribute to the different equiaxed and dendritic microstructures [18]. However, the average cooling rates for the whole fabrication process were around 1.4 × 105 K/s, which were similar regardless of the position of the samples. As a result, when constructing the CCT diagram, a constant average cooling rate assumed.
Continuous cooling transformation diagrams were used as a reference for phase transformation predictions at different cooling rates. Specifically, the IN718 CCT diagram was developed and compared with experimental results obtained using traditional furnace cooling and air cooling processes [8,36]. As AM processes involve multiple heating and cooling processes, the time (x-axis) measured in the CCT diagram was revised as the accumulated duration within the temperature range, and a constant average cooling rate was assumed. However, based on previous CCT diagrams [8,36], the accumulated duration required for γ′/γ″ precipitation is at least 432 s, which is 518% longer than the experimental result (83.38 s) in this study. Therefore, it would be of interest to determine how the boundary of the γ′/γ″ precipitation line is shifted for rapid cooling AM processes.
The critical temperature range can be determined by first referring to a traditional CCT/TTT diagram and then reviewing the temperature range determined based on the AM experimental data. Acharya et al. [7] demonstrated that based on hardness measurement results for IN718, γ′/γ″ occurred between 893 and 993 K. The hardness was continuously increasing with the increase in duration when the temperature was within the critical temperature range. However, when the temperature was between 1023 K and 1073 K, the hardness was first increased and then decreased due to over-aging. The high temperature dissolved the γ′/γ″ precipitates and resulted in the coarsening of the materials.
Based on microstructure observation and temperature simulation for SLM IN718 samples, a revised CCT diagram is proposed and presented in Fig. 11. The accumulated duration spent within each temperature range was calculated based on the thermal model. A series of experiments have been conducted by Makiewicz [13] to design samples with both a single layer and multiple layers. In the experiments, the processing conditions were adjusted, so that different durations below, within, and higher than the critical temperature were considered.
To create the CCT diagram, it was assumed that the average cooling rate was constant at different locations during the fabrication process in this study. The assumption is reasonable as the average cooling rates computed in Table 3 for each location were similar at around 1.4 × 105 K/s. Furthermore, a previous study [17] also used this assumption to develop the CCT diagram for the post-process heat treatment of an SLM IN718 sample.
As shown in Fig. 11, the boundary of the γ′/γ″ precipitation line was shifted to the left when comparing the traditional CCT process with the AM (SLM) process. The results of this experiment showed that the accumulated duration required for the formation of γ′/γ″ precipitates was shorter than the accumulated duration in the case of traditional CCT. This shift may be attributed to the cyclic heating and cooling processes, which could induce phase transformation in a shorter accumulated duration. The nominal aging temperature range (893–993 K) was identified as the critical temperature range for γ′/γ″ precipitation as there were no or very minimal precipitates found in samples at higher than 993 K and lower than 893 K [13].
In summary, the accumulated duration within the critical temperature range was identified as a key factor for precipitate formation. We showed that the effect of the segmented duration within the critical temperature range, which induced precipitate formation, could be cumulative. Although the presented CCT diagram might not be accurate due to the limited experimental data, the new general methodology is important for creating CCT diagrams for materials made by the AM process.
4 Summary and Conclusions
This study demonstrated the potential of a new general methodology for creating CCT diagrams for materials fabricated by AM. We showed that the effect of the segmented duration within the critical temperature range, which induced precipitate formation, could be cumulative. The accumulated duration within the critical temperature range was identified as a key factor for precipitate formation.
Three IN718 samples with different heights (25 μm, 2.5 mm, 5 mm) fabricated by SLM were analyzed. Microstructure examination revealed equiaxed microstructures that lacked γ′/γ″ precipitates at the top of all samples. However, some precipitate precursors were found at the bottom of the 2.5 mm sample, and some γ′/γ″ precipitates were found in the columnar dendritic microstructures at the bottom of the 5 mm sample. Based on the computed temperature profile, the accumulated duration required for γ′/γ″ precipitation was shifted from at least 432 s to 83.38 s. The created CCT diagram demonstrated the capability of microstructure design to meet product performance requirements when combined with a thermal model to quantitatively predict phase transformation.
Footnote
Acknowledgment
This work was supported by the National Science Foundation (CMMI) (Grant No. 1562960). Some ideas and materials presented in this paper are patent pending granted to the owner of Purdue University.
Conflict of Interest
There are no conflicts of interest.
Data Availability Statement
The datasets generated and supporting the findings of this article are obtainable from the corresponding author upon reasonable request. The authors attest that all data for this study are included in the paper.