Characterization of cold metal transfer and conventional short-circuit gas metal arc welding processes for depositing tungsten carbide-reinforced metal matrix composite overlays

This paper compares the processing characteristics of advanced CMT (cold metal transfer) and conventional GMAW-S (gas metal arc welding with short-circuit metal transfer) processes for depositing Ni-WC MMC (nickel-based metal matrix composites reinforced with WC) overlays. In contrast to common expectations, advanced CMT technology with mechanically assisted droplet transfer could not demonstrate significant advantages over the GMAW-S process; on the contrary, CMT exhibits marginal disadvantages in terms of carbide transfer efficiency, volume fraction of retained WC, and deposition rate. Some carbides originally contained in the core of the feed wire are blown away and expelled out of the processing zone leading to physical losses of WC particles during the deposition processes, which is more significant for the CMT process owing to much higher waveform cycle frequency and cyclic feed wire retractions. CMT exhibits superior waveform stability, better control over penetration depth, marginally lower dilution level, and exceptional arc stability. The main parameters affecting carbide transfer efficiency and volume fraction of retained WC are wire feed speed and travel speed for both processes; increased wire feed speed and travel speed generally lead to decreased carbide transfer efficiency and reduced volume fraction of retained WC. Shielding gas may have different effects on the outcomes for the CMT and GMAW-S processes. CMT overlays show comparatively higher W and lower Fe concentration in the matrix, while GMAW-S overlays show a higher concentration of Fe in the matrix (due to elevated dilution level) with marginally higher matrix microhardness and more herringbone-like secondary carbide precipitates.


Introduction
Nickel-based metal matrix composites reinforced with WC (Ni-WC MMC) provide robust protection for machinery and equipment operating in harsh environments such as mining, extending the lifespan of production-critical components and minimizing maintenance expenses [1]. Thanks to the unique characteristics of WC, Ni-WC MMC overlays can not only improve wear properties but also enhance the corrosion resistance of the heavy-duty components [2]. The dissolution of WC particles into the molten matrix during the weld overlay process may negatively impact the wear resistance of the final product. Therefore, it is suggested to employ low-heat-input processes for the deposition to minimize the likelihood of thermally unstable WC particles dissolving in the molten matrix [3].
Gas metal arc welding (GMAW) is widely used for the hardfacing of large mobile mining equipment. Utilization of GMAW is driven by factors such as cost and efficiency, with a particular emphasis on productivity. Companies utilizing heavy haulers and related equipment are highly focused on sustained production. Therefore, manufacturers invest considerable effort to improve productivity to fulfill market demand and maintain market competitiveness [4]. Unfortunately, high WC dissolution is reported while using the GMAW process to deposit Ni-WC MMC overlays [3]. Günther et al. [5] quantified the effect of the GMAW process with short-circuit transfer (GMAW-S) and globular transfer mode on WC dissolution. They kept the dilution level fixed by regulating the amount of external hot wire introduced into the weld pool and demonstrated that the studied power range has minimal impact on WC dissolution. The study found that the globular transfer mode is responsible for up to 40% increase in WC dissolution compared to the shortcircuit transfer mode, caused by higher process power and subsequently elevated droplet temperature.
As a new variant of GMAW, CMT process operates in short-circuit transfer mode, but it is integrated with sophisticated waveform control and reciprocating movement of feed wire (to facilitate the separation of the molten droplet from the tip of feed wire and its integration into the molten pool). This unique control offers improved arc stability and results in low-to-no spatter. Since Fronius International GmbH introduced CMT to the market in 2004, it has seen extensive use in a wide variety of welding applications, from welding aluminum [6,7] to joining dissimilar materials [8,9]. Additionally, CMT has gained significant attention from researchers and industry for its potential application in wire arc additive manufacturing due to its high production efficiency [10,11].
According to Galeazzi et al. [12], however, wire arc additive manufactured samples deposited by CMT using AISI 420 wire did not show any significant improvement or difference in comparison to the standard GMAW, even though they observed higher arc stability of CMT over standard GMAW. These findings contradict the existing body of knowledge regarding CMT and conventional GMAW in literature. When it comes to the deposition of Ni-WC MMC overlays, the situation could be much more complicated, as tubular cored feed wire can cause challenges in achieving a stable arc [13] compared to solid feed wire.
Although conventional GMAW processes have been employed for depositing Ni-WC MMC overlays using tubular feed wires, few systematic studies have been reported to address associated excessive WC dissolution and low carbide transfer efficiency (η). As to the CMT process, there is even less work published regarding its use for the deposition of Ni-WC MMC overlays. In the authors' previous investigation [14], CMT process was successfully used for the deposition of single weld bead tracks of Ni-WC MMCs on steel plates; the process was optimized by using Taguchi's design of experiments (DOE), and the effects of processing parameters on f WC , η and WC dissolution behavior were analyzed for the CMT process. Nonetheless, the frequently asked question, whether it is beneficial to use CMT for the deposition of Ni-WC MMC overlays as compared to conventional GMAW processes, was not addressed there. To answer that question, this comparative study is to reveal the differences in the processing characteristics of CMT and GMAW-S for depositing Ni-WC MMC overlays. GMAW-S was chosen for the comparative study because it is believed to be a conventional low-heat GMAW process with diptransfer mode, which offers the most similarities as compared to the advanced CMT process. Cladding test matrixes were designed by the Taguchi DOE to assess the processing characteristics.

Materials
A36 hot rolled flat bar steel with the dimension of 100 mm × 75 mm × 10 mm was used as the substrate. A commercial tubular cored wire by COR-MET (with a nominal composition of NiCrBSi-55 wt.% WC and 1.6 mm diameter), COR ® FACE 164M, was used as feed material [15]. The initial carbide volume fraction (f p ) of cored feed wire was calculated in our previous study [14] as 40.4%, representing the theoretical maximum amount of retained carbides that were achievable in the resultant overlays.

Experimental setup and design
The experimental setup is shown in Fig. 1 , and arc length correction of −15%) synergic modes programmed by the supplier. Taguchi's design of experiment L16 orthogonal arrays were used to investigate the effects of processing parameters (WFS, travel speed (TS), and shielding gas (SG)) on f WC , η, DL, deposition rate (DR), and weld bead geometry. Table 1 is a summary of the test matrixes listing the welding control parameters for CMT and GMAW-S, respectively. The waveform characteristics of CMT (including wait current (I w ), short-circuit current (I sc ), boost current (I b ), and duration of boost phase (t b )) and of GMAW-S (including arc current (I arc ) and arc voltage (V arc )) vary with changing WFS according to the selected synergic modes.

High-speed imaging
A monochromatic high-speed camera (Photron FASTCAM Mini AX100) at an acquisition rate of 2000 fps was used to visualize the deposition process. A bandpass filter of 808 nm, corresponding to the wavelength of the illuminating CAVILUX HF pulsed high-power diode laser, was used to reduce arc interference for the imaging process. These images were also used to determine the length of the molten pool (the distance from the molten pool front to the end of the trailing liquid-to-solid boundary, as illustrated in Fig. 2). Finally, dwell time of WC particles was calculated using Eq. 1: (1) Dwell time = Length of molten pool Cladding travel speed

Characterization of the overlays
To monitor the dynamics of welding waveforms, real-time measurements of the welding current and voltage were recorded at a sampling rate of 20 kHz. In cases where needed, the data recording was synchronized with a highspeed imaging process, by automatically trig-starting at the same time. Since there may be discrepancies between heat input calculated by traditional method (using average current and voltage) and instantaneous values, and since no single efficiency factor is universally accepted for the CMT process [16][17][18], average instantaneous power (AIP) is used to compare the arc power within various processing parameters, as determined by Eq. 2.
where I i is a current sample, V i is a voltage sample, n is the number of samples, and t is the sampling duration. For this evaluation, the time interval for the data recording is 0.05 ms.
The weight of each overlay sample was measured both before and after the deposition. The deposited weight was divided by the cladding time to obtain DR.
Microstructure evaluation of the overlays was carried out by cross-sectioning the samples at the first, middle, and third quarters along the cladding bead track, respectively. Samples were polished to a mirror surface and lightly etched by 2% Nital to reveal the interface of the overlays and the substrate. Cross-section images of the samples were taken by a KEYENCE VHX-7000 series digital microscope. These images were later used to calculate DL, f WC , and η.
As shown in Fig. 3, cross-section images from the optical microscope were converted to binary images by ImageJ software. Interested regions were segmented to calculate f WC , η, and DL using Eqs. 3-5, respectively. Please note that in the calculation of f WC , both retained and primary WC are captured in image processing. However, the amount of primary WC, which are primary precipitates from dissolved carbides during the solidification process of the molten pools, is negligible compared to that of retained WC.

Microhardness
Vickers microhardness was performed on the selected samples in a vertical line through the entire cross-section using a Buehler 1600 series microhardness tester with a 500 g load, 15 s loading time, and indentation made every 0.25 mm. It should be noted that the indenter was positioned in a way to target indents on the matrix while avoiding landing on the large WC particles. Measurements were also performed on the substrate and Ni-sheath of the feed wire for reference purposes. (

Weld bead geometry
The deposited samples were scanned with a 3D laser profile scanner perpendicular to the direction of the cladding direction, with 0.1 mm stepover and 0.01 mm resolution. The scanned results were subsequently processed using MAT-LAB to determine bead geometry by fitting a polynomial curve to the bead profile, achieving a minimum R 2 of 0.9 across all samples. The reported values for W, H, W/H, and A R are the average of the scanned data throughout the entire length of the overlay bead by discarding the irregularity at the beginning and end of the bead track, as shown in Fig. 4.

Waveform analysis
Current and voltage waveforms are indicators of the arc stability and comprise three main parts: a cyclical portion stemming from the current pattern, a transient element originating from recurring processes during welding (such as droplet detachment, arc ignition and extinguishing, and short circuiting), and noise signals generated by arc disruptions, such as arc blow, shielding gas flow disturbance, and spatters [19]. In the following subsections, cyclograms and dynamic characterization of the waveforms will be discussed to evaluate the stability of the CMT and GMAW-S processes.

Cyclogram examination
A cyclogram represents the cyclic variation of voltage and current in a welding process, providing a qualitative assessment of the processing stability by showcasing distinct phases. Figures 5 and 6 illustrate the cyclograms for the CMT and GMAW-S tests, respectively. Cyclogram data follows a counterclockwise oscillation, where the density distribution of points serves as an effective means to differentiate various phases of a process and their stability. For the CMT process, the boost, wait, and shortcircuit phases can be recognized by pinpointing areas with a higher density distribution. The overall elliptical shape of cyclograms for CMT samples stays consistent across various control parameters, while the area expands as WFS increases which correlates to an enlarged range of processing current and voltage. The position of high-density points shifts with higher WFS along the x-coordinate (current), while their location along the y-coordinate (voltage) remains relatively stable. Furthermore, at a constant WFS, TS does not significantly alter the shape or size of cyclograms; however, O 2 -containing SG appears to have a smaller cyclogram periphery, indicative of greater stability [20].
For the GMAW-S tests, cyclograms exhibit more dispersion and increased noise data, implying a lack of regularity and occurrences of unstable events. As shifts along the y-coordinate (voltage) for high-density points. Additionally, it appears that a SG mixture containing 25% CO 2 extends the periphery along the horizontal direction (i.e., x-coordinate), indicating reduced stability.

Waveform comparison
Typical waveforms for the CMT and GMAW-S processes with comparable processing parameters are shown in Fig. 7.
High-speed camera images are also presented, to demonstrate the characteristics of arcing and/or droplet forming/transfer at different moments of the waveform cycle. CMT cycle consists of two main phases: short-circuit and arcing ones. The arcing phase is further divided into the boost and wait phases. During the boost phase, the current quickly rises, reigniting the arc between the electrode and substrate, causing droplet formation at the electrode tip. The wait phase follows the boost phase; as the electrode moves toward the weld pool, the current decreases to prevent the globular transfer of the droplet, and the voltage changes (based on the selected program) to maintain the arc intensity. When the molten droplet gets in touch with the weld pool (initiating the short-circuit phase), the current starts to drop sharply, and the electrode retracts, facilitating the detachment of the molten droplet from the tip of the feed wire and its transfer into the weld pool. The duration of the short-circuit phase cannot be adjusted once a CMT synergic mode has been chosen [21]. The GMAW-S cycle includes short-circuit and arcing phases. The short-circuit phase begins with the molten droplet touching the welding pool (forming a metallic bridge), which is reflected by a sudden increase in the current and a rapid drop in voltage. This current increase leads to a reduced conduction area and intensifies the pinch force, leading to the droplet detachment from the tip of the feed wire at the peak of the current spike and the rupture of the metal bridge. Thereafter, the arcing phase restarts and the arc reignites, while the current drops and the voltage increases to the preset value, enabling the formation of a new droplet. The periodic oscillations in current and voltage throughout these phases are crucial in controlling the metal transfer and the arc stability for the GMAW-S process [22].

Analysis of waveform characteristics
Figures 8 and 9 demonstrate the effects of WFS, TS, and SG on the actual processing waveform characteristics (such as the mean current, power, and duration of each phase) of the CMT and GMAW-S tests, respectively, as compared with the preset values if applicable. Each data represents a mean value of measurements from several tests that have the same processing parameter under assessment (see Table 1 for experimental details).
In the case of the CMT process, the actual duration of the boost phase rises with increased WFS, following closely with the preset value (t b-preset ) as shown in Fig. 8a; nonetheless, t b remains constant with changing TS and SG. On the other hand, t w and t sc have not been preset in the CMT synergic program and change with the processing parameters (such as WFS, TS, and SG). t w is governed by the time needed for the droplet to touch the weld pool; therefore, it is dependent on WFS, the melting rate of the electrode, and the diameter of the molten droplet. It can be observed that increased WFS and TS lead to decreased t w (Fig. 8a). While the dependence of t w on WFS is obvious, the decrease of t w with increased TS may be due to a slight rise in arc intensity because the electrode interacts more with the substrate and less with the molten pool with increased TS.  seem to change t sc much, with an exception for 15% CO 2 SG mixture where higher t sc is observed.
In the case of GMAW-S, t arc fluctuates significantly (Fig. 8d), suggesting lower stability compared to the CMT tests. Although GMAW-S and CMT have different droplet detachment mechanisms, the range of their t sc is comparable.
As shown in Fig. 9, the average cycle frequency for CMT is 56 Hz, which is much higher than that for GMAW-S (19 Hz). Fluctuation of the cycle's frequency, ranging from 50 to 63 Hz, is small (less than ± 12% from their mean) for CMT, whereas the fluctuation of cycle frequency for GMAW-S, ranging from 11 to 27 Hz (i.e., about ± 42% from their mean), is much higher than that for CMT. This demonstrates the superior waveform stability of CMT as compared to GMAW-S.
In the case of CMT, I b , I w , and I sc all increase with rising WFS closely following their preset values, and they are not affected by TS and SG (Fig. 8b), indicating great waveform stability for the CMT process at various processing conditions. Comparatively, for GMAW-S, actual processing I arc also increases with rising WFS but is higher than the preset value (Fig. 8e). In addition, both I arc and I sc fluctuate slightly with changing TS and SG mixtures. All those are indications that the processing waveforms for GMAW-S are much more difficult to control and less stable, as compared to those for CMT. It is interesting to note that I arc in GMAW-S is in the same range as I b in the CMT tests. Nonetheless, I sc for GMAW-S is much higher than that of CMT in a comparable processing condition, that is because a higher current is needed for the droplet detachment during the GMAW-S   Mean of cycle frequency Fig. 9 Effects of processing parameters on the mean of cycle frequency for CMT and GMAW-S tests, respectively process leading to higher arc instability, while the droplet detachment is facilitated by wire retraction at a lower current level for the CMT process leading to less spatter. Consequently, the mean P sc for CMT (varying from 15 to 122 W as shown in Fig. 8c) is much lower than that of GMAW-S (varying from 209 to 1198 W as shown in Fig. 8f) in a comparable processing condition. Although P b for CMT is consistently greater than P arc for GMAW-S across all tests, the total power for CMT is lower than that of GMAW-S, due to much shorter t i than t arc .

Physical loss of WC during deposition
In addition to WC dissolution, black residues containing WC particles were found along bead tracks of Ni-WC MMCs deposited by the CMT process, providing conclusive evidence that some WC particles were not caught up into the molten pool during deposition operation as reported in detail elsewhere [14]. Physical loss during the deposition process was also found to be one of the causes for reduced f WC in similar weld overlays produced by the GMAW process with both short-circuit and globular transfer modes [23][24][25]. Guest et al. [24] determined the loss mechanism as the non-wetting behavior of WC particles. Utilizing high-speed camera images, they demonstrated that a portion of WC particles could not penetrate into the molten pool due to surface tension, either bouncing directly off or sitting on the molten pool surface before being further expelled by the welding arc. Figure 10 shows high-speed camera images of typical CMT and GMAW-S tests, obtained during this investigation. It can be observed that some WC particles are blown away and wasted without having a chance to encounter the weld pool. This is different from the non-wetting behavior of WC particles as Guest et al. [24] documented before. For CMT, there are mainly two instances for the physical loss of WC particles: (1) during arc ignition and/or re-ignition (at the beginning of the boost phase) and (2) after the detachment of the molten droplet while retracting the electrode (as shown in Fig. 10a). For the former case, it may be due to excessive and uneven electromagnetic force during the arc initiation phase in association with the tubular electrode; for the latter case, particles within the open tubular sheath may fall out due to gravitational effect exacerbated by the feed wire retracting movement before the formation of the next droplet, and then blown away by flowing protective gas and electromagnetic force.
For GMAW-S, the loss of WC particles primarily occurs either during arcing phases especially at the current peaks or after droplet detachments leaving the tip of the tubular feed wire open. In addition, the molten droplet can be expelled away from the processing zone from the open side of the electrode (as clearly demonstrated in Fig. 10b). Because of cathode spot relocations which could often be the case, especially for tubular electrodes [13], a weld droplet formed at the tip of the tubular feed wire can be attracted to one side at the tip, leaving another side of the electrode tip exposed and open from which WC particles are expelled away from the processing zone. This third type of physical losses is more prevalent for the GMAW-S process (which may be due to unstable arc), but can also be observed for the CMT process.
For the current investigation, the highest η achieved among all the tests is approximately 92.5% (see Table 2), indicating that physical loss of WC particles can be kept as low as 7.5% of the initial carbide amount in cored feed wire. For tests 1-4 of CMT and GMAW-S where the thermal input is at the lower range, low f WC should be mainly due to physical losses, as WC dissolution is negligible; on that base, it can be reasonably assumed that physical loss can go as high as 44.5% of initial carbide amount in cored feed wire. Lower f WC and η of CMT samples compared to GMAW-S ones with identical processing parameters (see the following subsection for details) can be attributed to the higher physical losses, due to more frequent events associated with detachments of molten droplets and arc re-initiations as well as exacerbated gravitational effects by wire-retracting actions for the CMT process. The increased cycle frequency (see Fig. 9) might lead to a larger physical loss of WC particles. Figure 11 presents the top view of CMT and GMAW-S bead tracks, respectively. Table 2 summarizes the experimental results regarding AIP, f WC , η, dwell time, and characteristics of bead geometry. The bead profiles are continuous and relatively smooth, and bead surfaces are uniform without any noticeable defects or irregularities. These characteristics are crucial for cladding purposes where subsequent tracks can be deposited avoiding porosities. In tests 1-4 of CMT, the bead profiles are narrow which is not ideal for cladding applications. Low WFS causes insufficient arc power and feed wire melting, which reduces DR and results in lower deposition volumes. Figure 12 exhibits the effects of WFS, TS, and SG on f WC , η, and weld bead characteristics (DL, P, and W/H), respectively. Each data represents a mean value of measurements from several tests that have the same processing parameter to be assessed.

Effect of processing parameters on the overlay characteristics
For CMT, it can be observed that increased WFS and TS lead to a decrease of f WC and η. For GMAW-S, a similar trend is generally evident with the exception of WFS of 3.9 m/min, where there is a slight increase in f WC and η compared to WFS of 2.1 m/min; this can be attributed to the low AIP for WFS of 2.1 m/min leading to an unstable arc and, thus, larger physical loss of WC particles.
In general, higher WFS (which is related to higher processing current and AIP as discussed earlier) correlates with elevated arc intensity (and thus higher electromagnetic force) and higher HI, facilitating both physical loss of WC particles and WC dissolution in molten pools during the deposition process for both CMT and GMAW-S.
It is commonly believed that increased TS correlates with reduced HI and, thus, lower WC dissolution; this should lead to larger f WC and η, a case that was observed previously with standard GAMW [5,26]. Obviously, the outcomes of the present study contradict this "common sense." For current tests of the low-HI processes, the effect of physical loss (rather than that of carbide dissolution) may be predominant, which may be intensified with increased TS resulting in lower f WC and η. In the case of the CMT process, it seems that both the cycle frequency of processing waveform and P b in each cycle rise with increased TS, which may be the main cause of intensified physical loss of WC particles. For GMAW-S, P arc in each cycle of the processing waveform seems to get slightly larger with increased TS, which may       be one contributing factor to the increased physical loss. In addition, with increasing TS, the arc intensity could be strengthened even at the same power level, because the electrode interacts more with the substrate and less with the molten pool, which may also promote the physical loss of WC particles during the deposition process. For the CMT tests, CO 2 content in the SG mixture does not appear to significantly impact f WC and η; however, an increasing trend in f WC and η is observed with the increasing level of O 2 in the SG mixture, which may be a reflection of the fact that O 2 addition results in lower waveform cycle frequency and P b leading to lower physical loss of WC particles during the deposition process. Guest et al. [24] observed improvements in η with the addition of up to 15% O 2 for GMAW processes with both short-circuit and globular transfer modes. Introducing a controlled amount of oxidizing gas, such as CO 2 or O 2 , to an Ar-rich SG mixture can alter the thermophysical characteristics of the plasma arc, enhancing welding arc stability, heat transfer, surface tension, and weldability [27][28][29]. However, in the case of CMT, CO 2 addition to the SG appears to be less effective in achieving arc stability as compared to O 2 addition for depositing the Ni-WC material system. Molten nickel is resistant to oxidation, which results in fewer electron emissions and reduced stability at the cathodic locations. The presence of O 2 in SG effectively promotes the oxidation of the nickel weld pool and improves arc stability during the droplet formation [25]. Improved arc stability leads to less physical loss of carbides and increased η and higher f WC .
For GMAW-S, in contrast, a higher level of CO 2 in the SG mixture increases f WC and η, whereas the increase of O 2 content to 5% lowers f WC and η although 2% O 2 seems to have a negligible effect. To that end, more studies are still needed to have a full understanding.
For both CMT and GMAW-S, WFS is the primary factor influencing weld bead characteristics and DR (see Fig. 12c-f). Higher WFS leads to an increase in AIP, resulting in greater substrate melting (directly increasing P, DL, and W/H).
Higher TS also increases DL, but its effect on the DR, P, and W/H remains either marginal or inconclusive for both CMT and GMAW-S. The effect of TS on bead geometrical characteristics is dependent on two combined effects of HI and arc impingement, which compete with each other [30,31] as also discussed in our previous work [14].
The effect of the SG mixture on DL for CMT seems to be also different from that for GMAW-S. For CMT, DL remains almost unchanged when using SG mixture with CO 2 addition; however, O 2 addition results in relatively lower DL. In contrast, O 2 addition results in relatively higher DL for GMAW-S. Furthermore, O 2 addition in the SG mixture seems to result in lower W/H as compared with those with CO 2 addition for CMT, whereas lower levels of W/H seem to be the case for GMAW-S with CO 2 -containing SG mixtures. SG compositions do not have a significant effect on the DR, except for slightly lower values for GMAW-S with 5% O 2 . Figure 13 illustrates the effect of AIP on the quality criteria of all individual overlays. There is no strong correlation between f WC or η and AIP. However, a subtle decreasing trend can be observed as AIP increases for both CMT and GMAW-S. As shown in Fig. 13, DR, dwell time, DL, P, A R , W, H, and W/H show in general a positive correlation with AIP.
Interestingly, GMAW-S and CMT results in terms of AIP suggest no superiority of CMT over GMAW-S, except for better control in P, and slightly lower DL. Overall, GMAW-S may even slightly outperform CMT with respect to f WC , η, and DR. W of the overlays shows comparable results and positive correlation with AIP for both CMT and GMAW-S tests. However, the impact of AIP is less profound on H. For instance, W of samples varies between 3.35-10.16 mm for CMT and 2-15.3 mm for GMAW-S, while H of samples varies only between 1.45-2.35 mm for CMT and between 1.3-3.1 mm for GMAW-S, respectively. In other words, higher AIP contributes to a higher melting rate of the electrode which mostly leads to wider beads rather than increasing bead height.

Microstructure and microhardness
CMT and GMAW-S test IDs 3, 7, 10, and 16, respectively, were selected to investigate the effect of processing parameters on the microstructure and microhardness of the overlays. The corresponding AIP for these tests is 0.85, 1.36, 1.89, and 2.95 kW for CMT and 1.28, 1.88, 2.84, and 3.64 kW for GMAW-S, respectively. Figure 14 shows the micrographs of these samples with EDS measurements for Fe and Ni.
A previous paper by the authors [14] extensively characterized the microstructure of CMT samples which consisted of the retained WC fairly uniformly distributed within the matrix of γ-Ni and eutectic Ni/Ni 3 B. Degradation seams are not observed around the retained WC particles. They exhibit an excellent metallurgical bond with the adjacent matrix. Nevertheless, WC particles inevitably experience partial dissolution in the molten pool, leading to an increased concentration of W in the Ni-based matrix. Accordingly, primary WC precipitates can be seen in all the overlay samples. As indicated by the EDS results presented in Fig. 14, the concentration of W (tungsten) increases toward the top of the overlay, possibly due to the effect of partially directional solidification starting from the interface with the substrate and due to the longer dwell time of WC in liquid metal at the upper portion of an overlay (increasing the likelihood of WC dissolution).
For CMT samples, secondary carbides are also observed in the form of W 2 C and/or W-rich carbides for tests with AIP smaller than 1.9 kW and in the form of M 6 C (Ni 2 W 4 C near the top and Fe 3 W 3 C closer to the interface) for those with AIP greater than 2 kW, respectively. Fe enters into an overlay due to dilution and, consequently, higher content of Fe in the matrix closer to the interface can be measured.
For GMAW-S samples, secondary carbides in the form of W 2 C, W-rich carbides, and/or M 6 C can be observed in all the tests. The major difference in the microstructure is the formation of herringbone carbides near the interface of GMAW-S samples at lower AIP (i.e., even for a microstructure-examined sample with the lowest AIP of 1.28 kW), while the same carbide is only observed at relatively higher AIP (greater than 1.89 kW) for CMT samples. This can be attributed to a higher concentration of Fe in the matrix due to elevated DL for GMAW-S specimens (as confirmed by EDS results shown in Fig. 14b) as compared to CMT ones, which facilitates the precipitation of secondary carbides (such as Fe 3 W 3 C) in GMAW overlays.
CMT samples exhibit higher W concentration and lower Fe concentration, as compared with GMAW-S ones with comparable AIP (i.e., GMAW-S samples 3, 7, and 10 versus CMT samples 7, 10, and 16, respectively). A possible explanation is that, for GMAW-S specimens, W is being consumed in the precipitation of secondary carbides (such as Fe 3 W 3 C) promoted by higher DL, leaving the matrix with less W concentration after solidification. Figure 15 shows indentation marks across the centerline of the overlays' cross-section and their average values. As references, the hardness of the tubular Nisheath and the substrate was measured, with the average value being about 202 and 218 HV 0.5 , respectively.
By avoiding large WC particles, the measured microhardness is most related to the matrix. That is in line with the fact that no cracks were observed around the majority of the Vickers indentations and that cracks can only be seen occasionally (in the matrix or when smaller WC particles were being pressed by the indenter). The large variations of the hardness numbers can be partially due to the interactions of the indenter with small secondary carbides (or occasionally even with larger primary WC), especially for GMAW-S samples where secondary carbides are more predominant. The much higher hardness of the matrix than the nickel sheath is attributed to a combination of several factors, such as solid solution strengthening (by alloy elements, such as Si and B contained in the core of the feed wire as well as W and Fe mixing into the substrate during the deposition process), dispersion strengthening mechanism [31] (e.g., associated with small secondary carbides), and grain refinement strengthening [32] (i.e., grain refinement due to the presence of retained and primary carbides).
In general, microhardness values of GMAW-S samples are slightly higher than that of CMT ones with roughly comparable AIP values. This may be attributed to the fact that there is a higher volume fraction of secondary carbide precipitates in the matrix of a GMAW-S sample compared to a comparable CMT one.

Conclusions
Short-circuit GMAW processes with/without mechanically assisted droplet transfer (i.e., CMT vs. GMAW-S) have been investigated for depositing Ni-WC MMC overlays, leading to the following conclusions: • In contrast to common expectations, advanced CMT technology could not demonstrate significant advantages over GMAW-S; on the contrary, GMAW-S exhibits marginal advantages over CMT in terms of carbide transfer efficiency, volume fraction of retained WC, and deposition rate. • Some carbides contained in the core of the feed wire are blown away and expelled out of the processing zone: (1) during arcing phase, especially at arc ignition and/or re-ignition stages, due to excessive and uneven electromagnetic force; (2) owing to particles dropping out of the open tip of the tubular feed wire (due to gravitational effect) after the detachment of a molten droplet (and then blown away); and (3) from local opening (due to a weld droplet unevenly attached to the tip of the tubular feed wire). The first two mechanisms prevail for the CMT process. • The physical loss of WC particles is more significant for the CMT process owing to much higher waveform cycle frequency and cyclic feed wire retractions, despite superior waveform stability and better control over penetration depth, marginally lower dilution level, and exceptional arc stability. • Cyclograms indicate that the CMT process has greater process stability, higher data consistency, and lower data noise, while the GMAW-S process exhibits a lack of regularity, lower process stability, and greater data scattering and noise. • For both processes, the carbide transfer efficiency and the volume fraction of retained WC are generally negatively correlated with the wire feed rate and the travel speed. • For GMAW-S tests, the higher level of CO 2 in shielding gas mixture improves the carbide transfer efficiency and the volume fraction of retained WC, whereas the addition of O 2 in shielding gas demonstrates the same effect for CMT. • The microstructural characteristics of CMT and GMAW-S overlays are mostly similar. CMT overlays exhibit comparatively higher W but lower Fe concentration in the matrix. Herringbone-like M 6 C secondary carbide precipitates are more prevalent in GMAW-S overlays, supposedly due to an elevated dilution level. • GMAW-S overlays demonstrate marginally higher matrix microhardness.