- Email: [email protected]
Microstructural analysis in GTA aluminum alloy welding using inverse problems
Applied Thermal Engineering 100 (2016) 333–339
Contents lists available at ScienceDirect
Applied Thermal Engineering j o u r n a l h o m e p a g e : w w w. e l s e v i e r. c o m / l o c a t e / a p t h e r m e n g
Research Paper
Microstructural analysis in GTA aluminum alloy welding using inverse problems Elisan dos Santos Magalhães, Edmilson Otoni Correa, Ana Lúcia Fernandes Lima E. Silva, Sandro Metrevelle Marcondes Lima E. Silva * Heat Transfer Laboratory – LabTC, Institute of Mechanical Engineering – IEM, Federal University of Itajubá – UNIFEI, Campus Prof. José Rodrigues Seabra, Av. BPS, 1303, 37500-903 Itajubá, MG, Brazil
H I G H L I G H T S
• • • • •
A numerical study of heating and cooling rate in a GTA aluminum welding process. A linear dependence of the grain size in relation to the positive polarity was observed. Correlations between the cooling rate and the grain size of heat affected zone (HAZ). A thermal analysis of the HAZ using software based on a numerical heat transfer model. The microstructural changes happen while the GTA arch torch was turned on.
A R T I C L E
I N F O
Article history: Received 21 October 2015 Accepted 11 February 2016 Available online 24 February 2016 Keywords: Inverse problems Heat flux Temperature estimation GTAW process Microstructural analysis
A B S T R A C T
Numerical software based on the tridimensional diffusion equation with a moving source, a phase change and heat flux estimation by the non-linear interactive Broydon–Fletcher–Goldfarb–Shanno (BFGS) inverse technique was used to study the heat affected zone (HAZ) in GTA aluminum 6065 T5 alloy welding. GTA welding experiments were performed using thin aluminum plates in four t+ experimental conditions. In previous studies, the authors determined that the peak temperature tends to increase as the positive polarity becomes higher. To confirm this behavior, the samples were cut on the welded region and later characterized using an optical microscope (OM) and a scanning electronic microscope (SEM). The heating and cooling rates, determined from an in-house code, were compared with the microstructures and the grain size of the HAZ found in the samples. The results revealed a linear correlation between the grain size on the HAZ and the positive polarity. The study also showed that a significant change in the microstructure occurs during the process while the GTAW torch is still turned on. After the welding, when the GTAW torch was turned off, the cooling rate was the same for all welded zones, which indicates that the microstructural changes had already occurred. © 2016 Elsevier Ltd. All rights reserved.
1. Introduction Although several new processes to join metals have been developed, the fusion welding process still plays a major role in engineering practice. Hence, a deep understanding of the behavior of microstructural welding is highly important for minimizing potential failures. The heat affected zone (HAZ) is a region where the microstructure is negatively affected during the welding process. The HAZ is fundamentally caused by the characteristic overheating and supercooling of the welding process. The thermal cycle that results in the recrystallization during welding is primarily composed of two steps: heating and cooling [1]. This cycle triggers a
* Corresponding author. Tel.: +55 35 3629 1362; fax: +55 35 3629 1148. E-mail address: [email protected] (S.M.M. Lima e Silva). http://dx.doi.org/10.1016/j.applthermaleng.2016.02.051 1359-4311/© 2016 Elsevier Ltd. All rights reserved.
microstructural change in the base material (BM), thereby creating the HAZ. Although the heating rate is important to achieve the fusion, the cooling rate is extremely important for the welding quality. During welding, a high cooling rate may generally be associated with the reduction of mechanical properties [2], cracking [3], deleterious phase precipitation [4], or susceptibility to intergranular corrosion [5]. Recently, an increasing interest in the impact of cooling rate on welding properties may be pointed out. For instance, Sivaprasad and Raman [6] analyzed the influence of cooling rate on microstructure and mechanical properties of alloy 718 when it is welded through two distinct welding processes, GTA (gas tungsten arc) and EB (electron beam). Manikandan et al. [7] studied the Laves phase formation in Inconel 718 in the cooling of a welding process. Aissani et al. [8] used a three-dimensional finite element model to analyze the evolution of the microstructure in the weld HAZ in a TIG
334
E. dos Santos Magalhães et al./Applied Thermal Engineering 100 (2016) 333–339
(tungsten arc gas) process. Di et al. [9] used experimental measurements to determine the average cooling rate in local dry underwater welding. The authors’ analyses suggest that a fast cooling rate, in this process, may improve the weld properties. A precise determination of the cooling rate is difficult to achieve. Several authors merely assume that the cooling rate is an intermediate value between the peak temperature and the room temperature divided by the time of cooling [10], while others obtain the value using mathematical models based on the one-dimensional heat diffusion equation, for example, in Bhattacharya et al. [11]. Another option is the use of numerical thermal models for the prediction of the microstructure parameters [12]. However, major simplifications are constantly being made to reduce the complexity of this type of model, e.g. Unfried et al. [13], who assumed a onedimensional heat flux and a one-dimensional thermal model based on Rosenthal’s equations. Another example is the work developed by Manvatkar et al. [14]. To estimate the peak temperature and cooling rate, the authors used an analytical expression proposed by Schmidt et al. [15]. The use of analytical expressions for the estimation is convenient; however, these models do not cover all experimental errors, thereby raising inaccuracies in the solution. The use of commercial software is another alternative to determine the cooling rate [16]. In contrast to the use of an analytical expression, the commercial software acts as a black box, i.e., the user does not have access to its methodology. In fact, due to certain experimental singularities, such as the non-uniformity of the torch arc and the electrical interference on measurements in aluminum welding, a three-dimensional numerical thermal model that covers all boundary conditions is highly complex. In addition, directly measuring the heat flux during the process is expensive. However, the use of inverse problems enables a quick determination of this parameter [17]. Lately, several authors have been using inverse models to estimate the amount of heat flux delivered during a diverse range of welding processes. For example, Gonçalves et al. [17] applied the Golden Section Inverse Technique to estimate the heat input in a TIG welding process. Yang et al. [18] used the conjugate gradient method and the discrepancy principle to minimize the heat flux generation in a rotary friction welding. The aforementioned work demonstrates the versatility of the use of inverse problems in the estimation of the welding heat flux. A common process extensively used to weld aluminum alloys is GTAW (gas tungsten arc welding). Recently, several authors have been conducting studies on the analysis of the microstructure of this welding process. Kumar et al. [19] presented a mathematical model to predict the grain size in the fusion zone for an AA6061 GTAW process. Zervaki et al. [20] analyzed the HAZ during aluminum welding using an inverse and direct modeling. The authors’ model satisfactorily predicted the hardness; this result demonstrated the feasibility of the inverse analysis as a way of control and quantitative prediction of the welding properties. In previous work, Magalhães et al. [21] developed a C++ code based on the non-linear three-dimensional heat diffusion equation with phase change and moving heat source to study a GTA welding process in AA6065 T5 samples. The inverse method Broydon–Fletcher–Goldfarb–Shanno (BFGS) was used to minimize the heat input. The software was validated by accomplishing lab experiments. This work has a focus on the heat diffusion in the welding of aluminum. In the present work, the code developed by Magalhães et al. [21] was slightly modified to determine the cooling rate at the FZ, HAZ and BM. The proposed analyses are an inexpensive alternative to precisely determine the instantaneous heating and cooling rate without the requirement of room-controlled temperature, calorimeters, or commercial software. The precise determination of those parameters is decisive to improve the fusion and welding quality. In summary, the study presents an alternative approach to inves-
tigate the cooling rate influence of GTA welding on the microstructure of AA6065 T5 samples. Furthermore, the correlations between the heat input on the weld, the time that the electrode remains on the positive polarity, t+, and the grain size of the HAZ were also investigated systematically. 2. Materials, experiments and simulations 2.1. Materials and methods The 6060-T5 250-mm long, 38-mm wide and 6.5-mm thick aluminum plates (Mn 0.15%; Fe 0.20%; Mg 0.45%; Si 0.40%; Cu 0.02%; Zn 0.05%; Ti 0.01%) were welded through a GTA process at LAPROSOLDA (Federal University of Uberlândia – Brazil). The experimental temperatures were obtained on accessible points of the sample. An automated system was used with the weld velocity at 250 mm/min. The time duration that the electrode remains in the positive polarity on the sample, t+, was tested in four experimental conditions, 2 ms, 7 ms, 11 ms and 13 ms. On each one, the three experiments were performed to ensure the repeatability of the results. The welding experiment procedure and the experimental apparatus were described in Magalhães et al. [21]. After the welding, the samples were cut with an abrasive saw, and the remaining deformed material was later removed by wet grinding and polishing. The samples were grinded using a Struers DPA motor-driven belt grinder in successive steps using silicon carbide abrasive papers of 400, 600, and 1200 grit. The mechanical polishing was accomplished in two steps: first, a suspension of 600-grit alumina (Al2O3) powder in distilled water was used in the Struers DPA motor as a rough polishing. The final polishing was performed using a 0.04-μm suspension of silicon dioxide (SiO2) in distilled water. A concentrated solution of Poultron’s reagent modified (50 mL of Poultron’s reagent (30 mL of HCl; 15 mL of HNO3; 2.5 mL of HF; 2.5 mL of H2O), 25 mL of HNO3 and 40 mL of a solution of 3 g of chromic acid per 10 mL of H2O) was prepared as an etchant in the microscopic examination [22]. The microscopic analysis was performed using an Olympus BM41M-LED optical microscopic and a ZEISS EVOMA15 scanning electron microscope (SEM). The aforementioned preparations and analyses were conducted at the Metallurgy and Materials Laboratory of the Federal University of Itajubá (UNIFEI). 2.2. Simulations To simulate the problem, the C++ in-house code accounted for a non-uniform mesh with 225.000 volumes. Details of the software’s theoretical development, boundary conditions and thermal properties have been reported in Magalhães et al. [21]. The software considered temperature values from six different points, as shown in Fig. 1, for all four experimental conditions. Points P1 and P2 are positioned in the fusion zone, points P3 and P4 are in the HAZ, and points P5 and P6 are in the base material. As the fusion zone and the HAZ size vary according to the positive polarity, different coordinates of the P points were adopted for each adjusted positive polarity condition. The coordinates of the z point are presented in Table 1. The z-axes orientation is defined as presented in Fig. 1. The heating rate was defined as the positive numerical derivative of the considered points and may be expressed as:
∂T ( x, y, z , t ) TPai +1 − TPai , = ∂t Δt
(1)
where T is the temperature, x, y and z are the Cartesian coordinates of the respective points Pi, i is the point index, and a is the time step.
E. dos Santos Magalhães et al./Applied Thermal Engineering 100 (2016) 333–339
335
Fig. 1. Collected temperature points.
in the fusion zone. The points in the HAZ, P3 and P4, have a lower heating rate in relation to P1 and P2. From the graphical analysis, the cooling rate of the points in the HAZ (P3 and P4), i.e., the negative heating rate presented in Fig. 4, is more stable than the cooling rate of the FZ points (P1 and P2). Because these points are in the HAZ, the heat delivered by the torch arch is not sufficient to reach the melting point. However, these points are heated sufficiently to reach the recrystallization point, i.e., these points have a larger grain size compared to the base metal. Points P5 and P6 are in the base metal part of the sample; consequently, they are not affected by the temperature gradient on the sample. From the heating rate analysis (Fig. 4), these points are heated in the beginning of the process. However, these points do not have a significant cooling rate after time 4 s. These points do not achieve the melting point, 655 °C, or the recrystallization temperature, 524 °C [22]. As the torch continues its movement, the temperatures tend to stabilize in these regions. Consequently, after the torch is turned off, these regions have a stable cooling rate, thereby leading to the same structure of the base material. By analyzing the data from Figs. 3 and 4, the heating rate tends to increase proportionally as the positive polarities of the welding increase. The welding condition t+ = 13 ms reaches the highest temperature; therefore, it has the highest heating rate. To determine the effects displayed by the thermal analyses, the grain size of the HAZ was measured. Table 3 presents the average grain size for each experimental welding condition. A correlation between the grain size and the evolution of the positive polarities on the plate must be highlighted. In fact, the grain size tends to decrease linearly as the adjusted welding condition increases. This fact could be explained by the thermal characteristic of the process. The adjusted condition t+ = 13 ms has the highest estimated power; consequently, it also induces the highest temperature on the plate. As shown in Fig. 4, this condition presents the highest heating and cooling rates. Thus, this rapid cooling results in a larger nucleation region of the material, thereby leading to a smaller grain size. However, the adjusted welding condition t+ = 2 ms has a lower estimated power; consequently, it will induce a lower temperature on the plate. As shown in Fig. 3, this condition has the lowest heating and cooling rates of the analyzed cases. Because the cooling rate is lower, grain growth may be verified. A linear dependence of the grain size in relation to the temperature was found. Fig. 5 presents a linear fit approximation for the data presented in Table 3. The cooling rate was also studied after the welding torch was turned off. In this analysis, the P points did not display a
3. Results and discussion 3.1. Numerical analysis For each t+ condition, the temperature field was determined through the C++ in-house code. The heating rate in the fusion zone must be calculated by the software. Fig. 2 presents the temperature fields in the analyzed section for the experimental condition t+ = 11 ms at the instant of 6.6 s. Fig. 2 also shows that the highest temperatures on the plate are achieved on the surface close to the GTA arch. To validate the program, the width and penetration of the weld bead were measured on the plates and numerically calculated. Table 2 presents the average of the obtained data of these parameters for the four positive polarities on the plate tested. The experimental width and penetration depth of the weld bead were obtained by microscopic analyses. In order to obtain these values numerically, the boundary line of the FZ was defined as the region that presents temperature equal to the melting point of the aluminum (655 °C [22]). The numerical values for those parameters were found by evaluating the difference between the extremity points of the coordinates in the boundary line of the FZ. Note that the numerical results present a low deviation compared to the experimental analysis. The difference in all cases is less than 4%, which could be considered a good result. As previously mentioned, six numerical sensors were coupled in the software. The heating rate calculated from the numerical temperatures is much more precise than the constant values adopted from handbooks. Fig. 3 presents the numerical temperature for the six sensors on four t+ experimental conditions. As mentioned by Magalhães et al. [21], the average temperature on the plate increases when the adjusted positive polarities on the plate become higher. This phenomenon can be observed in Fig. 3. The peak temperature, in this case, represented by the point closer to the plate surface P1, starts at approximately 1400 °C at t+ = 2 ms and reaches 2240 °C at t+ = 13 ms. As the temperature rises, a higher microstructural change occurs. The main effects responsible for this shift are the heating and cooling rates. Fig. 4 presents the heating rate for the four positive polarities on the sample tested. In all of the presented cases, point P1 is subject to the highest temperature gradient. Indeed, this point is closer to the surface and is most strongly influenced by the torch. Although point P2 is also in the fusion zone, it has a lower heating rate compared with P1. Naturally, the temperature in P2 tends to be less influenced by the voltaic arch, whereas the heat flux is distributed
Table 1 Coordinate z of the P points for each adjusted positive polarity. Adjusted t+ (ms)
P1 (mm)
P2 (mm)
P3 (mm)
P4 (mm)
P5 (mm)
P6 (mm)
2 7 11 13
0.30 0.30 0.40 0.40
1.3 1.4 1.6 1.7
1.4 1.5 1.8 1.9
1.5 1.6 1.9 2.0
2.0 2.0 2.5 2.5
2.5 2.5 3.0 3.0
336
E. dos Santos Magalhães et al./Applied Thermal Engineering 100 (2016) 333–339
significant difference compared to the same welding condition, i.e., all P points exhibited almost the same cooling rate. Consequently, the microstructural change for this welding experiment occurred when the GTA torch was turned on. The cooling rate also presented a similar profile for the different t+ welding conditions. Fig. 6 presents the cooling rate aforementioned as the negative heating rate for the different t+ welding conditions in relation to time. 3.2. Microstructure analysis
Fig. 2. Temperature [°C] distribution for t+ = 11 ms at instant 6.6 s.
Table 2 Numerical and experimental values of the width and penetration of the weld bead. Adjusted (t+) (ms)
Experimental penetration (mm)
Numerical penetration (mm)
Experimental width (mm)
Numerical width (mm)
2 7 11 13
1.43 1.61 1.72 1.73
1.38 1.58 1.70 1.71
6.22 6.66 6.76 6.77
6.12 6.59 6.70 6.74
As predicted by the software analysis, the HAZ size tends to increase as the positive polarity increases. This phenomenon is verified in Fig. 7, where the micrographs for one sample of each analyzed welding condition are presented. The microstructure found in the fusion zone (FZ) and changed to base material (BM) depends on the high temperatures reached at each welding condition. The microstructure found in all cases was largely of dendritic grains with a directional solidification toward the center of the fusion zone. Kou [23] described the effect of a supercooling solidification process mode as planar to cellular, cellular to columnar dendrite, and columnar dendrite to equiaxed dendrite. As the HAZ is directly related to the thermal gradient, it is expected that the welded samples at t+ = 13 ms present a larger HAZ than the samples welded at t+ = 2 ms. As previously mentioned, due to the increase of the thermal gradient, the grain size tends to decrease. This decrease is attributed to the higher presence of intermetallic compound precipitates when using high heat input.
Fig. 3. Numerical temperatures (°C) for t+ experimental conditions: a) 2 ms, b) 7 ms, c) 11 ms, and d) 13 ms.
E. dos Santos Magalhães et al./Applied Thermal Engineering 100 (2016) 333–339
337
Fig. 4. Numerical heating rate for t+ experimental conditions: a) 2 ms, b) 7 ms, c) 11 ms, and d) 13 ms.
The in-house software may also be applied to determine the partially melted zone (PMZ) size. As presented by Magalhães et al. [21], the code was written based on the enthalpy method. This mathematical approach considers the mush zone, i.e., the region located at the interface of liquid and solid phases. Due to the mesh settings, the C++ code only indicates the formation of a PMZ on the positive polarities t+ = 11 ms and t+ = 13 ms. During welding, this region is identified on the boundary of the columnar dendrite and the HAZ. The grain size of this phase is typically larger than that of the HAZ. Because this zone was heated at a temperature slightly above the solidus temperature, the grain of this region tends to remain heated for a longer period than the grain of the HAZ. This effect may be clearly identified in the hottest conditions in welding; in this case, t+ = 11 ms and t+ = 13 ms. As identified by the code, SEM analysis indicated that for the welding conditions t+ = 2 ms and t+ = 7 ms, the grain size of the PMZ is similar to the grain size of the HAZ; thus, it could be considered as part of the HAZ. Using an
appropriated resolution on the SEM imaging, Fig. 8a presents an internal view of the PMZ for the welding condition t+ = 13 ms. Fig. 8b presents the equiaxed dendrites found in the center of the FZ for the welding condition t+ = 13 ms. Similar structures were found in the remaining welding conditions. Fig. 8c presents the columnar dendrites found on the interface of the FZ and HAZ. As precipitation
Table 3 Average grain size for each welding condition. Adjusted (t+) (ms)
Adjusted (t-) (ms)
Estimated power (W)
Grain size (μm)
2 7 11 13
20 20 20 20
2037 2187 2335 2702
80.01 57.39 46.96 40.00
Fig. 5. Grain size versus positive polarity condition.
338
E. dos Santos Magalhães et al./Applied Thermal Engineering 100 (2016) 333–339
Si may be identified as the black precipitates present at the grain boundary shown in Fig. 8d.
4. Conclusions
Fig. 6. Numerical cooling rate for the tested welding conditions after the welding torch is turned off.
of alloys usually occurs along grain boundaries, traces of Mg and Si were found on the grain boundary of the HAZ. It was found that the Mg:Si ratio in the HAZ tends to increase as the welding polarity increases. Further investigation is warranted to quantify the ratio of Mg:Si according to the time of positive polarity. Fig. 8d presents a view of the microstructure of the grain in the HAZ. Mg and
This paper presented a thermal analysis of the HAZ using software based on a numerical heat transfer model. A linear dependence of the grain size in relation to the positive polarity was observed. The cooling rate of the six analyzed temperature points was found to increase as the points become closer to the surface of the sample and the positive polarity intensifies. Microstructural changes occurred when the GTA welding torch was on. The heat transfer loss (cooling rate) after the GTA torch was turned off is the same in the points of the FZ, HAZ and BM. The software was demonstrated to be a good and inexpensive method to analyze the welding in the HAZ. The software could also be adapted to other welding processes. Therefore, this software could improve welding processes in general.
Acknowledgements The authors thank CNPq, CAPES and FAPEMIG for their financial support. The authors acknowledge the Heat Transfer (LTCM) and Welding (LAPROSOLDA) laboratories in Universidade Federal de Uberlândia.
Fig. 7. Microstructural analysis for t+ welding conditions: a) 2 ms, b) 7 ms, c) 11 ms, and d) 13 ms.
E. dos Santos Magalhães et al./Applied Thermal Engineering 100 (2016) 333–339
339
Fig. 8. The microstructural changes for t+ = 13 ms: a) PMZ, b) equiaxed dendrite at the FZ, c) columnar dendrites, and d) precipitation at the grain boundary.
References [1] J.A. Goldak, M. Akhlaghi, Computation Welding Mechanics, Springer, 2005. [2] J.D. Jordatos, G. Fourlaris, G. Papadimitriou, The effect of cooling rate on the mechanical and corrosion properties of SAT 2205 (UNS 31803) duplex stainless steel welds, Scr. Mater. 44 (2001) 401–408. [3] M.A. Winarto, Taufiqullah, cooling rate on cold cracking in welded thich HSLA steel plate, Mater. Sci. Forum 689 (2011) 269–275. [4] S. Kwon, S. Bae, J. Do, J. Chang, H. Hyun, Characterization of the microstructures and the cryogenic mechanical properties of electron beam welded inconel 718, Metal. Mater. Trans. A 47 (2) (2016) 777–787. [5] H.T. Lee, J.L. Wu, The effects of peak temperature and cooling rate on the susceptibility to intergranular corrosion of alloy 690 by laser beam and gas tungsten arc welding, Corros. Sci. 51 (2009) 439–445. [6] K. Sivaprasad, S.G.S. Raman, Influence of welding cooling rate on microstructure and mechanical properties of alloy 718 weldments, Metal. Mater. Trans. A 39A (2008) 2115–2127. [7] S.G.K. Manikandan, D. Sivakumar, K. Prasad Rao, M. Kamaraj, Effect of weld cooling rate on Laves phase formation in Inconel 718 fusion zone, J. Mater. Process. Technol. 214 (2014) 258–264. [8] M. Aissani, S. Guessasma, A. Zitouni, R. Hamzaoui, D. Bassir, Y. Benkedda, Three-dimensional simulation of 304L steel TIG welding process: contribution of the thermal flux, Appl. Therm. Eng. 89 (2015) 822–832. [9] X. Di, S. Ji, F. Cheng, D. Wang, J. Cao, Effect of cooling rate on microstructure, inclusions and mechanical properties of weld metal in simulated local dry underwater welding, Mater. Des. 88 (2015) 505–513. [10] S. Zheng, Q. Wu, Q. Huang, S. Liu, Y. Han, Influence of different cooling rates on the microstructure of the HAZ and welding CCT diagram of CLAM steel, Fusion Eng. Des. 86 (2011) 2616–2619. [11] T. Bhattacharya, A. Bandyoupadhyay, P.K. Pal, An investigation on temperature distribution and cooling rate of ERW pipes during TIG welding, J. Manuf. Sci. Prod 14 (2014) 219–231.
[12] J. Dagner, J. Friedrich, G. Müller, Numerical study on the prediction of microstructure parameters by multi-scale modeling of directional solidification of binary aluminum – silicon alloys, Comput. Mater. Sci. 43 (2008) 872–885. [13] J.S. Unfried, C.M. Garzón, J.E. Giraldo, Numerical and experimental analysis of microstructure evolution during arc welding in armor plate steels, J. Mater. Process. Technol. 209 (2009) 1688–1700. [14] V. Manvatkar, A. De, L.E. Svensson, T. DebRoy, Cooling rate and peak temperatures during friction stir welding of a high-carbon steel, Scr. Mater. 94 (2015) 36–39. [15] H. Schmidt, J. Hattel, J. Wert, An analytical model for the heat generation in friction stir welding, Model. Simul. Mater. Sci. Eng. 12 (2004) 143–157. [16] G.L. Datta, A.K. Pathak, Estimation of temperature distribution and cooling rate in arc welding using three-dimensional finite element analysis, Mater. Sci. Forum 426–432 (2003) 4099–4104. [17] C.V. Gonçalves, S.R. Carvalho, G. Guimaraes, Application of optimization techniques and the enthalpy method to solve a 3D-inverse problem during a TIG welding process, Mater. Des. 30 (2010) 2398–2402. [18] Y-C. Yang, W-L. Chen, H-L. Lee, A nonlinear inverse problem in estimating the heat generation in rotary friction welding, Num. Heat Transf. A 59 (2011) 130–149. [19] T.S. Kumar, V. Balasubramanian, S. Babu, M.Y. Sanavullah, Effect of pulsed current GTA welding parameters on the fusion zone microstructure of AA6061 aluminum alloy, Met. Mater. Int. 13 (2007) 345–351. [20] A.D. Zervaki, G.N. Haidemenopoulos, S.G. Lambrakos, Analysis of heat affected zone in welded aluminum alloys using inverse and direct modeling, J. Mater. Eng. Perform. 17 (2008) 402–410. [21] E.S. Magalhães, S.R. Carvalho, A.L.F. Lima e Silva, S.M.M. Lima e Silva, The use of non-linear inverse problem and enthalpy method in GTAW process of aluminum, Int. Comm. Heat Mass Transfer 66 (2015) 114–121. [22] R.H. Stevens, Aluminum Alloys: Metallographic Techniques and Microstructures. Metallography and Microstructures, vol. 9, ASM Handbook. ASM International, 2004. [23] S. Kou, Welding Metallurgy, J. Wiley & Sons, Hoboken, NJ, 2003.
Contents lists available at ScienceDirect
Applied Thermal Engineering j o u r n a l h o m e p a g e : w w w. e l s e v i e r. c o m / l o c a t e / a p t h e r m e n g
Research Paper
Microstructural analysis in GTA aluminum alloy welding using inverse problems Elisan dos Santos Magalhães, Edmilson Otoni Correa, Ana Lúcia Fernandes Lima E. Silva, Sandro Metrevelle Marcondes Lima E. Silva * Heat Transfer Laboratory – LabTC, Institute of Mechanical Engineering – IEM, Federal University of Itajubá – UNIFEI, Campus Prof. José Rodrigues Seabra, Av. BPS, 1303, 37500-903 Itajubá, MG, Brazil
H I G H L I G H T S
• • • • •
A numerical study of heating and cooling rate in a GTA aluminum welding process. A linear dependence of the grain size in relation to the positive polarity was observed. Correlations between the cooling rate and the grain size of heat affected zone (HAZ). A thermal analysis of the HAZ using software based on a numerical heat transfer model. The microstructural changes happen while the GTA arch torch was turned on.
A R T I C L E
I N F O
Article history: Received 21 October 2015 Accepted 11 February 2016 Available online 24 February 2016 Keywords: Inverse problems Heat flux Temperature estimation GTAW process Microstructural analysis
A B S T R A C T
Numerical software based on the tridimensional diffusion equation with a moving source, a phase change and heat flux estimation by the non-linear interactive Broydon–Fletcher–Goldfarb–Shanno (BFGS) inverse technique was used to study the heat affected zone (HAZ) in GTA aluminum 6065 T5 alloy welding. GTA welding experiments were performed using thin aluminum plates in four t+ experimental conditions. In previous studies, the authors determined that the peak temperature tends to increase as the positive polarity becomes higher. To confirm this behavior, the samples were cut on the welded region and later characterized using an optical microscope (OM) and a scanning electronic microscope (SEM). The heating and cooling rates, determined from an in-house code, were compared with the microstructures and the grain size of the HAZ found in the samples. The results revealed a linear correlation between the grain size on the HAZ and the positive polarity. The study also showed that a significant change in the microstructure occurs during the process while the GTAW torch is still turned on. After the welding, when the GTAW torch was turned off, the cooling rate was the same for all welded zones, which indicates that the microstructural changes had already occurred. © 2016 Elsevier Ltd. All rights reserved.
1. Introduction Although several new processes to join metals have been developed, the fusion welding process still plays a major role in engineering practice. Hence, a deep understanding of the behavior of microstructural welding is highly important for minimizing potential failures. The heat affected zone (HAZ) is a region where the microstructure is negatively affected during the welding process. The HAZ is fundamentally caused by the characteristic overheating and supercooling of the welding process. The thermal cycle that results in the recrystallization during welding is primarily composed of two steps: heating and cooling [1]. This cycle triggers a
* Corresponding author. Tel.: +55 35 3629 1362; fax: +55 35 3629 1148. E-mail address: [email protected] (S.M.M. Lima e Silva). http://dx.doi.org/10.1016/j.applthermaleng.2016.02.051 1359-4311/© 2016 Elsevier Ltd. All rights reserved.
microstructural change in the base material (BM), thereby creating the HAZ. Although the heating rate is important to achieve the fusion, the cooling rate is extremely important for the welding quality. During welding, a high cooling rate may generally be associated with the reduction of mechanical properties [2], cracking [3], deleterious phase precipitation [4], or susceptibility to intergranular corrosion [5]. Recently, an increasing interest in the impact of cooling rate on welding properties may be pointed out. For instance, Sivaprasad and Raman [6] analyzed the influence of cooling rate on microstructure and mechanical properties of alloy 718 when it is welded through two distinct welding processes, GTA (gas tungsten arc) and EB (electron beam). Manikandan et al. [7] studied the Laves phase formation in Inconel 718 in the cooling of a welding process. Aissani et al. [8] used a three-dimensional finite element model to analyze the evolution of the microstructure in the weld HAZ in a TIG
334
E. dos Santos Magalhães et al./Applied Thermal Engineering 100 (2016) 333–339
(tungsten arc gas) process. Di et al. [9] used experimental measurements to determine the average cooling rate in local dry underwater welding. The authors’ analyses suggest that a fast cooling rate, in this process, may improve the weld properties. A precise determination of the cooling rate is difficult to achieve. Several authors merely assume that the cooling rate is an intermediate value between the peak temperature and the room temperature divided by the time of cooling [10], while others obtain the value using mathematical models based on the one-dimensional heat diffusion equation, for example, in Bhattacharya et al. [11]. Another option is the use of numerical thermal models for the prediction of the microstructure parameters [12]. However, major simplifications are constantly being made to reduce the complexity of this type of model, e.g. Unfried et al. [13], who assumed a onedimensional heat flux and a one-dimensional thermal model based on Rosenthal’s equations. Another example is the work developed by Manvatkar et al. [14]. To estimate the peak temperature and cooling rate, the authors used an analytical expression proposed by Schmidt et al. [15]. The use of analytical expressions for the estimation is convenient; however, these models do not cover all experimental errors, thereby raising inaccuracies in the solution. The use of commercial software is another alternative to determine the cooling rate [16]. In contrast to the use of an analytical expression, the commercial software acts as a black box, i.e., the user does not have access to its methodology. In fact, due to certain experimental singularities, such as the non-uniformity of the torch arc and the electrical interference on measurements in aluminum welding, a three-dimensional numerical thermal model that covers all boundary conditions is highly complex. In addition, directly measuring the heat flux during the process is expensive. However, the use of inverse problems enables a quick determination of this parameter [17]. Lately, several authors have been using inverse models to estimate the amount of heat flux delivered during a diverse range of welding processes. For example, Gonçalves et al. [17] applied the Golden Section Inverse Technique to estimate the heat input in a TIG welding process. Yang et al. [18] used the conjugate gradient method and the discrepancy principle to minimize the heat flux generation in a rotary friction welding. The aforementioned work demonstrates the versatility of the use of inverse problems in the estimation of the welding heat flux. A common process extensively used to weld aluminum alloys is GTAW (gas tungsten arc welding). Recently, several authors have been conducting studies on the analysis of the microstructure of this welding process. Kumar et al. [19] presented a mathematical model to predict the grain size in the fusion zone for an AA6061 GTAW process. Zervaki et al. [20] analyzed the HAZ during aluminum welding using an inverse and direct modeling. The authors’ model satisfactorily predicted the hardness; this result demonstrated the feasibility of the inverse analysis as a way of control and quantitative prediction of the welding properties. In previous work, Magalhães et al. [21] developed a C++ code based on the non-linear three-dimensional heat diffusion equation with phase change and moving heat source to study a GTA welding process in AA6065 T5 samples. The inverse method Broydon–Fletcher–Goldfarb–Shanno (BFGS) was used to minimize the heat input. The software was validated by accomplishing lab experiments. This work has a focus on the heat diffusion in the welding of aluminum. In the present work, the code developed by Magalhães et al. [21] was slightly modified to determine the cooling rate at the FZ, HAZ and BM. The proposed analyses are an inexpensive alternative to precisely determine the instantaneous heating and cooling rate without the requirement of room-controlled temperature, calorimeters, or commercial software. The precise determination of those parameters is decisive to improve the fusion and welding quality. In summary, the study presents an alternative approach to inves-
tigate the cooling rate influence of GTA welding on the microstructure of AA6065 T5 samples. Furthermore, the correlations between the heat input on the weld, the time that the electrode remains on the positive polarity, t+, and the grain size of the HAZ were also investigated systematically. 2. Materials, experiments and simulations 2.1. Materials and methods The 6060-T5 250-mm long, 38-mm wide and 6.5-mm thick aluminum plates (Mn 0.15%; Fe 0.20%; Mg 0.45%; Si 0.40%; Cu 0.02%; Zn 0.05%; Ti 0.01%) were welded through a GTA process at LAPROSOLDA (Federal University of Uberlândia – Brazil). The experimental temperatures were obtained on accessible points of the sample. An automated system was used with the weld velocity at 250 mm/min. The time duration that the electrode remains in the positive polarity on the sample, t+, was tested in four experimental conditions, 2 ms, 7 ms, 11 ms and 13 ms. On each one, the three experiments were performed to ensure the repeatability of the results. The welding experiment procedure and the experimental apparatus were described in Magalhães et al. [21]. After the welding, the samples were cut with an abrasive saw, and the remaining deformed material was later removed by wet grinding and polishing. The samples were grinded using a Struers DPA motor-driven belt grinder in successive steps using silicon carbide abrasive papers of 400, 600, and 1200 grit. The mechanical polishing was accomplished in two steps: first, a suspension of 600-grit alumina (Al2O3) powder in distilled water was used in the Struers DPA motor as a rough polishing. The final polishing was performed using a 0.04-μm suspension of silicon dioxide (SiO2) in distilled water. A concentrated solution of Poultron’s reagent modified (50 mL of Poultron’s reagent (30 mL of HCl; 15 mL of HNO3; 2.5 mL of HF; 2.5 mL of H2O), 25 mL of HNO3 and 40 mL of a solution of 3 g of chromic acid per 10 mL of H2O) was prepared as an etchant in the microscopic examination [22]. The microscopic analysis was performed using an Olympus BM41M-LED optical microscopic and a ZEISS EVOMA15 scanning electron microscope (SEM). The aforementioned preparations and analyses were conducted at the Metallurgy and Materials Laboratory of the Federal University of Itajubá (UNIFEI). 2.2. Simulations To simulate the problem, the C++ in-house code accounted for a non-uniform mesh with 225.000 volumes. Details of the software’s theoretical development, boundary conditions and thermal properties have been reported in Magalhães et al. [21]. The software considered temperature values from six different points, as shown in Fig. 1, for all four experimental conditions. Points P1 and P2 are positioned in the fusion zone, points P3 and P4 are in the HAZ, and points P5 and P6 are in the base material. As the fusion zone and the HAZ size vary according to the positive polarity, different coordinates of the P points were adopted for each adjusted positive polarity condition. The coordinates of the z point are presented in Table 1. The z-axes orientation is defined as presented in Fig. 1. The heating rate was defined as the positive numerical derivative of the considered points and may be expressed as:
∂T ( x, y, z , t ) TPai +1 − TPai , = ∂t Δt
(1)
where T is the temperature, x, y and z are the Cartesian coordinates of the respective points Pi, i is the point index, and a is the time step.
E. dos Santos Magalhães et al./Applied Thermal Engineering 100 (2016) 333–339
335
Fig. 1. Collected temperature points.
in the fusion zone. The points in the HAZ, P3 and P4, have a lower heating rate in relation to P1 and P2. From the graphical analysis, the cooling rate of the points in the HAZ (P3 and P4), i.e., the negative heating rate presented in Fig. 4, is more stable than the cooling rate of the FZ points (P1 and P2). Because these points are in the HAZ, the heat delivered by the torch arch is not sufficient to reach the melting point. However, these points are heated sufficiently to reach the recrystallization point, i.e., these points have a larger grain size compared to the base metal. Points P5 and P6 are in the base metal part of the sample; consequently, they are not affected by the temperature gradient on the sample. From the heating rate analysis (Fig. 4), these points are heated in the beginning of the process. However, these points do not have a significant cooling rate after time 4 s. These points do not achieve the melting point, 655 °C, or the recrystallization temperature, 524 °C [22]. As the torch continues its movement, the temperatures tend to stabilize in these regions. Consequently, after the torch is turned off, these regions have a stable cooling rate, thereby leading to the same structure of the base material. By analyzing the data from Figs. 3 and 4, the heating rate tends to increase proportionally as the positive polarities of the welding increase. The welding condition t+ = 13 ms reaches the highest temperature; therefore, it has the highest heating rate. To determine the effects displayed by the thermal analyses, the grain size of the HAZ was measured. Table 3 presents the average grain size for each experimental welding condition. A correlation between the grain size and the evolution of the positive polarities on the plate must be highlighted. In fact, the grain size tends to decrease linearly as the adjusted welding condition increases. This fact could be explained by the thermal characteristic of the process. The adjusted condition t+ = 13 ms has the highest estimated power; consequently, it also induces the highest temperature on the plate. As shown in Fig. 4, this condition presents the highest heating and cooling rates. Thus, this rapid cooling results in a larger nucleation region of the material, thereby leading to a smaller grain size. However, the adjusted welding condition t+ = 2 ms has a lower estimated power; consequently, it will induce a lower temperature on the plate. As shown in Fig. 3, this condition has the lowest heating and cooling rates of the analyzed cases. Because the cooling rate is lower, grain growth may be verified. A linear dependence of the grain size in relation to the temperature was found. Fig. 5 presents a linear fit approximation for the data presented in Table 3. The cooling rate was also studied after the welding torch was turned off. In this analysis, the P points did not display a
3. Results and discussion 3.1. Numerical analysis For each t+ condition, the temperature field was determined through the C++ in-house code. The heating rate in the fusion zone must be calculated by the software. Fig. 2 presents the temperature fields in the analyzed section for the experimental condition t+ = 11 ms at the instant of 6.6 s. Fig. 2 also shows that the highest temperatures on the plate are achieved on the surface close to the GTA arch. To validate the program, the width and penetration of the weld bead were measured on the plates and numerically calculated. Table 2 presents the average of the obtained data of these parameters for the four positive polarities on the plate tested. The experimental width and penetration depth of the weld bead were obtained by microscopic analyses. In order to obtain these values numerically, the boundary line of the FZ was defined as the region that presents temperature equal to the melting point of the aluminum (655 °C [22]). The numerical values for those parameters were found by evaluating the difference between the extremity points of the coordinates in the boundary line of the FZ. Note that the numerical results present a low deviation compared to the experimental analysis. The difference in all cases is less than 4%, which could be considered a good result. As previously mentioned, six numerical sensors were coupled in the software. The heating rate calculated from the numerical temperatures is much more precise than the constant values adopted from handbooks. Fig. 3 presents the numerical temperature for the six sensors on four t+ experimental conditions. As mentioned by Magalhães et al. [21], the average temperature on the plate increases when the adjusted positive polarities on the plate become higher. This phenomenon can be observed in Fig. 3. The peak temperature, in this case, represented by the point closer to the plate surface P1, starts at approximately 1400 °C at t+ = 2 ms and reaches 2240 °C at t+ = 13 ms. As the temperature rises, a higher microstructural change occurs. The main effects responsible for this shift are the heating and cooling rates. Fig. 4 presents the heating rate for the four positive polarities on the sample tested. In all of the presented cases, point P1 is subject to the highest temperature gradient. Indeed, this point is closer to the surface and is most strongly influenced by the torch. Although point P2 is also in the fusion zone, it has a lower heating rate compared with P1. Naturally, the temperature in P2 tends to be less influenced by the voltaic arch, whereas the heat flux is distributed
Table 1 Coordinate z of the P points for each adjusted positive polarity. Adjusted t+ (ms)
P1 (mm)
P2 (mm)
P3 (mm)
P4 (mm)
P5 (mm)
P6 (mm)
2 7 11 13
0.30 0.30 0.40 0.40
1.3 1.4 1.6 1.7
1.4 1.5 1.8 1.9
1.5 1.6 1.9 2.0
2.0 2.0 2.5 2.5
2.5 2.5 3.0 3.0
336
E. dos Santos Magalhães et al./Applied Thermal Engineering 100 (2016) 333–339
significant difference compared to the same welding condition, i.e., all P points exhibited almost the same cooling rate. Consequently, the microstructural change for this welding experiment occurred when the GTA torch was turned on. The cooling rate also presented a similar profile for the different t+ welding conditions. Fig. 6 presents the cooling rate aforementioned as the negative heating rate for the different t+ welding conditions in relation to time. 3.2. Microstructure analysis
Fig. 2. Temperature [°C] distribution for t+ = 11 ms at instant 6.6 s.
Table 2 Numerical and experimental values of the width and penetration of the weld bead. Adjusted (t+) (ms)
Experimental penetration (mm)
Numerical penetration (mm)
Experimental width (mm)
Numerical width (mm)
2 7 11 13
1.43 1.61 1.72 1.73
1.38 1.58 1.70 1.71
6.22 6.66 6.76 6.77
6.12 6.59 6.70 6.74
As predicted by the software analysis, the HAZ size tends to increase as the positive polarity increases. This phenomenon is verified in Fig. 7, where the micrographs for one sample of each analyzed welding condition are presented. The microstructure found in the fusion zone (FZ) and changed to base material (BM) depends on the high temperatures reached at each welding condition. The microstructure found in all cases was largely of dendritic grains with a directional solidification toward the center of the fusion zone. Kou [23] described the effect of a supercooling solidification process mode as planar to cellular, cellular to columnar dendrite, and columnar dendrite to equiaxed dendrite. As the HAZ is directly related to the thermal gradient, it is expected that the welded samples at t+ = 13 ms present a larger HAZ than the samples welded at t+ = 2 ms. As previously mentioned, due to the increase of the thermal gradient, the grain size tends to decrease. This decrease is attributed to the higher presence of intermetallic compound precipitates when using high heat input.
Fig. 3. Numerical temperatures (°C) for t+ experimental conditions: a) 2 ms, b) 7 ms, c) 11 ms, and d) 13 ms.
E. dos Santos Magalhães et al./Applied Thermal Engineering 100 (2016) 333–339
337
Fig. 4. Numerical heating rate for t+ experimental conditions: a) 2 ms, b) 7 ms, c) 11 ms, and d) 13 ms.
The in-house software may also be applied to determine the partially melted zone (PMZ) size. As presented by Magalhães et al. [21], the code was written based on the enthalpy method. This mathematical approach considers the mush zone, i.e., the region located at the interface of liquid and solid phases. Due to the mesh settings, the C++ code only indicates the formation of a PMZ on the positive polarities t+ = 11 ms and t+ = 13 ms. During welding, this region is identified on the boundary of the columnar dendrite and the HAZ. The grain size of this phase is typically larger than that of the HAZ. Because this zone was heated at a temperature slightly above the solidus temperature, the grain of this region tends to remain heated for a longer period than the grain of the HAZ. This effect may be clearly identified in the hottest conditions in welding; in this case, t+ = 11 ms and t+ = 13 ms. As identified by the code, SEM analysis indicated that for the welding conditions t+ = 2 ms and t+ = 7 ms, the grain size of the PMZ is similar to the grain size of the HAZ; thus, it could be considered as part of the HAZ. Using an
appropriated resolution on the SEM imaging, Fig. 8a presents an internal view of the PMZ for the welding condition t+ = 13 ms. Fig. 8b presents the equiaxed dendrites found in the center of the FZ for the welding condition t+ = 13 ms. Similar structures were found in the remaining welding conditions. Fig. 8c presents the columnar dendrites found on the interface of the FZ and HAZ. As precipitation
Table 3 Average grain size for each welding condition. Adjusted (t+) (ms)
Adjusted (t-) (ms)
Estimated power (W)
Grain size (μm)
2 7 11 13
20 20 20 20
2037 2187 2335 2702
80.01 57.39 46.96 40.00
Fig. 5. Grain size versus positive polarity condition.
338
E. dos Santos Magalhães et al./Applied Thermal Engineering 100 (2016) 333–339
Si may be identified as the black precipitates present at the grain boundary shown in Fig. 8d.
4. Conclusions
Fig. 6. Numerical cooling rate for the tested welding conditions after the welding torch is turned off.
of alloys usually occurs along grain boundaries, traces of Mg and Si were found on the grain boundary of the HAZ. It was found that the Mg:Si ratio in the HAZ tends to increase as the welding polarity increases. Further investigation is warranted to quantify the ratio of Mg:Si according to the time of positive polarity. Fig. 8d presents a view of the microstructure of the grain in the HAZ. Mg and
This paper presented a thermal analysis of the HAZ using software based on a numerical heat transfer model. A linear dependence of the grain size in relation to the positive polarity was observed. The cooling rate of the six analyzed temperature points was found to increase as the points become closer to the surface of the sample and the positive polarity intensifies. Microstructural changes occurred when the GTA welding torch was on. The heat transfer loss (cooling rate) after the GTA torch was turned off is the same in the points of the FZ, HAZ and BM. The software was demonstrated to be a good and inexpensive method to analyze the welding in the HAZ. The software could also be adapted to other welding processes. Therefore, this software could improve welding processes in general.
Acknowledgements The authors thank CNPq, CAPES and FAPEMIG for their financial support. The authors acknowledge the Heat Transfer (LTCM) and Welding (LAPROSOLDA) laboratories in Universidade Federal de Uberlândia.
Fig. 7. Microstructural analysis for t+ welding conditions: a) 2 ms, b) 7 ms, c) 11 ms, and d) 13 ms.
E. dos Santos Magalhães et al./Applied Thermal Engineering 100 (2016) 333–339
339
Fig. 8. The microstructural changes for t+ = 13 ms: a) PMZ, b) equiaxed dendrite at the FZ, c) columnar dendrites, and d) precipitation at the grain boundary.
References [1] J.A. Goldak, M. Akhlaghi, Computation Welding Mechanics, Springer, 2005. [2] J.D. Jordatos, G. Fourlaris, G. Papadimitriou, The effect of cooling rate on the mechanical and corrosion properties of SAT 2205 (UNS 31803) duplex stainless steel welds, Scr. Mater. 44 (2001) 401–408. [3] M.A. Winarto, Taufiqullah, cooling rate on cold cracking in welded thich HSLA steel plate, Mater. Sci. Forum 689 (2011) 269–275. [4] S. Kwon, S. Bae, J. Do, J. Chang, H. Hyun, Characterization of the microstructures and the cryogenic mechanical properties of electron beam welded inconel 718, Metal. Mater. Trans. A 47 (2) (2016) 777–787. [5] H.T. Lee, J.L. Wu, The effects of peak temperature and cooling rate on the susceptibility to intergranular corrosion of alloy 690 by laser beam and gas tungsten arc welding, Corros. Sci. 51 (2009) 439–445. [6] K. Sivaprasad, S.G.S. Raman, Influence of welding cooling rate on microstructure and mechanical properties of alloy 718 weldments, Metal. Mater. Trans. A 39A (2008) 2115–2127. [7] S.G.K. Manikandan, D. Sivakumar, K. Prasad Rao, M. Kamaraj, Effect of weld cooling rate on Laves phase formation in Inconel 718 fusion zone, J. Mater. Process. Technol. 214 (2014) 258–264. [8] M. Aissani, S. Guessasma, A. Zitouni, R. Hamzaoui, D. Bassir, Y. Benkedda, Three-dimensional simulation of 304L steel TIG welding process: contribution of the thermal flux, Appl. Therm. Eng. 89 (2015) 822–832. [9] X. Di, S. Ji, F. Cheng, D. Wang, J. Cao, Effect of cooling rate on microstructure, inclusions and mechanical properties of weld metal in simulated local dry underwater welding, Mater. Des. 88 (2015) 505–513. [10] S. Zheng, Q. Wu, Q. Huang, S. Liu, Y. Han, Influence of different cooling rates on the microstructure of the HAZ and welding CCT diagram of CLAM steel, Fusion Eng. Des. 86 (2011) 2616–2619. [11] T. Bhattacharya, A. Bandyoupadhyay, P.K. Pal, An investigation on temperature distribution and cooling rate of ERW pipes during TIG welding, J. Manuf. Sci. Prod 14 (2014) 219–231.
[12] J. Dagner, J. Friedrich, G. Müller, Numerical study on the prediction of microstructure parameters by multi-scale modeling of directional solidification of binary aluminum – silicon alloys, Comput. Mater. Sci. 43 (2008) 872–885. [13] J.S. Unfried, C.M. Garzón, J.E. Giraldo, Numerical and experimental analysis of microstructure evolution during arc welding in armor plate steels, J. Mater. Process. Technol. 209 (2009) 1688–1700. [14] V. Manvatkar, A. De, L.E. Svensson, T. DebRoy, Cooling rate and peak temperatures during friction stir welding of a high-carbon steel, Scr. Mater. 94 (2015) 36–39. [15] H. Schmidt, J. Hattel, J. Wert, An analytical model for the heat generation in friction stir welding, Model. Simul. Mater. Sci. Eng. 12 (2004) 143–157. [16] G.L. Datta, A.K. Pathak, Estimation of temperature distribution and cooling rate in arc welding using three-dimensional finite element analysis, Mater. Sci. Forum 426–432 (2003) 4099–4104. [17] C.V. Gonçalves, S.R. Carvalho, G. Guimaraes, Application of optimization techniques and the enthalpy method to solve a 3D-inverse problem during a TIG welding process, Mater. Des. 30 (2010) 2398–2402. [18] Y-C. Yang, W-L. Chen, H-L. Lee, A nonlinear inverse problem in estimating the heat generation in rotary friction welding, Num. Heat Transf. A 59 (2011) 130–149. [19] T.S. Kumar, V. Balasubramanian, S. Babu, M.Y. Sanavullah, Effect of pulsed current GTA welding parameters on the fusion zone microstructure of AA6061 aluminum alloy, Met. Mater. Int. 13 (2007) 345–351. [20] A.D. Zervaki, G.N. Haidemenopoulos, S.G. Lambrakos, Analysis of heat affected zone in welded aluminum alloys using inverse and direct modeling, J. Mater. Eng. Perform. 17 (2008) 402–410. [21] E.S. Magalhães, S.R. Carvalho, A.L.F. Lima e Silva, S.M.M. Lima e Silva, The use of non-linear inverse problem and enthalpy method in GTAW process of aluminum, Int. Comm. Heat Mass Transfer 66 (2015) 114–121. [22] R.H. Stevens, Aluminum Alloys: Metallographic Techniques and Microstructures. Metallography and Microstructures, vol. 9, ASM Handbook. ASM International, 2004. [23] S. Kou, Welding Metallurgy, J. Wiley & Sons, Hoboken, NJ, 2003.