**Numerical and Metallurgical Analysis of Laser Welded, Sealed Lap Joints of S355J2 and 316L Steels under Di**ff**erent Configurations**

**Hubert Danielewski 1,\*, Andrzej Skrzypczyk 1, Marek Hebda 2, Szymon Tofil 1, Grzegorz Witkowski 1, Piotr Długosz <sup>3</sup> and Rastislav Nigroviˇc <sup>4</sup>**


Received: 18 November 2020; Accepted: 16 December 2020; Published: 20 December 2020

**Abstract:** This paper presents the results of laser welding of dissimilar joints, where low-carbon and stainless steels were welded inthe lap joint configuration. Performed welding of austenitic and ferritic-pearlitic steels included a sealed joint, where only partial penetration of lower material was obtained. The authors presented acomparative study of the joints under different configurations. The welding parameters for the assumed penetration were estimated via anumericalsimulation. Moreover, a stress–strain analysis was performed based on theestablished model. Numerical analysis showed significant differences in joint properties, therefore, further study was conducted. Investigation of the fusion mechanism in the obtained joints wascarried out using electron dispersive spectroscopy (EDS) and metallurgical analysis. The study of the lap joint under different configurations showed considerable dissimilarities in stress–strain distribution and relevant differences in the fusion zone structure. The results showed advantages of using stainless steel as the upper material of a microstructure, and uniform chemical element distribution and stress analysis is considered.

**Keywords:** laser beam welding; sealed lap joints of dissimilar materials; austenitic and ferritic-pearlitic steels; numerical simulation; microstructure analysis

#### **1. Introduction**

A number of analytical, numerical, and empirical studies have shown the potential of laser beam welding (LBW). First commercially applied in 1970 [1–3], LBW has become one of the most widely studied techniques of joining metallic parts.A focused laser beam causes rapid vaporization and ionizationof the metal. The keyhole effectallows lap welding [4,5]. The weld bead geometry and the joint properties are related to the process parameters and the properties of the welded materials [6,7]. Some process parameters depend on the laser type used (wavelength, transverse mode, pulse/continuous-wave operating mode), the laser machine system (spot size, single or multi-spot optics, focal length), and the programmed parameters (emission frequency, output power, welding speed, focal point position). The material properties, such as thermal conductivity, specific heat capacity, latent heat, and thermal diffusivity, affect the welding dynamics andoutcome [8]. The laser beam penetration is related to the surface reflectivity and the ionization effect of the metalvapor, therefore, the intensity of plasma generation varies for different materials [9]. The beam penetration through two materials is a complex phenomenon, moreover, in dissimilar lap joints, the results of the welding process depend on the welded material configuration.

Numerous studies of laser welding technology, including lap welding, are being carried out. Many of those studies focus on joining special application materials such as titanium, aluminum and nickel-based alloys, zinc-coated steels, or a combination of these materials [10,11]. However, as theresearchersfocus on one configuration type, there is a lack of publications showinghow joint properties change under different configurations. Therefore, the authors presented acomparative study of sealed lap joints with partial penetration, where low-carbon and stainless steels are welded alternately using a laser beam [12]. The joints are intended for use in pipeline components and large crude oil storage tanks, where high joint quality and strength are critical characteristics [13,14].

Nowadays, the development of computing power makes it possible to perform advanced calculations of welding processes, based on the finite element method (FEM) and computer-aided design (CAD) geometry, divided into finite elements (FEs). By solving the heat transfer problem, using differential equations, many physical phenomena, including phase transformation and different heat transfer mechanisms, can be taken into account [15–17]. Nevertheless, in numerical simulations, some parameters cannot be used in the model directly as input data, and some simplificationsare necessary [18]. Simple numerical models of laser welding provide a thermal solution where the fusion zone (FZ) and the heat-affected zone (HAZ) dimensions can becalculated. Nevertheless, a more advanced thermo-mechanical solution, a stress–strain analysis, can be carried out. The simulation of a lap joint is complex, and the complexity of the process increases when materials with different thermophysical material properties are to be joined. Realistic results can be obtained by developing an accurate heat source (HS) model andadequate boundary conditions [19,20].

The welding of dissimilar materials is problematic, especially when a sealed lap joint is considered. Many publications report the study of joint properties based on process parameters, while neglecting other aspects [21]. In this paper, the authors proposed a new approach to the problem, by analyzing the joint depending on the configuration of welded materials. On the basis of a numerical simulation, welding parameters were estimated, and stress–strain analysis was performed. The structure of the welds wasstudiedusing metallographic and electron dispersive spectroscopy (EDS) analyses. The researchshowedadvantages and disadvantages of austenitic and ferritic-pearlitic steels welded under different configurations ina sealed lap joint.

#### **2. Material and Experimental Design**

#### *2.1. Materials*

The materials used in the experiment were 4 mm thick steel sheets in grades S355J2 (according to EN 10025-2 [22]) and 316L (according to ASTM A240 [23]), with dimensions of 50 mm × 20 mm. The S355J2 steel is a commonly used, unalloyed, low-carbon construction steel with a ferritic-pearlitic structure. The other material is austenitic stainless steel 316L, with a high content of chromium and nickel. Both steels are characterized as materials of good weld ability, however, the differences in the chemical composition (specified in Table 1) affect the welding process and the properties of the joints.


**Table 1.** Chemical composition of materials used.

The materials used have different thermophysical properties that are not constant and change with temperature (Figures 1 and 2). This phenomenon affects the laser beam absorption, which increases with material temperature during welding and plays a significant role in the process dynamics. Figures 1 and 2 show the material database with the modified specific heat used to model the latent heat effect, calculated using JMatPro (Sentesoftware, Guildford, Surrey, UK) included in the Simufact Welding 8 software library (MSC Software Company, Hamburg Germany).

**Figure 2.** The temperature-dependent thermophysical properties of S355J2 steel.

The thermophysical properties for the abovementioned materials have a dissimilar range. For austenitic steel, the thermal conductivity uniformly increases in the range of 14 to 34 W/(m·K). Moreover, the specific heat capacity has linear characteristics in the range of 0.49 to 0.59 J/(g·K), and from 600 ◦C it is linearhorizontal. However, for low-carbon steel, these properties are related to the material phase, and for the austenitic phase, the linear dependence of thermal conductivity in the range from 15 to 33 W/(m·K) can be observed. On the other hand, for the ferrite and pearlite phase, the thermal conductivity decreases from 46 W/(m·K), until it reaches 800 ◦C, where it overlaps within austenite conductivity. In S355J2 steel, for the austenitic phase, specific heat capacity increases from 0.3 to 0.62 J/(g·K), with a characteristic similar to linear for the ferrite and pearlite phases, and specific heat capacity rapidly increases to 800 ◦C, reaching the value of 0.86 J/(g·K), and then rapidly decreases; from 900 ◦C, it overlaps within the austenite value.

#### *2.2. Numerical Simulation Procedure*

For a programming simulation of the welding process, particularly LBW, calibration of the HS model is required and, therefore, at the preliminary stage, by comparing the results of trial welding with the simulation, an accurate model was obtained [24]. Softwarecommonly used for welding simulationsare based on solving heat-mass flow phenomena (ANSYS with the Fluent module (Ansys Inc., Southpointe, Canonsburg, PA, USA), software with an additional welding module (Abaqus), and software dedicated towelding applications, such as SYSWELD and Simufact Welding [25,26]. From the aforementioned software, Simufact Welding was used for estimating welding parameters, and performing stress and strain analysis. The selected program, based on Marc solver (MSC Software

Company, Hamburg Germany), is software dedicated to welding applications, and provides sufficient accuracy of results with a relatively quick calculation time. Simulations of conventional arc welding use the double-ellipsoid Goldak model, however, simulations of laser welding are based on volumetric heat sources (conical and cylindrical), with uneven power distribution (Gaussian parameter). Moreover, some methods use a combination of HSs, with a conical (1) source for simulating the keyhole effect, and a disc-shaped source for simulating laser beam absorption by the material surface (Figure 3) [27,28].

**Figure 3.** Heatsource model of laser lap joint welding.

Conical volumetric heat source with the Gaussian distribution can be described by the following equation:

$$Q(x, y, z) = Q\_0 \exp\left(-\frac{x^2 + y^2}{\left(r\_i + \frac{r\_l - r\_i}{z\_l - z\_i}(z - z\_i)\right)^2}\right) \tag{1}$$

where *Q*0—the maximum heat flux density in a volumetric heat source, *rt* − *ri*—the dimensions of the upper and lower conical radius, *zt* − *zi*—the depth of the conical heat source, *x*, *y*, *z*—the heat source coordinates.

Solving the governing heat equation based on Fourier's law for three-dimensional heat conduction (2), witha partial differential equation in a nonlinear form, is done with the following equation:

$$
\rho c(T)\frac{\partial T}{\partial t} = \frac{\partial}{\partial \mathbf{x}} \Big( k(T)\frac{\partial T}{\partial \mathbf{x}}\Big) + \frac{\partial}{\partial y} \Big( k(T)\frac{\partial T}{\partial y}\Big) + \frac{\partial}{\partial z} \Big( k(T)\frac{\partial T}{\partial z}\Big) + q\_{\overline{v}} \tag{2}
$$

where *c(T)*—temperature-dependent specific heat capacity; *k(T)—*temperature-dependent thermal conductivity; *qv—*volumetric internal energy; *x, y, z—*space coordinates; *T—*temperature; ρ *—*density; and *t—*time.

The conical heat sourcecan be described as follows:

$$q\_l(\mathbf{x}, \ y, z) = \frac{9\eta\_l \eta\_l e^3}{\pi (e^3 - 1)(z\_l - z\_i) \left(r\_l^2 + r\_l r\_i + r\_i^2\right)} \exp\left(-\frac{3\left[\left(\mathbf{x} - vt\right)^2 + y^2\right]}{\left(r\_l - \left(r\_l - r\_i\right)\frac{z\_l - z}{z\_l - z\_i}\right)}\right) \tag{3}$$

where *ql*—heat flux, *zt* − *zi* —z coordinates (heat source depth), *rt* − *ri*—upper and lower conical radius, *<sup>e</sup>*—natural logarithm, *rt* <sup>−</sup> (*rt* <sup>−</sup> *ri*) *zt*−*<sup>z</sup> zt*−*zi* —linear decrease in distribution along theconical heat source, *Pl*—laser power, η*l*—laser heat source efficiency.

Thermal conductivity, specific heat, and emissivity in the heat transfer analysis depend on the temperature, however, the used model is based on the assumption that mass density is constant. By an extrapolating or interpolating procedure, the temperature-dependent properties are averaged. Latent heat is related to phase transformation from solid to liquid metal, or vice versa. Phase change modeling is a complex process, however, using simplification, where latent heat is uniformly released in the solid–liquid range, it can be calculated [29,30].

Due to the convection–diffusion effect involved, the Petro–Galerkin model with nodal velocity vectors was used, which can be described as follows:

$$\frac{\partial T}{\partial t} + v \cdot \nabla T = \nabla \cdot (\kappa \nabla T) + Q \tag{4}$$

where *v*—nodal velocity vector, *T*—temperature, κ—diffusion tensor, *Q*—source term.

The surface energy is determined by calculating the thermal energy that affects the boundary conditions, including thermochemical ablation. These phenomena affect the convective heat flux, and the mass and the enthalpy flow are related to molecular diffusion. The surface energy is correlated with the heat transfer to the material by conduction through a heat-mass flow towards the surface as a result of the evaporation of the material.

Specimens with a size of 50 mm × 20 mm were used for the FE model. Three-dimensional solid hexahedral finite elements were used while meshing. The sizes of the elements were determined by adjusting the resolution and accuracy of the temperature distribution in the regions of severe thermal gradients. A mesh convergence study was performed, within FE sizes of 1.0000, 0.7500, 0.5000, 0.2500, 0.1250, and 0.0625 mm. For 0.0625 and 0.1250 mm, no relevant differences in weld geometry and heat distribution were observed, however, there were some differences between 0.1250 and 0.2500 mm. Therefore, the nominal FE size was set as 0.2500 mm, and in an area where a temperature exceeding 400 ◦C may occur (Figure 4, region 4), an FE refinement was performed, and the FE sizewas set as 0.1250 mm. Two 4 mm thick sheets (Figure 4, elements 1 and 2) were oriented in a lap configuration and fixed in three-dimensional space by additional elements (Figure 4, element 3). The welding trajectory was set in the center of the refinement area (9.5 mm from the sheet edge).

**Figure 4.** Laser welding finite element (FE) model with defined elements: 1—topsheet 2—bottom sheet, 3—fixed bearings,4—refined FE area, 5—heat sources, 6—welding trajectory.

The geometries of the heat sources (Figure 3) (3), are related to the used welding optics. In this case, the focal length was equal to 270 mm with a focal point diameter of 0.3 mm. Nevertheless, the HS geometry calibration for obtaining more accurate results was required. Therefore, a trial calibration weld at the speed of 1 m/min and output power equal to 4 kW was performed (to achieve a keyhole effect). A comparison with the simulation results showed a small discrepancy (width of the face of the weld, with anerror less than 15%), and so the heat source geometry was adjusted until the error was less than 1% [31–33]. According to the preliminary calibration performed, a conical heat source

with a depth equal to 7 mm, upper radius rt equal to 0.4 mm, and lower radius ri equal to 0.2 mm, as well as a disc-shaped heat source, with a radius equal to 0.5 mm and depth equal to 0.05 mm, were programmed. The volume heat fraction defined the laser power division between the conical and disc HS and for laser keyhole welding, the value was set as 0.95, which means that 95% of the total power was assigned to the conical HS. Moreover, the Gaussian distribution parameter for the disc HS was defined as 1.0 and for the conical HS as 2.8, and they are related to TEM01\* (CO2 laser, transverse mode). The materials used in the simulation were structural steel ofgrade S355J2 and 316L austenitic stainless steel. Laser welding with a keyhole effect makes it possible to obtain deep material penetration, however, for a CO2 laser, the metal surface has high reflectivity and, therefore, HS efficiency for S355J2 was assumed as 0.6, and for 316L as 0.7. The programmed efficiencies are related to laser beam interaction with the materials used, where the reflectivity and the ionization effect have a significant influence on those aspects [34–36].

The simulation of dissimilar, sealed, lap joint laser welding, for two configurations of the top material, was carried out using the Simufact Welding software. Simufact solves a heat transfer problem in solid material, considering the convection–radiation effect, however, no mixing effect of welded material can be taken into account. Figure 5a shows a keyhole in the cross-section during the lap welding, and Figure 5b shows the top hat of the keyhole, during the formation of the face of the weld. The results are presented for the temperature scale in the range of 20 to 3070 ◦C (from normal condition to the boiling point).

**Figure 5.** Simulation of laser lap joint welding using the deep penetration mode: (**a**) view of the keyhole effect in the cross-section, (**b**) top view of a moving keyhole, and weld formation.

The first joint configuration assumes low-carbon S355J2 steel placed on the top, and the second configuration assumes the opposite: S355J2 steel on the bottom and 316L stainless steel on the top. The research assumed obtaining a sealed lap joint, with partial weld penetration in the bottom sheet [37,38]. Numerical simulations of laser welding were performed at a constant speed of 1 m/min and variable output power in the range of 3 to 6 kW, which increased with each subsequent step by the value of 0.5 kW until the assumed penetration was obtained. The heat source operating time, based on the HS speed and the trajectory length, was equal to 1.2 s, however, the full simulation time, including cooling, was programmed as 30 s. According to the performed simulations, the obtained weld bead geometry and stress–strain distribution (according to the determined measurement points) were studied.

#### *2.3. Experimental Welding Procedure*

The first step of the experimental research was the HS validation. For this purpose, a preliminary test according to the procedure described in Section 2.2 was carried out. Based on the aforementioned procedure, calibration of the HS geometry and the efficiency coefficient was carried out. The parameters estimated in the numerical simulation, welding speed equal to 1 m/min and output power 6 kW, were used to perform trial welds with a TrumpfTruFlow6000 CO2 CW laser (wavelength 10.6 μm)

(Trumpf Group, Ditzingen, German), using welding optics with the focal length of 270 mm and the spot diameter equal to 0.3 mm. This type of laser has worse parameters than fiber or disc lasers, however, further planned research requires the use of single and twin spot optics for an extended weld area, which are available for this type of laser. To shield the welding zone, helium (5.0) was used, with the flow rate equal to 20 l/min conveyed coaxially. The laser beam was focused on the surface of the top plate in a PA (flat) position and the welding trajectory was established based on the boundary conditions of the simulation [39]. Using single pass welding, lap joints in two configurations were obtained: the 1st with the low-carbon steel placed on the top (and the 316L sheet at the bottom), and the 2nd with the 316L steel placed on the top (and the S355J2 sheet at the bottom). Welding procedure qualification was performed according to PN-EN ISO 15609-4: 2009 [40] and the joint quality level was specified according to PN-EN ISO 13919-1 [41]. The results of the simulation and the metallographic analysis of the trial joints were studied.

#### *2.4. Microstructure Analysis*

Validation of the simulation results is related to a weld bead build analysis, therefore, measurements of characteristic geometries were carried out. Specimens for further analysis were prepared by cutting them in half (according to the cross-section of the weld), polishing, and etching with Adler reagent. A metallographic analysis of weld structures according to PN-EN ISO 17639 [42], using a Hirox KH-8700 (Hirox Co Ltd., Tokyo, Japan) confocal digital microscope and a scanning electron microscope JSM-7100F (JEOL Ltd., Tokyo, Japan), was carried out. By visual tests and weld bead build evaluation, welding defect detection and inclusion analysis wasperformed. Investigation of the weld uniformity by element distribution analysis, using a JEOL scanning electron microscope with an electron dispersive X-ray spectroscopy analyzer, was carried out [43].

#### **3. Results**

#### *3.1. Global Observation*

Due to the significant mismatch of thermal conductivities, specific heat capacity, and surface absorptivities in the welded materials, the FZ and HAZ are different. Moreover, a numerical simulation analysis showed further differences. The research aimed to obtain sealed lap joints, with a bottom sheet welded approximately to 3/4ths of its thickness. A macroscopic analysis showed only partial material mixing in the bottomsheet (Figure 6). The decision of which configuration to choose for better properties is complex and requires an extended thermalstress–strain numerical analysis, as well as micro- and macro-structure examinations. The preliminary visual tests (VTs) showed a good weld build, lack of defects, a convex face of the weld, and assumed partial penetration of the bottom plate was obtained, therefore, according to therequirement of the PN-EN ISO 13919-1 standard, a B quality level was obtained [44,45].

In the obtained welds beads, some separate lap joint characteristic regions were indicated, where: S1—root of the weld, S2—intersheet area, S3—uniform mixing of top plate weld area, S4—HAZ region (Figure 6). Potentially non-uniform mixing of fused materials was detected, which can affect the joint strength and corrosion resistance and can lead to micro-cracking. Therefore, a microstructure analysis, extended to include a numerical stress–strain study for lap welds evaluations, was carried out [46,47].

**Figure 6.** The build of welds with the region defined for further investigation, (**a**) 1st joint configuration (S355J2 steel on the top), (**b**) 2nd joint configuration (316L steel on the top).

#### *3.2. Numerical Simulation*

The weld bead geometry based on the solid–liquid range was analyzed. On the basis of the programmed HS parameters and established boundary conditions, a numerical simulation of laser lap welding was performed at a constant speed and varying output power. The total thickness of the materials positioned in the lap configuration was equal to 8 mm. As sealed joints were to be obtained, the partial penetration of the weld was up to 7 mm at an output power equal to 6 kW. The differences in the welded materials, especially the absorption coefficient, the thermal conductivity, and the vaporization–ionization effect, made it necessary to change the HS efficiency for each simulation configuration. Therefore, the aforementioned HS efficiency for the low-carbon steel was adopted as 0.6 and for the stainless steel as 0.7 [48,49].

For the 1st configuration, aweld depth equal to 7.1 mm, a weld face width equal to 2.87 mm, and a width of the overlap region equal to 1.92 mm were calculated, with experimental values equal to 7.16 mm, 2.51 mm, and 1.83 mm, respectively (Figure 7). The calculation results for configuration 2 showed similar values, where the depth was equal to 7.15 mm, the weld face width to 3.1 mm, and the width of the overlap region to 1.7 mm. The experimental values were equal to 7.3 mm, 3.15 mm, and 1.59 mm, respectively (Figure 8). Some differences in the weld build were observed. The weld obtained in the 2nd configuration had a typical U-groove weld build, with a wider face of the weld, however, the weld waist was narrower. The 1st configuration showed a narrower face of the weldand a wider weld waist.

**Figure 7.** Comparison of the simulation and the trial joint weld build—1st configuration (S355J2 steel on the top).

**Figure 8.** Comparison of the simulation and the trial joint weld build—2nd configuration (316L steel on the top).

Figures 7 and 8 compare the simulation and experimental results. The predicted weld bead geometries were compared to the measurements obtained from the trial joints [50]. The results showed good agreement between the predicted and measured characteristic dimensions of weld geometries.

The studied weld bead build showed some differences but no welding defects or incorrect weld structures were observed. Therefore, in order to decide which configuration gives better results, a further study, starting from a stress–strain distribution analysis, was carried out.

Considering the stress and strain analysis, it is relevant to study the variability of those phenomena in time. Therefore, measurement points, located across determined lines (characteristic joints areas), as shown in Figure 9, were used to perform the analysis, referred to the welding and cooling stages [51–53].

**Figure 9.** Defined points and lines for stress–strain numerical analysis, selected in critical joint areas.

The measurement points (1–12) were defined along the M1 and M2 measurement lines—parallel to the weld axis, where M1 is the inside line (near fusion line) and M2 is the outside line (near the HAZ line), and D1, D2, and D3 lines are perpendicular to the weld axis, where D1 is the top line, D2 is the central line, and D3 is the bottom line of weld profile. Based on the defined measurement points and lines, a stress–strain analysis for the 1st and 2nd configurations was carried out (Figures 10–14).

**Figure 10.** The effective plastic strain curve for the 1st (S355J2 steel on the top) and 2nd (316L steel on the top) configurations along the M1 line.

**Figure 11.** The effective plastic strain curve for the 1st (S355J2 steel on the top) and 2nd (316L steel on the top) configurations along the M2 line.

**Figure 12.** The maximum principal stress curve for the 1st (S355J2 steel on the top) and 2nd (316L steel on the top) configuration along the D1 line—top.

**Figure 13.** The maximum principal stress curve for the 1st (S355J2 steel on the top) and 2nd (316L steel on the top) configuration along the D2 line—central.

**Figure 14.** The maximum principal stress curve for the 1st (S355J2 steel on the top) and 2nd (316L steel on the top) configuration along the D3 line—bottom.

Effective plastic strain is a monotonically increasing scalar value which is calculated incrementally, as a function of the plastic component of the rate of the deformation tensor. The effective plastic strain increases whenever the material is actively yielding [54]. An analysis of the effective plastic strain results is shown below, according to the defined M-lines. Welding time alone (1.2 s) during the performed simulation is shown by the perpendicular line.

The tensorial strain values are not monotonically increasing as they reflect the current, the total (elastic+plastic) state of deformation, primarily in the HAZ and across the fusion line where, due to the constituent thermal expansion mismatch, the phase transformation is the most intense. The value of effective plastic strain is the integral of stepwise increments of plastic deformation continuing for a period of time; therefore, the analyzed values are shown according to the welding time. The results show the highest effective plastic strain value across the M1 line and are similar for both joint configurations, however, in the 2nd case, the curve is sharper. The values obtained across the M2 line show greater differences and, in the 2nd configuration, they were almost 4.7 times greater than in the 1st configuration. While the values for the M1 line obtained at the surface are similar (points 1 and 2), in the central zone of the 1st configuration, they are equal to the bottom zone values (points 5 and 6). For the 2nd configuration, three separate regions of effective plastic strain can be identified. The computation times of the welding process (without cooling) were equal to 1.2 s, therefore, for these specific points, some changes in the plots can be observed [55–57].

A maximum principal stress (MPS) analysis, concerning values changing in time, was performed as well (Figures 12–14). The points selected for the MPS analysis were related to the weld depth (D-lines). By performing a distribution analysis of the maximum principal stress, the areas of stress concentrations can be identified. These particular areas show where crack propagation could start when critical material parameters are reached. An analysis of the results can shown whether in particular joints configurations, the MPS did not exceed the critical values, disqualifying one of the lap joints [58,59].

Analysis of MPS showed higher stress concentrations across the D2 line, particularly in points 5 and 6 (in the middle of the HAZ). The greatest calculated values, equal to 540 MPa, occurred in the 1st configuration, where low-carbon steel is located on the top. In the 2nd configuration, the highest stress occurred across the D2 line as well, however, it did not exceed 400 MPa. At the surface (D1 line), the maximum principal stress exceeded 300 MPa only in point 4. Across the D3 line, the maximum principal stress is similar in both configurations, with slight differences in austenitic steel (in 1st configuration). An increase in the maximum principal stress in D1 and D2 occurred after the welding, and MPS values were related to the movement of the HS, and increased rapidly during the cooling stage, after the heat source was turned off [60–63]. One must keep in mind that the measurement points were located on the sectional plane in the middle of the welding trajectory.

#### *3.3. Metallographic Analysis*

Important differences between stress–strain distribution showed significant dissimilarity in joints properties, however, considering only simulation, it is hard to define which configuration provides better results, and further analysis is required. Therefore, the microstructure of characteristic weld areas, chosen alloyed elements distribution (Figure 21), and the precipitates in the fusion zone were studied and the results are shown below.

Figure 15 shows the microstructure of the base material (BM). The structures shown are typical for the used materials, with a ferritic-pearlitic microstructure of low-carbon S355J2 steel (Figure 15a) and an austenitic structure of stainless 316L steel (Figure 15b). The HAZ areas (Figure 16) in both materials are different. In 316L steel, the HAZ is narrow, with grain refining along the fusion line, and an increase in the volume fraction of ferrite δ. The HAZ identified in S355J2 steel is much wider and is made up of three separate regions: overheated zone (OZ), normalization zone (NZ), and partial recrystallization zone (RZ) [64,65]. The HAZ was investigated according to the S3 region (Figure 6).

**Figure 15.** The microstructure of welded base materials: (**a**) 316L steel, (**b**) S355J2 steel.

**Figure 16.** The microstructure of the heat-affected zone (HAZ): (**a**) 316L steel, (**b**) S355J2 steel.

The weld composition is related to the mixture of the low-carbon and stainless steel alloying components, and the weld structure is a result of the solidification process of this newly formed material (Figure 17).

**Figure 17.** The weld structure approximately in the middle of the upper plate: (**a**) 1st configuration (S355J2 steel on top), (**b**) 2nd configuration (316L steel on top).

The obtained structure has a dendritic build, however, in the case where low-carbon steel is the topsheet, the structure is composed of coarse-grained dendrites and acicular ferrite content. Meanwhile, for the configuration where stainless steel is the topsheet, the observed structure has a typical pillar dendritic build. In the 1st case, an austenitic-ferritic structure can be observed, however, in the 2nd case, there is only an austenitic structure. Moreover, fusion zones in the bottom sheet showed regions with non-uniform structures (Figure 18), according to region S1 (Figure 6) [66,67].

The SEM analysis showed differences in the weld structures, where separate areas inside the FZ were identified (Figure 19A,B). In area A, a pillar-dendritic structure was observed, while in area B, a dendritic cellular structure dominated.

In the root of the weld (region S1), where the weld penetration was achieved, some differences were observed. The bottom weld area in austenitic steel had a structure similar to that of the BM (Figure 20a), while in ferritic-pearlitic steel, the structure was similar to that of the HAZ (Figure 20b).

**Figure 18.** Dissimilarity in the weld structure, down plate: (**a**) 1st configuration (S355J2 steel on top), (**b**) 2nd configuration (316L steel on top).

**Figure 19.** The weld structure approximately in the middle of the down plate, 2nd configuration (316L on top), with identified differences in structure (central picture) and the enlarged areas: (**A**)—pillar dendritic structure, (**B**)—cellular dendritic structure.

**Figure 20.** The fusion line of the weld in the down plate: (**a**) 1st configuration (S355J2 steel on top), (**b**) 2nd configuration (316L steel on top).

The non-uniform fusion zone structurewas confirmed through the EDS analysis. Therefore, according to the defined regions (presented in Figure 6), important differences in chemical compositionwere studied (Table 1) based on iron, nickel, and chromium distributions in the weld cross-section (Figure 21).

**Figure 21.** Iron, chromium, and nickel distribution according to S3 region: (**a**) 1st configuration, (**b**) 2nd configuration.

The uniform distribution in the topsheet FZ proves the high mixture factor in this region, but the distribution in the bottomsheet FZ is more problematic (Figure 22).

**Figure 22.** Bottom fusion zone iron and chromium map distribution (region S1): (**a**) 1st configuration (S355J2 steel on top), (**b**) 2nd configuration (316L steel on top).

The linear distribution of Fe, Cr, and Ni in the upper plate (region S3) showed greater differences for the 1st configuration (low-carbon steel as the topsheet), while in the 2nd configuration, the distribution was more uniform, however, an analysis showed largedifferences in the fusion line region. Moreover, according to the iron and chromium distributions, the joints made in the 1st configuration showed low iron content and a largeamount of chromium, while in the 2nd configuration, the iron distribution was almost uniform, with slightly increased chromium content.

#### **4. Discussion**

The welded materials have important differences in thermo-physical properties related to their chemical composition (Table 1). Both materials are characterized by good weldability, but due to the differencesdiscussed above, joining them by welding is problematic, especially when we consider a lap joint with partial penetration in the bottom plate, as described in this paper.

Global observation showed differences in the weld builds for both assumed joint configurations. The macroscopic analysis showed a potentially non-uniform structure in the fusion zones. Therefore, in order to investigate the quality level of the joints and to choose which configuration gives better results, numerical analysis and then metallographic studies were carried out.

Based on the developed numerical model, calibrated in the preliminary research, the welding parameters for the sealed lap joint were estimated. According to the established boundary conditions and the programmed simulation parameters, welding speed equal to 1 m/min with 6 kW of output power results in the assumed partial penetration. The performed simulations provided results where the differences in the width of the face of the weld were less than 14.5% compared to the trial joint for the 1st configuration and 1.62% for the 2nd configuration. The difference in the weld depth, calculated and obtained via experimental welding for the 1stconfiguration, was equal to 0.85% and for the 2nd configuration, to 2.1%. The width of the overlap region resulted in a 4.9% mismatch, comparing the simulation and the trial joint results for the 1st configuration, and 6.9% for the 2nd configuration. The numerical model obtained gave accurate results of the estimated weld bead geometry, however, the mismatch in the weld width for the 1st configuration, where low-carbon S355J2 steel was the topsheet, resulted in a 14.5% difference. The numerical model did not include the Marangoni effect and the surface tension, which has a great impact on the face of weld geometry. However, based on the small differences in other (Figures 7 and 8) weld geometries, the established model gave realistic results and was used for joint stress–strain analysis [68–70].

In order to define the properties of the obtained joints based on the simulation results, the distribution of the effective plastic strain and the maximum principal stress was studied according to the determined lines (Figure 9). The M lines define points distributed parallel to the weld, near the fusion and heat-affected zones lines, and D lines are related to the weld penetration depth and are distributed perpendicular to the weld axis. According to the M1 and M2 lines, the effective plastic strain analysis showed amaximum value equal to 0.035 and was similar for both joint configurations, however, in the 2nd case, the values increased more rapidly. For the 1st case, the increase was slower, however, after the end of the welding cycle (1.2 s), it was still rising in the cooling stage, to 2.6 s. For the M2 lines (located 3.6 mm from the weld axis), the effective plastic strains were much smaller and the maximum calculated value (which occurred in the 2nd configuration) did not exceed 0.015, while for the 1st configuration, it was equal to 0.03. The differences in the effective plastic strain can be related to the different values of thermal conductivity (curves from 0 to 1.2 s), which for low-carbon steel resulted in a higher accumulation of thermal energy in the fusion zone and phase transformation (after 1.2 s) in the cooling stage [71].

The maximum principal stress values across the Dlines showed that the stress characteristic was divided into two separate cycles. The first was related to heating and cooling during movement of the HS, however, this value did not exceed 150 MPa. The second was related to material cooling, when the heat source was turned off. The maximum principal stress increased on the surface of the topsheet and its value ranged from 60 to 290 MPa for the 1st configuration and 20 to 320 MPa for the 2nd (the curve characteristics were similar for all measurement points). In the center line (D2), a greater stress value occurred in joint configuration number 1 and was equal to 540 MPa, while for the 2nd configuration, the value did not exceed 400 MPa. For the D3 lines, the values were less than 240 MPa (1st configuration) and equal to 200 MPa (2nd configuration). The maximum principal stress had a greater value in the measurement points located closer to the weld axis, withthegreatest difference occurring across the D2 line, where at points 5 and 6 they were more than 5.5 times greater than in points 7 and 8. The calculated maximum principal stress had a greater value in the 1st configuration and was related to the thermal gradient and the phase transformation of low-carbon steel [72,73].

A metallographic analysis confirmed the typical structure of the welded materials in the BM region, austenitic in 316L steel and ferritic-pearlitic in S355J2 steel. Significant differences in the heat-affected zones were observed. In stainless steel, the HAZ was very narrow and no growth of austenite grains was observed, while the elongated grains of the ferrite formed a discontinuous network around the austenite grains. These regions occur in steels with the structure of a native material consisting of austenite with the participation of ferrite δ. At high temperatures near the fusion line, the transformation of γ→δ occurs, which begins in the existing ferrite δ grains and progresses in areas with increased chromium concentration. After re-cooling, these areas do not achieve an equilibrium structure and the proportion of ferrite δ is increased. The HAZ of low-carbon steel is considerably wider and is made up of threeregions: an overheated zone, a normalization zone, and a partial recrystallization zone. These phenomena are related to phase transformation and structural changes during the solidification and cooling process. The obtained HAZs (including microstructure and geometry) of welded materials are typical for this steel grade welded using the LBW method [74].

In the global observation, some non-uniform microstructure of the fusion zones was detected, therefore, an extended weld microstructure analysis was carried out. The structure observed in the topsheet FZ had a typical uniform dendritic build, however, in the 1st configuration, coarse-graineddendritescontaining acicular ferrite were observed, while in the 2nd configuration, a typical pillar-dendritic build occurred. When the keyhole only partially penetrates the bottom workpiece, the influence of the flow field is clearly evident from the corresponding solidified structure. In the case of partial penetration, discrete growth bands occurred in the entire solidified weld bead, suggesting severe fluctuations in the flow field and the growth process [75]. On the contrary, in the topsheet, a full penetration keyhole was performed, where a columnar structure and an equiaxed zone along the centerline, similar to that shown in Figure 17b, can be observed. Discrete growth bands or striations occurred occasionally in this case, but the structural pattern was maintained within these bands. It should be noted that the laser output power in both cases was similar.

The fusion zone structure in the overlap region are similar to those observed in the FZ upper region, however, in the bottom plate, non-uniform structures were identified (Figure 18). There were differences in the fusion structure (Figure 19), where a pillar-dendritic build in region A and a cellular-dendritic structure in region B were observed. In the root of the weld, at the bottom of the fusion zone where weld penetration reaches, the structure was similar to the BM for stainless steel and the HAZ structure for low-carbon steel. In this region, the temperature exceeded the melting point, but only slightly, and according to the direction of heat diffusion in the XYZ axis, the heat flow to the BM was higher [76].

Homogenous weld builds are related to the uniform distribution of chemical elements. The welded materials have important differences in chemical composition, where S355J2 steel has a high content of iron and only trace amounts of chromium and nickel, while stainless steel contains more than 16% chromium, 11% nickel, and a balanced content of iron. Therefore, an analysis of defined regions in the cross-section of the welds was performed based on the linear and map distribution of these elements. The study showed bigger differences between the BM and the FZ in the distribution of Fe, Cr, and Ni in the 1st configuration. A clear boundary between the base material and the fusion zone was observed, however, the distribution curve had an almost vertical characteristic. For the 2nd configuration, the differences were smaller, with peak changes along the fusion line, which is related to chromium diffusion and concentration across the fusion line (phase transformation and increasing

content of ferrite δ). An EDS analysis of the FZ bottom region showed further differences, particularly in chromium and iron distribution [77]. For the 1st configuration, small amounts of iron were detected and considerably greater differences in the detection of chromium were observed (Figure 22a). In the 2nd joint configuration, the iron distribution in the fusion zone and the base material was almost uniform, and some differences in chromium detection were observed (Figure 22b).

#### **5. Conclusions**

The study of laser lap joint welding, where low-carbon S355J2 and stainless 316L steel were joined together in two configurations (1st with low-carbon steel on the top, 2nd with stainless steel on the top), showed significant differences in the obtained results. According to the developed numerical model and the established boundary conditions, an accurate match of the simulated and experimental results of weld geometry was obtained. Assumingthat the obtained model gave accurate results, a numerical stress–strain analysis was performed. Higher values of effective plastic strain occurred in the 2nd configuration, however, the maximum principal stress values were greater in the 1st configuration. Greater differences in the measurement points occurredfor the configuration with low-carbon steel placed on the top. The microstructure analysis showed further differences, most of all in the fusion zone structure, where in the bottom plate, two separate structures were detected. A more uniform structure was observed in the 1st configuration, however, greater differences in Cr, Ni, and Fe distribution between the FZ and the BM were detected in this configuration. No porosity, cracks, or welding defects were identified, therefore, according to therequirement of PN-EN ISO 13919-1, a B quality level was assumed. According to the performed study, and based on the maximum principal stress distribution, differences in the weld structure, and the distribution of the alloying elements, the second configuration showed better properties. Therefore, the second configuration, where stainless steel is placed on the top, was chosen as a dominant joint. The microstructure study showeda non-uniform mixture of welded material in the root of the weld, however, no welding defects or precipitations in the inter-plate region were observed. This study confirmsthe impact of the welded material configuration on joint properties and shows which configuration ensures the lowest stress and strain concentration and a more uniform weld structure. Therefore, according to:


When laser lap welding of S355J2 and 316L steels is considered, the configurationin which stainless steel is placed on top provides better joint properties and is recommended.The presented results showed that when ahigh-qualitysteel joint with good strength characteristics is required, such as pipeline components or crude oil storage tanks, the welding position should be related to the stainless steel side.

Further investigation of this joint type is required, therefore, the possibility of usingtwin spot laser welding optics is planned. Theresearch will also be extended to mechanical tests, including the tensile strength and hardness tests.

**Author Contributions:** Conceptualization, H.D., A.S., and R.N.; methodology, H.D., A.S., M.H., and P.D.; software, H.D., G.W., and S.T.; validation, H.D., A.S., M.H., P.D., and R.N.; formal analysis, H.D., A.S., R.N., S.T., and P.D.; investigation, H.D. and G.W.; resources, H.D.; writing—original draft preparation, H.D., A.S., M.H., P.D., and R.N.; writing—review and editing, H.D., A.S., M.H., and G.W.; visualization, H.D. and G.W.; ervision, A.S. and M.H.; project administration, H.D.; funding acquisition, H.D. All authors have read and agree to the published version of the manuscript.

**Funding:** This research was funded by NCBiR, grant number LIDER/31/0173/L-8/16/NCBR/2017: Technology of manufacturing sealed weld joints for gas installation by using concentrated energy source.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Review* **Manufacture and Performance of Welds in Creep Strength Enhanced Ferritic Steels**

#### **Jonathan Parker \* and John Siefert**

Electric Power Research Institute, Charlotte, NC 28262, USA **\*** Correspondence: jparker@epri.com; Tel.: +1-704-595-2791

Received: 3 June 2019; Accepted: 11 July 2019; Published: 13 July 2019

**Abstract:** Welding is a vital process required in the fabrication of 'fracture critical' components which operate under creep conditions. However, often the procedures used are based on 'least initial cost'. Thus, it is not surprising that in many high energy applications, welds are the weakest link, i.e., damage is first found at welds. In the worst case, weld cracks reported have had catastrophic consequences. Comprehensive Electric Power Research Institute (EPRI) research has identified and quantified the factors affecting the high temperature performance of advanced steels working under creep conditions. This knowledge has then been used to underpin recommendations for improved fabrication and control of creep strength enhanced ferritic steel components. This review paper reports background from this work. The main body of the review summarizes the evidence used to establish a 'well engineered' practice for the manufacture of welds in tempered martensitic steels. Many of these alternative methods can be applied in repair applications without the need for post-weld heat treatment. This seminal work thus offers major benefits to all stakeholders in the global energy sector.

**Keywords:** steel; weld; high temperature; creep; fracture; advanced methods

#### **1. Introduction**

EPRI is a not for profit organization that has been providing independent technical support to global stakeholders in the electricity supply industry for over 40 years. Within EPRI's generation sector, a key research imperative is knowledge creation and technology transfer linked to the reliable, safe, and economically flexible operation of power plants. Collaborative achievements have included contributions to the development of databases of key properties for high temperature alloys, publication of recommended guidelines for design and fabrication as well as compiling case studies of in-service problems and facilitating expert root cause assessment. Technology transfer has ensured that lessons learned can be used to establish best practices; these activities include annual workshops, the publication of summary documents and identification of additional research. In all cases, excellence in science and engineering is necessary to underpin technology which will help to meet challenges associated with the safe and reliable operation of a plant.

The present paper reviews key findings from over 10 years of collaborative research which was carried out to identify and quantify the factors affecting the high temperature behaviour of tempered martensitic steels. In particular, selected information is provided from a series of EPRI-coordinated, industry-sponsored projects on the manufacture and performance of welds made in 9%Cr creep strength-enhanced ferritic (CSEF) steels. The initial work in this area provided the basis for a meaningful asset management strategy for Grade 91 steel components. This seminal project involved more than 40 participants providing over \$4 million of industry funding. The learnings and findings were realized with direct input and perspective from stakeholders representing the entire electricity supply chain. The knowledge created established that in addition to operating variables such as stress state and temperature, the high temperature behaviour of CSEF steels was dependent on a

Metallurgical Risk Factor. Meaningful assessment of component creep behaviour could thus only be achieved by integrating information from steel making and fabrication with microstructure and operating conditions, with particular emphasis on complex transient loading.

A key outcome from this research was that for metallurgically complex steels such as CSEFs, it is vital that research programmes should establish high temperature performance on steel sections which are carefully chosen and well characterized [1]. Thus, it is important that the full pedigree of the steel heat or cast selected for a research project is known or checked. For example, when research is assessing the creep behaviour of welds, the welds must be made in the same base substrate so that results can be directly compared. Similarly, it should not be assumed that the behaviour of welds made in plate will exhibit the same performance trends as welds made in tubes or pipes.

Achievements directly linked to factors controlling the high temperature performance of welds in CSEF steels are reviewed in this paper with detailed background provided in the original reports and papers [1–3]. The primary areas covered are summarized below:


**Figure 1.** Micrographs showing detail of cracking at a stub tube to header weld. Cracking found at this location at an earlier inspection had been removed by local grinding. It is noteworthy that the re-cracking event did not take place at the base of the excavation. Instead, the creep damage is focused on the metallurgically susceptible structures in the weld HAZ.

It is apparent that the uncertainty regarding component performance is in part a consequence of inadequacies during the design and fabrication. Sometimes less than optimal approaches are justified on the basis of least cost. Failure to use best practice specification, design and manufacturing methods may also be the result of a lack of knowledge or lack of control or both. In all cases, excessive variability in the condition of a component or structure at the start of service life results in major downstream uncertainties which may include problems with fractures and forced outages. With particular reference to the fabrication and high temperature performance of CSEF steels, it is apparent that the EPRI knowledge base offers technical information and guidance to minimize variability in behaviour and thus greatly simplifies asset management [12]. The information described in this review paper should be used by designers, manufactures and end users to help to meet the challenges of economic, safe and reliable operation of high energy components.

#### **2. Metallurgical Risk Factor**

The high temperature performance of CSEF steels is dependent on the details of steel making, steel processing, composition, and heat treatment [1]. Moreover, it is not possible to simply 'recover' the performance of a poorly made steel batch by application of a subsequent heat treatment. Thus, the final quality or the steel to be used starts with high quality in composition control, steel making, poring etc. While a steel section with a poor microstructure cannot be easily fixed by heat treatment, it is now very well established that the poor control of heat treatment can lead to different problems. Thus, for example, when irregularities of heat treatment result in a ferrite microstructure (rather than the martensitic structure expected), the creep strength is significantly reduced compared to the creep strength of tempered martensite.

EPRI recommendations clearly advocate performing detailed metallographic characterization and documentation of the damage present after creep testing. This assessment is particularly important for tests of long duration since these results are generally considered most representative of behaviour of components in service. In CSEF steels these long-term tests generally fracture with very little reduction of area. This creep brittle behaviour is a consequence of the formation and growth of cavities and the subsequent formation of micro and macro cracks. Careful examination using modern techniques has shown that creep cavities are formed both on prior austenite grain boundaries and other features in the martensitic microstructure such as at lath boundaries and at inclusions. The diversity of the microstructural sites which develop cavities is illustrated in Figure 2. It appears that the formation of voids at these diverse sites makes the formation of micro cracks more difficult.

**Figure 2.** Micrograph showing detail of creep cavities formed in the base metal of a Grade 92 CSEF steel. These creep voids nucleate at a size below 1 μm and grow during component life. The sample shown here is close to final fracture, yet no macro cracks had formed.

The diversity of void nucleation sites creates a challenge to tracking in-service component damage using traditional inspection methods. While the details of the number of voids formed, and the tendency for reductions in strain to fracture, is different for the different CSEF steels, research to date [1] shows that void nucleation is related to the presence of trace elements and hard nonmetallic inclusions. In Figure 3 detailed characterization of the inclusions present has been performed. The steel identified as Barrel 2 has a relatively high inclusion number density and exhibits low creep ductility. In contrast, the steel identified as Tee piece 1 has a relatively low number density of inclusions and exhibits a high creep ductility.

**Figure 3.** Number density of inclusions for two different but ASME code acceptable Grade 91 steels. A higher density of MnS was measured in the creep damage susceptible steel which contained 0.009 wt.% S (labelled barrel 2) as compared to the cavity resistant steel which contained 0.002 wt.% S (labelled tee piece 1).

An example of the link between the nucleation of a creep void, the presence of inclusions and segregation of trace elements is illustrated by the analysis results shown in Figure 4. The analysis reveals that high levels of copper are found associated with the creep void. Copper in this location can act to promote cavitation by reducing the local surface energy. A further key factor in determining whether inclusions aid the nucleation of voids is the particle size. Thus, only inclusions of a sufficient size (the critical inclusion size is directly linked to the creep stress) will be thermodynamically stable and thus able to act directly as nucleation sites.

**Figure 4.** Detailed micrograph showing inclusions associated with an individual creep cavity as well as compositional maps showing the distribution of the elements Cr, Nb, V, Cu, N, and Al associated with this location.

It is clear from root cause analyses of Grade 91 steel components [1,4] that steel composition and processing variables are linked to low creep ductility in base metal and creep cracking in weld HAZs. The observed in-service component damage cannot be simply explained as a 'one off' anomaly. Examination of failures in several Grade 91 welded components [13,14] indicates a trend where an increasing number of failures of Grade 91 welds are occurring in times below that expected based on simple design rules. The trend in the reduced creep ductility in Grade 91 steel and the link to very low HAZ creep life may be a characteristic of a significant number of components which entered service with compositions which show a high susceptibility for cavity formation. It should be emphasized that the poor creep performance of weld HAZs is a key reason for the introduction of weld efficiency factors. These factors were aimed at reducing the risk of in-service damage by increasing the component thickness. Further, reductions in weld creep performance or indeed greater uncertainty over long term performance would be expected to lead to increases in recommended weld strength reduction factors (WSRF) or weld-efficiency factors. The evidence from EPRI research shows using Grade 91 steel with low densities of inclusions and controlled levels of deleterious trace elements should significantly reduce the risk of creep cracking associated with weldments. Selection of improved base material would provide significant benefit for high energy systems.

In order to properly assess metallurgical risk, it is necessary to obtain a full chemical composition for Grade 91 steel base material. As highlighted previously [1,15,16], there is concern that the influence of tramp elements such as As, S, Sn, Sb and Cu has been underappreciated and that these elements are playing a role in the reduction in creep ductility in martensitic CSEF steels. In general, the analysis of elements can be grouped into two sets: elements required by common specifications for Grade 91 steel (14 total elements) [1] and elements for informational purposes (typically 10+ additional elements). The following approaches are typically used to determine the composition of each of the elements in EPRI research. Inductively coupled plasma optical emission spectrometry (ICP-OES) was utilized to determine the values for: Al, B, Ca, Co, Cr, Cu, La, Mn, Mo, Nb, Ni, P, Si, Ta, Ti, V, W, Zr. Inductively coupled plasma mass spectrometry (ICP-MS) was used to determine the amounts of As, Bi, Pb, Sb, Sn. Finally, combustion was necessary to determine the C, S levels while insert gas fusion (IGF) to assess the amount of O and N in the steel. It should be noted that, to ensure that sufficient information is provided for each of the requested elements, the number of required significant digits should also be specified in any specification to the laboratory performing the analysis.

Discussions of industry experience in general, and the EPRI recommendations in particular, have now resulted in agreement within ASME that for Grade 91 steels there should be a Type I and a Type II designation [16]. It has been recommended that the steel made with greater controls designated Type II, will have higher Allowable Design stress values [16]. This recommendation is reflected in the ASME Code Case 2864, Table 1 and associated documentation. Similar specifications have been approved by ASTM for the component specific requirements. Discussions continue about the influence of trace elements on creep performance of other CSEF steels. However, the position recommended by EPRI is clear. In all cases, purchase of CSEF steel components should recognize the complexity of the metallurgy, the need for care to achieve excellent creep performance without excessive variability. Thus, well controlled composition and fabrication should be mandated.

**Table 1.** Composition for Controlled Quality Grade 91 steel specified in ASME Code Case 2864 [16].


#### **3. Weld Manufacture and HAZ Microstructural Characterization**

Proper characterization of the microstructure in metallurgically complex steels is complicated by the fact that most of the features which define creep strength and ductility cannot be resolved or studied using optical metallography. Moreover, the diversity of thermal cycles experienced by multi-pass fusion welds further complicates meaningful characterization because without care there is a very significant spatial variation throughout the weld and HAZ.

The preferred approach to overcome the problems of relevant examination and recording meaningful information, it is usually necessary to balance the results from macro-, micro- and nanoevaluation with appropriate analysis. Thus, it is important to take an overall view to microstructure and then increase detail based on the information recorded. In view of the challenges associated with nano level examination this level of detail can only be performed selectively. It is thus vital to make sure that the methods used for this selection are rigorous.

This section summarizes information regarding EPRI recommended approaches [17] for this characterization for welds manufactured in CSEF steels. It should be emphasized that in all cases the procedures used have been validated and checked against calibrated sections.

#### *3.1. Weld Manufacture*

EPRI has published a series of documents which describe procedures for weld manufacture. The outline below is included here because it is critical to research projects that the base sections used is well pedigreed and the subsequent welding is performed in a controlled manner. Lack of control of fusion welding will lead to a very wide range of local temperature cycles and therefore microstructures in any weld.

In thick section components such as pipes EPRI has typically used a machined U-groove with a 15◦ bevel and using best practice guidance for the shielded metal arc welding (SMAW) process as detailed previously [10]. For Grade 91 steels the process included a minimum preheat temperature of 150 ◦C (300 ◦F), a maximum interpass temperature of 315 ◦C (600 ◦F), stringer beads only, and removal of slag after each weld layer through light grinding. The filler material used to make the weldments was consistent with an American Welding Society (AWS) type E9015-B9 filler material. Stipulating stringer beads without weaving and using only 3.2 mm (0.125 inches) diameter electrodes limited the variability in the heat input. The completed weldment including a macro sample, documented fill sequence, and the recorded data for amperage, voltage, travel speed, and interpass temperature are provided elsewhere [10].

Following welding, the weldment was allowed to cool to room temperature. In some cases, EPRI research into weld repair investigated the performance of welds without post weld heat treatment (PWHT). In other cases, and for the weld shown in Figure 5, a relatively low temperature of PWHT, namely at 675 ◦C (1250 ◦F) held for 2 h, was used. Full details of the development and testing of welds made using alternative weld and PWHT procedures have been reported [10].

**Figure 5.** (**A**) Macro Sample of the As-fabricated Weldment in the Post Weld Heat Treated Condition. (675 ◦C, 1250 ◦F for 2 h) as shown the pipe thickness is 50 mm; (**B**) Fill Sequence used to complete the Weldment. Note that the darkened fill passes constitute a fill pass that was monitored for voltage, amperage, travel speed and interpass; (**C**) Details for the Monitored Fill Passes.

The weld macro section shown in Figure 5A demonstrates that the appearance of the weld was consistent with the expected bead sequence. This is shown in Figure 5B. Thus, it was clear that the requirements of the procedure had been followed. No obvious regions of inhomogeneity in the weld macrostructure were identified. Because of the limitations of optical metallography regarding defining microstructural differences, to provide visual images of the macroscopic bead size and shape and the overall pattern of structure EPRI has developed procedures for macro-analysis which includes hardness mapping [17]. The equipment utilized for the hardness mapping characterization was a LECO Automatic Hardness Tester, Model AMH-43. Hardness mapping was conducted so that the requirements in both ASTM E384-11 (ASTM 2011) and ISO 6507 (ISO 2005) were met. One of the key requirements in these two standards is that for a given hardness load (e.g., for this study 0.5 kgf), the indents should be at least 2.5 d apart (where d = mean diagonal distance of the measured Vickers indent in the material being examined).

To ensure that sufficient resolution in the data was obtained, i.e., to achieve enough indents in the heat affected zone, a very large area around the weld fusion boundary was analyzed. Thus, in contrast to simple hardness line scans carried out in some traditional studies, in the present research representative portions of the base metal, heat affected zone (HAZ) and deposited filler metal were captured in the hardness map. This approach resulted in a final hardness map size that was 25 mm × 25 mm and included a total of 10,000 indents. The location examined is shown as the highlighted box in Figure 6A with the hardness map produced shown in Figure 6B.

**Figure 6.** (**A**) Macro Sample of the As-fabricated Weldment in the Post Weld Heat Treated Condition (675 ◦C, 1250 ◦F for 2 h) with the yellow box showing the region examined by hardness mapping; (**B**) The hardness map produced with a scale showing hardness ranges for each colour.

It should be noted that the regions of highest hardness in the weld were recorded at, or close to, the centre of the individual beads. The pattern exhibited by these regions shows the uniformity of the beads deposited. The highest hardness recorded were values that are <350 HV 0.5. This observed maximum value is higher than for typical weldment manufacturing where the hardness values are reduced to ~<300 HV 0.5. The difference is attributed to the specified PWHT namely (675 ◦C for 2 h), which is consistent with the recognized minimum in the National Board Inspection Code Part 3 Repairs and Alterations Supplement 8 (NBIC 2017). This Code provides post construction repair requirements for power generation components and materials including Grade 91 steel and is commonly utilized in North America and in some southern European countries.

It should also be noted that recently ASME B&PV Code Section I reduced the specified minimum PWHT requirements for Grade 91 steel type welds in new construction. The stipulated minimum conditions are 705 ◦C (1300 ◦F) for weld thickness > 13 mm (0.50 inches) and 675 ◦C (1250 ◦F) for weld thickness ≤ 13 mm (0.50 inches) [18].

#### Heat Affected Zone Microstructure

Historically, the microstructures in Low Alloy Steel welds have been classified based on the prior austenite grain size and whether the original precipitates present had tempered. This type of characterization is not technically justified for tempered martensitic steels since properties are controlled by substructure as well as the type and size of precipitates. Recent research has therefore been performed to properly characterize the HAZ regions in 9 wt. % Cr CSEF steels. This research [5,6] involved systematically investigating the microstructural distribution in the HAZ of single pass and multipass welds and performing microstructural simulations under known conditions. The multipass welds capture the accumulated influences from multiple weld thermal cycles.

The advanced characterization techniques used included a Nova 600 Nanolab dual beam (Thermo Fisher Scientific, Hillsboro, OR, USA) focused ion beam field emission gun scanning electron microscope was used to collect the electron backscatter diffraction (EBSD) data from the matrix. Ion-beam-induced secondary electron imaging was used to evaluate the distribution of precipitate particles. EBSD maps were collected using an EDAX Hikari camera (EDAX Inc., Mahwah, NJ, USA), at an accelerating voltage of 20 kV and a nominal beam current of 24 nA.

The output of the characterization made using these advanced techniques is illustrated with reference to the images shown in Figure 7. The observations and subsequent analysis of the results obtained [5,6] demonstrated that a new description for the HAZ regions in martensitic CSEF steels based on the transformation behaviour was required. This description which emphasizes the effect of the welding thermal cycles on the precipitates present is summarized as follows:


**Figure 7.** Ion-beam-induced secondary electron micrographs (**a**–**f**) showing the secondary precipitate particles in the HAZ.

These descriptions more closely match the original regions described in the documentation from ORNL [19] and are justified by the extensive nature of the characterization techniques used [5,6].

#### *3.2. Heat A*ff*ected Zone Damage*

The heterogeneous microstructures in weld HAZ's are such that during creep, multiaxial stresses can develop in large cross weld specimens under uniaxial loading. These multiaxial stresses are established because the large samples constrain the deformation until cracking occurs. Metallographic examination of feature test samples has shown that creep damage in the HAZ is a function of the metallurgical risk inherent in the base steel and the local stress state. In the HAZ the welding thermal cycles modify the parent microstructure. The local stress state established during creep is influenced by the weld geometry, orientation and the properties of the specific microstructural zones.

After the creep failure or termination of the creep test, a macro sample was removed from the post-test feature creep test using fine wire, electrostatic discharge machining (EDM). All specimens were removed from the approximate center (the line in the top image in Figure 8) to analyze the most representative distribution of damage well-controlled, feature-type cross-weld creep tests.

**Figure 8.** Example of Feature Creep Tests Used in the Evaluation of Damage in the Heat Affected Zone of Grade 91 Steel. Note: the feature test sample contains the entirely of the weld, the HAZ on both sides of the weld and sufficient base metal to promote stress states in the weldment.

When these feature specimens are creep tested to the end of life for low stresses, i.e., long times the fracture path ran through the weld HAZ, Figure 9. Fractures of this nature take place with very little ductility or tearing, and the cross sections shows very little or no reduction in area. Fractures of this type are typical of the damage observed in CSEF welds in service.

**Figure 9.** Macro Image from a feature sample of Grade 91 Weldment which was creep tested to failure.

Details of the creep damage developed in the specimens have been establish using specialist metallographic techniques. Information regarding the sample preparation and characterization methods have been published [17]. The present paper therefore only summarizes key information used for the laser metallography. However, even a basic appreciation of these factors emphasizes the level of effort invested to obtaining relevant and accurate results. It is apparent that the usefulness of the data is dependent upon careful preparation, observation, recording of data and then analysis. Other researchers are encouraged to comment on specific techniques used in studies documenting the microstructure and creep damage in tempered martensitic steels. The following are the main stages involved with the sample preparation and examination in EPRI research:

• Following the fine wire electro discharge machining the samples were prepared using metallographic preparation procedures which included initial grinding on 240 to 1200 grit SiC and then subsequent polishing on standard cloths with diamond suspension down to 1 μm. A final extended chemo-mechanical polishing procedure was carried out. This was performed using 0.06 μm colloidal silica suspension. This final polish ensured that all evidence of surface deformation that was introduced in the abrasive stage of preparation was eliminated.


An example of the examination approach is shown in the highlighted region of Figure 10. As with all metallographic techniques there is frequently the need to balance being able to resolve specific items in the specimen and obtaining sufficient images to ensure meaningful observations. Very high magnification can aid in the identification of fine detail, but the small fields involved makes capture of sufficient numbers of features difficult. In the present research a 20× objective was established as appropriate for the analysis of creep cavitation in the HAZ of Grade 91 steel welds. This objective provided a magnification of ~400× on a 15-inch monitor [20].

**Figure 10.** (**A**) Macro Image from the Examined Grade 91 Weldment and Highlighted Region of Interest along the fusion line and HAZ of the Weld; (**B**) Magnified View of Creep Damage in the HAZ of the Weldment.

The magnification is not the only critical variable in the assessment of creep cavitation since the number of pixels in the obtained image can also be altered. The default size for each image collected is 1024 × 768 pixels. The complied macro image can be saved using the full resolution. The number of pixels used for analysis of creep cavitation was 6346 × 3303 pixels. The image size (i.e., the number of pixels) is an important factor in determining the threshold for counting cavities in any material.

The damage developed in cross weld feature testing of a Grade 91 steel which is inherently damage susceptible is illustrated in Figure 11. The creep voids present are shown as dark contrast in Figure 11a. For this feature test, it is apparent that even though high number densities of voids have developed in the weld HAZ there are no cracks present. Thus, for structures which have a uniform susceptibility to void formation, and a similar stress across the section, only very limited periods of stable creep crack growth can take place. The specific region within the HAZ where the highest number of voids was developed is shown in Figure 11b. Based on the detailed analysis performed, it appears that the highest damage levels are found in the region which experienced a thermal cycle in

the range 1050 to 920 ◦C during welding. Whereas, traditional descriptions have termed this as 'fine grained' or 'intercritical', recent EPRI research has established that this region should more correctly be defined as partially transformed [5,7].

**Figure 11.** Macrograph showing a Grade 92 steel cross weld feature test sample after long term testing, note using the EPRI classification system the steel parent is considered as 'susceptible' to void formation (**a**) and, histogram showing the number of voids present as a function of the distance from the weld fusion line (**b**). The basic EPRI procedures used for sample preparation, data capture and analysis have been detailed previously [17].

#### **4. Continuum Damage Mechanics**

A creep continuum damage mechanics (CDM) constitutive model for Grade 91 steel developed as part of a multi-year collaborative research project has been used to assess the creep performance of structures and components. As described in previous papers [9,20], EPRI research in the area of CDM has been guided by a systematic, step wise methodology. The following list summarizes the important steps followed during the development and testing of this model, and its ability to describe high temperature behaviour:


The broad scope of these tasks requires an integrated and sustained approach implemented and reviewed over several years. The basis for this research has been described in an EPRI published report [9]. The overall plan required that there would be key points over the course of the model development and validation where the details of the work were reviewed and further evaluated. At each of these review points, important knowledge gaps or needs, such as the requirement for additional metallurgical data, were identified and the necessary testing and analysis undertaken.

Using the logical approach outlined above the expression for the CDM was formulated recognizing that the creep deformation and fracture behavior was different at high stresses and at low stresses. The relationship used to describe the stress dependence of the minimum creep rate involved summing two power laws. This overall relationship is shown in Equation (1). This considers the two different high temperature damage mechanisms, namely strain softening and cavitation.

$$
\dot{\varepsilon}\_{\rm min} = A\_{\rm HT} \sigma^{\prime \prime H} + A\_{\rm MT} \sigma^{\prime \prime\_{\rm m}} \tag{1}
$$

where

$$A\_{HT} = A\_H \exp\left(-\mathbf{Q}\_H/\mathbf{RT}\right) \tag{2}$$

and

$$A\_{\rm MT} = A\_{\rm M} \exp(-\mathbf{Q}\_{\rm M}/\mathbf{RT})\tag{3}$$

In these equations, . ε*min* is the minimum creep rate, *R* is the ideal gas constant, *Q* represents the Activation Energy and *T* is the temperature in Kelvin.

The damage and strain softening state variable are incorporated into the strain rate equation as shown in Equation (4). Damage as a result of strain softening was defined by the state variable G, which is 0 in the initial condition and when ductile failure occurs. The softening rate is proportional to the strain rate, see Equation (5). Cavitation damage was described associated with the formation and growth of creep cavities. Following the work of Kachanov [21], cavitation was represented by the state variable, ω, and this also varies from 0 to 1. The classical damage rate equation with (1 − ω) is used to describe the relationship between the rate of cavitation and the value of the maximum principal stress, see Equation (6).

$$\dot{\varepsilon}\_{\dot{\varepsilon}\dot{\jmath}}^{\mathcal{L}} = \frac{\mathbf{3} \mathbf{s}\_{\dot{\imath}\dot{\jmath}}}{\mathbf{2} \sigma\_{\varepsilon}} \Big| A\_{\text{HT}} \sigma\_{\varepsilon}^{\prime \mathcal{U}\mathbf{H}} + A\_{\text{MT}} \Big( \frac{\sigma\_{\varepsilon}}{(\mathbf{1} - \boldsymbol{\omega})(\mathbf{1} - \boldsymbol{G})} \Big)^{\mathcal{U}\mathbf{M}} \Big| \tag{4}$$

$$
\dot{G} = k \dot{\varepsilon}\_e^c \tag{5}
$$

$$
\dot{\boldsymbol{\omega}} = \frac{A\_{\rm DT} \sigma\_I^{\rm x}}{\left(\mathbf{1} - \boldsymbol{\omega}\right)^{\rm \varphi}} \tag{6}
$$

In these equations . <sup>ω</sup> is the rate of damage due to cavitation,σ*<sup>I</sup>* is the maximum principal stress, . *G* is the strain softening rate, . ε *c <sup>e</sup>* is the equivalent creep strain rate, . ε *c ij* is the creep strain tensor, *sij* is the deviatoric stress tensor, and σ*<sup>e</sup>* is the von Mises equivalent stress.

Extensive analysis of the Grade 91 heat affected zone microstructure by EPRI has identified distinct microstructural regions in the HAZ. The first, termed the completely transformed zone (CTZ), is located adjacent to the weld fusion line. This zone has historically been referred to as the coarse-grained heat-affected zone is exposed to the highest temperatures during the welding process and the microstructure which develops in the CTZ is similar to the microstructure observed in the base metal. For this reason, in CDM analyses of cross-weld behavior, material properties in this zone have been assumed to be the same as those of the base metal. The second zone is termed the partially transformed zone (PTZ). The temperatures reached in this region are less than those achieved in the completely transformed zone and a desired dislocation and precipitate substructure is not present. The PTZ is associated with lower strength and greater susceptibility to creep cavitation than in Grade 91 base metal or the CTZ.

The structure of the constitutive model for the partially transformed zone is the same as that of the Grade 91 base metal model. In order to represent the difference in strength between the partially transformed zone and the base metal, a strength factor *F* was introduced, where 0 < *F* ≤ 1. When *F* = 1, base metal strength is achieved; when *F* < 1, the creep strength is less than that of base metal:

$$\dot{\varepsilon} = A\_{HT} \left( \frac{\sigma}{F} \right)^{n\_H} + A\_{MT} \left( \frac{\sigma}{F(1 - G)(1 - \omega)} \right)^{n\_M} \tag{7}$$

The strength difference observed between the partially transformed zone and the base metal is described using the factor, *F*. Since *F* is a function of temperature an Arrhenius temperature dependence was incorporated as follows:

$$F = \mathbf{1} - F\_0 \mathbf{exp}(-\mathbf{Q}\_F/\mathbf{RT})\tag{8}$$

The stress redistribution leads to enhancement of the maximum principal stress in the PTZ. This was considered when determining material parameters for the damage mechanism.

Determining material property coefficients for the PTZ model requires data from a minimum of three test types (each performed over a range of temperatures and stresses): uniaxial smooth bar tests of the base metal to define base metal creep strength; uniaxial smooth bar tests of simulated PTZ material to define the creep strength of the PTZ; and a test imposing triaxial constraint on the PTZ or simulated PTZ material to determine the cavitation damage response. At present, cross-weld tests have been used to define the damage response of the PTZ. To ensure the coherence and relevance of the developed data, all the proceeding tests should be performed on material from the same heat.

Testing according to the above methodology was performed by Hongo et al. [22], representing a coherent and relevant dataset that could be used to determine material parameters for the partially transformed zone (note that the partially transformed zone is referred to as the fine-grained HAZ in [22]). A comparison of the measured and predicted time to rupture for uniaxial creep rupture tests of Grade 91 steel base metal, simulated HAZ (PTZ), and cross-welds is shown in Figure 12. A key item to note is that the model predicts failure by the development of damage in the HAZ for cross-welds tested at low stresses, but failure by creep rupture of the base metal at high stresses. The behavior is thus directly in accordance with the trends noted in tests reported elsewhere [22].

As an example of the application of CDM to the behaviour of a component, the physically-based creep continuum damage constitutive model was applied to the serviceability assessment of a 90◦ large bore, welded branch connection [9]. This connection had developed a steam leak which occurred as the result of in-service cracking. The connection was removed from the component and selected samples were metallographically prepared and examined in detail, Figure 13. The model based predicted time to crack initiation (when), location of crack initiation (where), and direction of subsequent crack growth (how) were shown to agree with the observed trends in the reported failure. The agreement between observation and analysis underpins confidence that cracking initiated at the surface. This indicates that nondestructive examination techniques such as surface metallurgical replication and magnetic particle testing can be used to detect damage. However, some caution is required since the

operating time between initiation and through-wall cracking can be short. For the exemplar component a leak-break-before break assessment was performed to provide an estimate of damage tolerance, i.e., an estimate of how long the crack would take from initiation to through wall growth. This further analysis demonstrated that the crack growth rate was relatively slow and there was a very low risk of the connection to fracture catastrophically. The expectation for a leak type failure was based on the fact that crack extension of greater than 90◦ around the branch pipe would be necessary for catastrophic rupture of the connection to occur. The prediction of leak not break is consistent with experience of cracks in similar components. Damage for the typical component geometries is found by steam leaks, not catastrophic rupture.

**Figure 12.** Comparison of measured and model predicted time to rupture for Grade 91 base metal, simulated HAZ, and cross-welds all tested at 600 ◦C and under uniaxial loading. The data shown in this Figure were published previously [22], the lines shown are as estimated from the CDM analysis.

**Figure 13.** Metallographic section showing in-service cracking at a welded tee piece connection. The scale shown in graduated in inches.

#### **5. Advanced Weld Repair Technologies**

EPRI has a long-established reputation in the development and validation of weld repair methodologies. Research supporting fossil-fired asset management has developed viable weld repair techniques for mainstay power generation CrMo steels which have been widely used for coal-fired boiler or combined cycle heat recovery steam generator (HRSG) components. A series of reports provide research results and document weld repair procedures specific to CrMo steel Grades 11, 12 and 22 [23]. It is clear the knowledge base published provides an important strand for cost-effective life

management, i.e., informed run/repair/replace decision-making. The research performed through the EPRI studies, in part, led to the acceptance of Welding Method 4 and Welding Method 5 in the NBIC Part 3 Repairs and Alterations. These weld repair methods are often cited as 'temperbead methods'. Although the primary consideration for CrMo and CrMoV repair is the avoidance of reheat cracking through the concept of 'grain refinement' rather than simply tempering the microstructure.

EPRI-coordinated, industry-sponsored research projects in CSEF steels began in 2007 with a major effort to establish sound life management approaches for Grade 91 steel components. The learnings and findings, realized with direct input and perspective from over 40 stakeholders representing the entire electricity supply chain, underpin component specific repair methodologies for the CSEF steel Grade 91 in use today. Key evidence for these 'alternative' repair methods are summarized below.

#### *5.1. Development of Alternative Repair Methodologies*

Historically, it has often been the case that weld repairs performed after periods of in-service operation were based on the procedures mandated for new construction. Clearly since the equipment is not new it is important to consider whether this approach is reasonable. A major EPRI initiative investigating and quantifying the factors affecting weld repair of CSEF steels started in 2011. After this initial thrust, a series of research programs has been undertaken to develop well-engineered weld repair procedures, initially for Grade 91 steel components [10]. The variables studied in these studies are summarized in Table 2. It is important to emphasize that this research resulted in acceptance of weld repair procedures which could be applied without the need for PWHT. The development of weld repair methodologies has been largely proactive, so that accepted procedures were established before widespread failures were reported. Today, the research continues to evolve the repair methods for dissimilar metal welds to Grade 91 steel and for the next generation of 9 to 12%Cr CSEF steels such as Grade 92.


**Table 2.** Variables Studied in the Examination of Well-Engineered Repair Methods in Grade 91 Steels.

Note: Welds made using Controlled Fill were NOT heat treated prior to testing.

The power generation industry is now recording operating times in boilers and piping systems using Grade 91 steel which exceed 100,000 h. The likelihood for damage increases with increasing operating time, so that as time in service increases so does the need for more regular and widespread repairs. More than two dozen EPRI reports underpin the recommended repair practices now recognized by the NBIC Part 3 Repairs and Alterations as Welding Method 6 (first published in 2015 edition) and Welding Supplement 8 (first published in 2017 edition). These methods include options without a mandatory post weld heat treatment (PWHT). Examples of publicly available information can be found EPRI documents for example [10,24,25].

The design and implementation of well-engineered weld repairs in Grade 91 steel components required a number of considerations be taken into account to provide a best practice approach. It was apparent that there is no one-size-fits-all procedure for all alternative weld repairs. The critical decisions and recommendations that were reviewed during the code approval process for Welding Method 6 and Welding Supplement 8 have been explained in a key document. Some of the important questions raised and answered during the code approval process have been documented [26]. These summaries concerned the following issues on performance:


The knowledge base supporting these repair methods will be updated periodically as needed to address questions specific to new or emerging issues associated with alternative weld repairs in Grade 91 steel.

#### *5.2. Validation of Alternative Repair Methodologies*

Although significant information was produced to validate the alternative repair methods during the first phases of research on-going studies include refinement of performance expectations for the weld repairs. These further approaches for quantification of behavior include the use of state-of-the-art testing and evaluation methods to provide a comprehensive basis for informed decision making. These independent approaches include:


A typical feature test cross weld creep specimen in shown in Figure 14A. This specific example shows a weld which simulates the manufacture of repair in a thick section joint between Grade 91 and Grade 22 steels. The width of the repair is purposely designed to extend beyond the HAZ of the original weld profile. This extended fill creates a step in the repair profile, and this has been shown to increase the damage tolerance of the joint. As shown in Figure 14B, creep damage which develops in the HAZ of the repair cannot easily propagate because the base metal exhibits greater resistance to creep damage than the HAZ. This is a very important benefit because in the majority of cases weld repair is required after removal of damage, i.e., the location needing to be repaired is known to be susceptible to forming damage. In the current case, the weld HAZ has developed stable micro and even macro defects which should be readily detected using recommended nondestructive testing methods.

**Figure 14.** Example of a feature test cross weld specimens used by EPRI, (**A**) containing a simulated weld repair and (**B**) showing HAZ creep damage in a dissimilar metal repair weld between Grade 91 and Grade 22 Steels.

**Figure 15.** Example of a full-size component vessel test containing innovative weld repairs in Grade 91 Steel performed without post weld heat treatment.

An example of one of the EPRI designed vessel creep tests is given in Figure 15. This vessel was manufactured from a cylindrical section of Grade 91 steel taken from a superheater outlet header. The vessel contains several features representative of the geometric complications found in power and petro-chemical plant. The features included in the vessel were a flat end cap and an end plug, a longitudinal seam weld and stub tube to header welds. In the original pressure boundary circumferential weld, a partial weld repair was made using one the newly approved repair approaches. The repaired region is shown in the schematic section of Figure 15. As previously the excavation and subsequent fill extend beyond the HAZ of the original weld. This vessel has been on test at 625 ◦C and the results will be reported in detail in due course.

**Figure 16.** Evaluation of vessel test behavior using finite element analysis coupled with physically-informed constitutive models for creep damage initiation and crack growth, (**A**) showing local stress concentrations at the end caps and (**B**) showing damage estimates for the weldments.

The vessel type creep tests offer very valuable direct evidence of high temperature performance both in terms of the time to cracking, the locations of first cracking and cracks growth leading to failure. As these tests are undertaken in purpose-built facilities it is possible to run tests to fracture as all risks are minimized. It is, however, not possible to perform these specialist experiments on all combinations of pressure, temperature and vessel design. Thus, in addition to extracting maximum value from the direct experimental observations further benefits are derived if the actual results from the vessels are assessed using the CDM analysis methods described earlier. The EPRI recommended approach has been to seek to benchmark this type of analytical assessment using results from in-service Case studies and vessel data. Figure 16A shows the results of stresses developed in the vessel as a consequence of internal pressure. Details of damage analysis for the HAZ of pressure boundary welds are presented in Figure 16B.

A further important method for evaluation of performance involves collating and analysis of information regarding the behaviour of actual repairs in installed components. Development of this database permits analysis of the performance of the overall performance of repairs and more detailed assessment by component. In the case of Grade 91 steel, EPRI has documented thousands of repairs over the last five years in multiple counties which were fabricated in accordance with the NBIC and EPRI reports. Tracking these cases is vital for the industry to provide best practices should issues arise, and such that the continued success of specific approaches is supported by statistical relevant information.

Future work is necessitated to evaluate repair options for other CSEF steels allowed by Design Codes. These include Grade 92, and a new generation of emerging CSEF steels such as VM12SHC (ASME Code Case 2781), Grade 93 (ASME Code Case 2839) and THOR 115 (ASME Code Case 2890). Furthermore, common place failures in traditional and A-ASS materials in tube to attachment welds is necessitate the development of novel test techniques to screen material performance and provide insightful guidance for remediation using weld repair.

#### **6. Discussion**

#### *6.1. Creep and Fracture of Welds*

Assessment of the creep performance of welds is frequently studied using cross weld testing. Critical issues in setting up these programmes include selecting test conditions and specimen geometries that result in damage mechanisms which are relevant to long term service. It is now widely accepted that the results of creep tests at relatively high stress and temperature are not typical of long-term damage in component welds. Thus, information from tests under non-representative conditions should

not be used as a guide to in-service behaviour. EPRI has established a testing protocol which involves performing cross weld creep tests on feature test type specimens. These sample examined in this study has a cross sectional area in the gauge of ~965 mm2 (1.5 inch2). This is about 30<sup>×</sup> greater than typical cross-sectional areas for standard round bar creep tests machined to a diameter of 6.35 mm (0.252 inches).

The characterization of damage in these cross weld tested samples typically involves [17,27] visual examination followed by optical metallography. The microscopy was carried out on sections which were prepared using careful polish/etch techniques to reveal damage present on the central plane of each specimen. This approach has been shown to be necessary since the tendency to form service relevant HAZ creep damage is related to hydrostatic stresses and is thus minimized at the specimen surfaces.

Experience has shown that as a rule the welds present in the pressure boundary of CSEF steel components are susceptible to creep damage and fracture before base metal locations. Observations of this type have therefore focused research to evaluate the factors affect weldment performance. It is historically the case that research of this type involves undertaking laboratory creep tests of specimens cut from the selected welds. In most cases decisions regarding the geometry of the samples are made considering the size of the welded section available and the load capacity of the available creep test machines. In most cases then the specimen geometry chosen is that of a cylindrical section with a diameter in the range 4 to 10 mm. EPRI has established a testing protocol which involves performing cross weld creep tests on feature test type specimens. The types of specimen involved in EPRI research assessing the creep behaviour of welds are shown in Figure 17. These samples typically have a cross sectional area in the gauge of ~965 mm2 (1.5 inch2). This is about 30<sup>×</sup> greater than typical cross-sectional areas for standard round bar creep tests machined to a diameter of 6.35 mm (0.252 inches). Figure 17 compares the feature type specimens with a small cylindrical sample of the type used in other testing.

**Figure 17.** Examples of the feature test cross weld specimens used by EPRI to evaluate the factors affecting the creep behaviour of pressure boundary joints in tempered martensitic steels.

In addition to selection of a meaningful specimen geometry it is also important to consider the test conditions since all of these factors will influence whether damage mechanisms in the laboratory tests are relevant to long term service. Experience has shown that the results of cross weld creep tests which are performed at relatively high stress and temperature are not typical of long-term damage in component welds. Thus, information from tests under non-representative conditions should not be used as a guide to in-service behaviour. It is critical then that the details of the fracture location and relationship of the damage present with the constituent microstructure of cross weld tests is established using accurate posttest metallographic preparation and examination. This type of characterization provides important knowledge as to how to use cross weld creep test results.

Details of the post-test characterization methods recommended by EPRI regarding the sample preparation and characterization methods have been published previously [17]. It is clear that to accurately document the creep damage developed in the specimens requires the use of specialist metallographic techniques. The information summarized in the present paper provides an important appreciation of the level of effort invested to obtaining relevant and accurate results using laser metallography. It should be emphasized that the usefulness of the data is dependent upon careful preparation, observation, recording of results and then analysis. This process must start by ensuring that the sectioning is performed to reveal damage present on the central plane of each specimen. This is necessary since the tendency to form service relevant HAZ creep damage is related to hydrostatic stresses and is these hydrostatic stresses are minimized at the specimen surfaces. In addition to the need for care during sectioning and specimen preparation, it is important when documenting creep cavity size and shape that no etching of the specimen is undertaken. It is clear that etching will modify both the size and character of cavities. In general, etching will increase the size of cavities present and will tend to 'round' the cavity shape. While the use of laser microscopy provides a reasonable record of the number density of creep voids it is clear that the most accurate observations of creep void shape are obtained using ion beam techniques [17,27]. In this advanced research, the cavities are shown to be angular in shape and usually associated with the inclusions or other hard particles present.

In addition to care in preparation, examination and recording of the creep cavities present effort should also be invested in careful selection of both magnification and image resolution. Earlier research [17] has shown that for images recorded from the same specimen using the same magnification, the cavitation densities shown were markedly different as a consequence of image resolution. Thus, for the lower resolution used, the theoretical minimum cavity diameter that was counted was 2.04 μm. In contrast, using an improved image resolution, improved the minimum cavity diameter that was counted to 0.70 μm. Because the digital processing of images using light microscopes is becoming increasingly common, it is critical that that appropriate settings and procedures are selected. In the case of the equipment used in EPRI studies, there are options for utilizing "super fine" image capturing that would further increase the resolution of the selected objective to 0.34 μm. The difference in cavity size recorded between the low resolution and "super fine" resolution represents a range of almost an order of magnitude.

The parent metal condition has a direct effect on the cross-weld creep performance in Grade 91 steel. This is most often put into the context of a deformation-related mechanism (i.e., strength), but as presented in this review paper there is a clear influence of the damage resistance (i.e., ductility) on cross-weld creep performance. Thus, the results from EPRI research demand that designers and alloy developers place an equal emphasis on parent metal creep strength and ductility. Inherent base metal performance is a key aspect linked to the susceptibility to damage in the HAZ. Detailed testing supported by comprehensive metallurgical characterization has established the factors which influence cross-weld creep performance. For a set of Grade 91 materials which exhibited the same strength, an increase in the ductility from 15% to 83% ROA resulted in increased the cross-weld creep life by 5× at a test condition of 625 ◦C (1157 ◦F) and 60 MPa (8.7 ksi). Ongoing assessment is working to more definitively link risk factors in the microstructure which are responsible for damage (i.e., inclusions and/or other cavitation-susceptible features in the material). However, it is apparent that although weld performance maybe life limiting for in-service components a significant contribution effecting the life and manner of creep fracture can be directly attributed to the base steel. As shown in Figure 18, the creep performance of cross weld tests and of simulated HAZ microstructures are very similar under low stress long–time conditions.

The effect of filler metal strength on weldment performance has also been included in EPRI research. Cross weld creep tests on samples made using a weld filler metal with an over-matching strength to the base metal did not show any trend on life. For example, there was no significant influence on life for creep tests performed at an applied stress of 60 MPa (8.7 ksi). In the tests at a higher applied stress of 80 MPa (11.6 ksi), there was some evidence of reduced creep life as filler metal strength increased but the results of all tests were still above the lower bound of acceptable performance, i.e., the test duration was greater than the mean-20% bound. In contrast, cross-weld testing of samples made with under-matching filler material consistently failed at the mean-20% bound, Figure 18. It appears

that the fact that welds with lower creep strength filler show creep failure in the HAZ can be explained by the fact that the B8 type filler material used was close to the strength of the PTZ region of the HAZ. It was also of potential practical benefit to note that the cross-weld tests for the under-matching filler exhibited the greatest damage tolerance of all the weld metals studied. Thus, in these tests there appeared to be evidence that macro-cracking started at an earlier life fraction than in tests with the higher strength filler metals. It is clear that a relatively slow rate of crack propagation increases the opportunity for defect detection by non-destructive testing during service. It is apparent therefore that designers should consider the geometry of the weld and the specific properties of the filler metal on component creep performance. Ensuring damage tolerance is a key benefit to component design for long-term operation under high temperature creep conditions.

**Figure 18.** Relationship showing the creep performance of the HAZ defines the properties of cross weld specimens in long times.

#### *6.2. Management of Assets Made from CSEF Steels*

The life management approach at the fleet-, plant-, system- or component-level is conventionally divided into a staged approach [24]. The Level 0, I, II or III methodology incorporates important concepts such as risk-ranking or risk-based inspection (RBI) and fitness for service (FFS). The level approach is summarized in the following information:


Fundamentally, this level of detail helps to refine prior recommendations and assess 'how' the component will fail, e.g., leak or break. The end user must be involved in performing a meaningful risk assessment, particularly to ensure that the consequences of fracture are properly established and factored into the risk calculation.

As a minimum EPRI recommends that defining the consequence of the event should consider the following factors in relation to component failure:


Risk-ranking can be performed to a progressively more detailed level of probabilistic and consequential assessment which typically requires more certain engineering factors (e.g., stress/temperature/metallurgy/dimensional measurements). Basic concepts which should be considered in the risk-ranking of 9%Cr steel components are covered in more detail in [10,12,28,29]. The general approach to risk-ranking is defined in ASME Post Construction Code 3 (PCC-3) with a more thorough approach defined in EN 16991:2018 [29]. Initially, effective asset management should target components which have a high expectancy to fail early-in-life. This decision-making is predicated on identification of the factors which contribute to reduced performance (e.g., design, fabrication, operation and metallurgy). In many cases, a detailed understanding of the historical issues identified through available service experience reports, such as in [13,14], are of immense value to the engineer to provide needed perspective and begin to target the highest risk locations

#### **7. Concluding Comments**

Components in power generating plants must operate safely under complex conditions, including a high temperature, high pressure and severe environments. New or post construction codes provide a basic set of rules to ensure that a minimum expected performance will be achieved. Fabrication of power generation components necessitates joining sections to establish a suitable pressure boundary through conventional fusion welding processes. Throughout the course of a component's life, damage at the most susceptible, high-risk locations will necessitate run/repair/replace decision-making as part of an integrated life management philosophy.

The present review emphasizes that the high temperature performance of tempered martensitic steels is complex. However, by appropriate planning and introducing good practice procedures, significant benefit to 'through life' costs can be achieved. These savings can be realized without having to compromise equipment safety and reliability.

**Author Contributions:** J.P. and J.S. contributed equally to Conceptualization of the work; establishing appropriate Methodologies for the research; Validation, of methods and Preparation of the manuscript. Submission of the manuscript and final Editing of the proofs was performed by J.P.

**Funding:** The original research described in this review paper was funded through the support to EPRI from Stakeholders in the Electricity Supply Industry.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Review* **Welding Joints in High Entropy Alloys: A Short-Review on Recent Trends**

#### **Fabio C. Garcia Filho \* and Sergio N. Monteiro**

Department of Materials Science, Military Institute of Engineering—IME, Praça General Tibúrcio 80, 22290-270 Rio de Janeiro/RJ, Brazil; snevesmonteiro@gmail.com

**\*** Correspondence: fabiogarciafilho@gmail.com

Received: 13 February 2020; Accepted: 16 March 2020; Published: 20 March 2020

**Abstract:** High entropy alloys (HEAs) emerged in the beginning of XXI century as novel materials to "keep-an-eye-on". In fact, nowadays, 16 years after they were first mentioned, a lot of research has been done regarding the properties, microstructure, and production techniques for the HEAs. Moreover, outstanding properties and possibilities have been reported for such alloys. However, a way of jointing these materials should be considered in order to make such materials suitable for engineering applications. Welding is one of the most common ways of jointing materials for engineering applications. Nevertheless, few studies concerns on efforts of welding HEAs. Therefore, it is mandatory to increase the investigation regarding the weldability of HEAs. This work aims to present a short review about what have been done in recent years, and what are the most common welding techniques that are used for HEAs. It also explores what are the measured properties of welded HEAs as well as what are the main challenges that researchers have been facing. Finally, it gives a future perspective for this research field.

**Keywords:** high entropy alloys; welding techniques; welding zone microstructure; welding joint properties

#### **1. Introduction**

High entropy alloys (HEAs) emerged in the beginning of this century as promising materials for engineering applications. Their unique compositions, microstructure, and properties are the most attractive characteristics and an incredible number of papers have been published regarding this subject. According to the Scopus database, publications include more than 7,000 research papers, 300 review articles, and eight books over the 16 years, since these alloys were firstly reported by Cantor's [1] and Yeh's research groups [2–6]. One should be wondering what make these alloys so special. In fact, the multi-main element in contrast to the traditional just one or two main elements that mankind have been using in the development of metallic alloys over the centuries leads to interesting discoveries. Multi-main elements would be responsible for increasing the configurational entropy of the alloy and, therefore, favoring a single phase material instead of a microstructure with several phases [6]. This incredible microstructure led to the remarkable mechanical and functional properties reported for these alloys [7–13]. Gludovatz et al. [7] studied the fracture resistance of HEAs. The authors reported that these HEAs were able to combine high fracture toughness, above 102 MPa.m1/2, with high yield strength, above 1.1 GPa. Moreover, these HEAs exhibit the highest relation between strength and ductility among all materials when compared to other class of materials as stainless steels, low alloy steel, nickel-based super alloys, metallic glasses, polymers, and ceramics, even at cryogenic temperature. This is a notable characteristic of these alloys that encourages their use in engineering applications. Nevertheless, it is critical to understand the behavior of these alloys during the processing and fabrication in order to achieve practical use.

Welding is a fabrication technology that is used in a variety of industries. Moreover, it is certainly the most used jointing techniques for metallic materials. Welding techniques could be divided into solid and liquid state processes. Liquid-state welding relies on the fusion of the metal to make the weld. Techniques that are based on gas oxyacetylene, shield metal arc (SMAW), gas-tungsten arc (GTAW), gas-metal arc (GMAW), electroslag (ESW), and others, as well as high-energy beam such as electron beam (EBW) and laser beam (LBW) welding are the most commonly used liquid-state ones [14]. Solid-state welding is defined as a joining process without any liquid/vapor phase formation and with the use of pressure. Friction (FW) and friction stir (FSW) welding are the some notable techniques in such class [15–18]. It is obvious that each welding technique presents significant differences in terms of joint preparation and sample thickness, as well as speed and energy input into the welding joint. For instance, the heat density of a GMAW technique is around 105 W/cm2, while, for the EBW, this density can reach 108 W/cm2 [19]. Such differences will directly impact the quality and properties of the welded joint [20,21].

Few works actually combined both subjects despite the aforementioned importance of the HEAs and the commonly used welding techniques for jointing metallic materials. When comparing the number of publications regarding HEAs properties or characterization and those that somehow discussed the welding in these alloys, less than 0.5% indeed concerned it. The behavior of a novel alloy during welding is considered to be a key technological issue. HEAs are no exception, and the behavior of the alloy when exposed to a welding thermal cycle needs to be explored and understood. The investigation of the alloy can be welded or joined without degradation, whether that is detrimental to the weldment microstructure or properties (during/after welding) and for the duration of intended service, is mandatory for its potential use as an engineering material for structural applications [22]. In this context, this short-review objective is to cover what has been reported on HEAs welded so far, including the influence of the welding techniques and the parameters in the material's properties and microstructure, as well as to critically assess the achievements reported and outlook what of is yet to come in future years.

#### **2. Welding Techniques in HEAs**

Gas tungsten arc welding (GTAW) is a well-known technique among the arc welding techniques for jointing metallic materials, especially high alloy materials [14]. In this process, a non-consumable tungsten electrode is used to produce the welded joint by melting the base metal. In addition, the weld area is protected from atmospheric contamination by an inert shielding gas, such as argon or helium. In one of the first papers regarding the welding of HEAs, Sokkalingam et al. [23] used the GTAW technique to weld an Al0.5CoCrFeNi alloy. The parameters of 40 A for current, 12 V for voltage, and 80 mm min-1 for weld velocity were chosen to joint 2.5 mm HEA plates. The main achievement of this work was a remarkable refinement of the grains from 60 μm in the base metal to a range of 8–12 μm grain size in the fusion zone. Nevertheless, a reduction of approximately 6.4 and 16.5 % were reported for strength and ductility, respectively in comparison with the base metal. Figure 1 displays the microstructure of the welded zones as well as the measured tensile properties of that work. Similarly, Wu et al [24] welded a CoCrFeMnNi alloy while using GTAW. Sheets with 1.6 mm of thickness were butt-welded using a voltage of 8.4 V, current of 75 A, and velocity of 25.4 mm min−<sup>1</sup> as the welding parameters. No cracks or significant microsegregation were produced.

Recently, high-energy beam welding were reported in several publications regarding high entropy alloys weld [25–28]. The high concentrated heat from laser or electron beam create a narrow heat affected zone in the material and allow for high welding rates to be achieved in comparison to arc welding techniques. Chen et al. [25] were able to successfully use high-power solid-state laser to weld CoCrFeMnNi high entropy alloy plates. The laser power was 2 kW, defocusing amount of +2 mm, laser spot moving rate of 1 m min<sup>−</sup>1, and shielding argon gas flow rate of 30–40 L min.-1 were the parameters used in YLS10000 fiber laser (American IPG Company, Oxford, MA, USA to produce the welded joint. When considering the same composition HEA Cantor alloy processed in two different conditions (cast and rolled), Nam et al. [26] used laser beam welding to study the impact of different welding velocities in their microstructures. The butt welding specimens dimensions were 100 <sup>×</sup> 20 <sup>×</sup> 1.5 mm<sup>3</sup> and Nd:YAG laser power of 3.5 kW, beam diameter of 300 μm, and focal legth of 304 mm, with no shielding gas was used for welding. The welding velocity was in the range of 6–10 m min<sup>−</sup>1. Figure 2 exhibits the result of such investigation. The authors reported the good weldability of the CoCrFeMnNi alloy with no macro-defects, such as internal pores and cracks under all welding conditions. Furthermore, shrinkage voids were observed in the interdendritic region near the center line of the weld metal, and the volume fraction of these voids decreased as the welding velocity increased.

**Figure 1.** Gas-tungsten arc (GTAW) technique to weld an Al0.5CoCrFeNi alloy (**a**) base metal zone (BM) microstructure, (**b**) BM+HAZ+FZ region, (**c**) FZ microstructure, and (**d**) comparison of the mechanical strength of the BM and welded sample. Adapted from [23].

Sokkalingam et al. [27] welded an Al0.5CoCrFeNi HEA with plate thickness of 2.5 mm while using an optic fiber laser source with a beam power and transverse speed of 1.5 kW and 600 mm min−1, respectively. Figure 3 shows that the laser weld resulted in a microstructure of the welding zone (WZ) with a lower degree of Al-Ni segregation in comparison with the base metal zone (BM). This microstructural difference was associated with the lack of diffusion in Al-Ni phase formation at rapid solidification.

**Figure 2.** Comparison of the welding microstructure of the CoCrFeMnNi alloy as cast (left) and rolled (right). Samples welded by LBW in different velocities (**a**) 6 m min<sup>−</sup>1, (**b**) 8 m min<sup>−</sup>1, and (**c**) 10 m min<sup>−</sup>1. Adapted from [26].

**Figure 3.** Comparison of the microstructure of the (**a**) BM and (**b**) WZ of Al0.5CoCrFeNi HEA. Inset of EDX mapping of Al and Ni of both regions. Adapted from [27].

Regarding solid-state welding techniques for HEAs, Zhu et al. [29,30] discussed the possibility of using friction-stir welding (FSW) as welding technique to obtain sound joints in HEAs. Indeed, the authors were able to weld a Co16Cr28Fe28Ni28 and CoCrFeNiAl0.3 high entropy alloys while using this technique. Figure 4a shows how this process runs, the parameters, and the microstructural zones generated through this technique. The welding process was conducted using a load-controlled FSW machine with the WC-Co based welding tool. The welding speeds were 30 mm min−<sup>1</sup> and 50 mm

min<sup>−</sup>1, while the rotation rate and load force were kept constant at 400 rpm and 1500 kg, respectively. The diameter of the shoulder and pin are 12 mm and 4 mm, respectively, with a pin length of 1.8 mm. The authors reported that W-rich particles were detected within the stir zone, which was associated with the friction between the WC-Co based rotating tool and the material, although, in both cases, the weld was considered to be successful. Li et al [31] studied a variant of the Friction-Stir welding by a rotary friction welder (HSMZ-20, Harbin Welding Institute, Harbin, China) to weld an AlCoCrFeNi2.1 alloy. Such a welding technique is a solid-state welding process with some interesting advantages, such as: high welding productivity, low heat input, excellent welding quality, and, unlike FSW, no contamination was reported. Figure 4b illustrates the Rotary Friction Welding (RFW) process, as well as the samples that were produced by Li et al in that work. As for the parameters, the rotation speed was kept constant at 1500 rpm, while the friction pressure varies from 80 to 200 MPa. The authors showed that four different zones could be observed in the microstructure of these materials: base metal (BM), heat-affected zone (HAZ), thermal-mechanically affected zone (TMAZ), and dynamic recrystallization zone (DRZ). In addition, it was disclosed that an increase of the pressure up to 200 MPa notably enhanced the quality of the weld in a way that the tensile test for that condition resulted in the fracture being propagated in the BM zone, unlike the other conditions in which the specimens fractured in the welding zone (WZ).

**Figure 4.** Schematic illustration and produced specimens of (**a**) friction stir (FSW) and (**b**) Rotary Friction Welding (RFW). Adapted from [29–31].

Another interesting welding technique that was recently reported for HEAs is the diffusion bonding [32,33]. As one should expect, in this technique the metallurgical bonding of the alloys depend on the diffusion of the elements from one metallic block to the other. Therefore, two metallic blocks are placed in a vacuum chamber, under a certain pressure, temperature (between 0.6 and 0.8 of the melting temperature), and during a determinate time. The applied parameters depend on the materials that have been welded in this diffusion-base welding method. Lei et al. [32] investigated the dissimilar joint of a single phase face center cubic Al0.85CoCrFeNi alloy and a TiAl intermetallic while using direct diffusion bonding under vacuum. For this work, temperatures in the range from 750–1050 ◦C, holding time of 30–120 min. and constant pressure of 30 MPa, were used to evaluate the weldability of this dissimilar joint. Figure 5a shows the typical interfacial microstructure that was observed for the TiAl/Al0.85CoCrFeNi. One should notice that the emergence of such graded microstructure is directly related to the diffusion velocity of each element of the HEA, as well as the TiAl. The authors suggest four stages for the diffusional bonding of the HEA/TiAl joint. In the first stage, there was physical contact of the base materials with a low degree of atomic diffusion and no reaction layer formed. In the second stage, a large number of Ni and Co atoms diffused into the TiAl. Consequently, it is observed that the formation of α2-phase and the solid solution strengthened γ-TiAl. In the third stage, it is noticed the formation of Ti(Ni, Co)2Al and Cr(Fe, Ni)ss layer. At this stage, the diffusional layers are formed and a reliable metallurgical bond can be observed. Finally in the fourth stage the growth of diffusion layers takes place. This late grow could be associated with the sluggish diffusion character of FCC-structured AlCoCrFeNi HEA. Figure 5b presents the compositional elemental distribution over the graded microstructure marked with yellow points in Figure 5a. The Figure 5b helps to understand the microstructural evolution of the joint and with the appearance of intermetallic phases. Such intermetallics as Cr- rich, Ti3Al, and FeNi phases were held responsible for increasing the hardness in each region, as shown in Figure 5c.

**Figure 5.** Typical insterfacial microstructure of a dissimilar diffusion bonding (**a**), compositional elemental distribution through the graded microstructura (**b**), and (**c**) measured Vickers hardness in the different layers that formed the welded joint. Adapted from [32].

#### **3. Properties of Welded HEAs**

The welding process of HEAs presents direct impact in several properties of such materials. In fact, many of these properties are measured to assess whether the quality of the welding was, indeed, satisfactory. Microstructure might present grain size modification, secondary phases precipitation, segregation, and lattice distortions. Therefore, corrosion, fatigue, and creep service conditions are significantly modified. In addition, mechanical properties, such as tensile strength, ductility, and hardness, are also important parameters to appraise the welding.

Al0.5CoCrFeNi alloy was reported to present better corrosion resistance than 304 stainless steel, but with an increased strength [34]. Sokkalingam et al. [27] studied the corrosion resistance of that welded alloy to verify whether the welding process could deteriorate it. It was observed that the WZ exhibited higher corrosion current density (2.83 <sup>×</sup> 10−<sup>5</sup> mA/cm2) than the BM (8.63 <sup>×</sup> 10−<sup>6</sup> mA/cm2), which showed a higher corrosion rate. The welded joint BM+WZ resulted in the WZ acting like cathode and BM acting as anode, which could result in the secondary phases and particles in the BM been corroded first, as the weldment is exposed to a corrosive environment. Furthermore, deep pit corrosion was observed in the interface between the WZ and the BM zone. This phenomenon was associated with the dissolution of Al-rich particles at the BM zone near the interface. Figure 6a shows the corrosion behavior of such welded joint, while, in Figure 6b, one might observe the deep pit corrosion that occurs near the interface WZ+BM.

**Figure 6.** (**a**) Corrosion behavior and (**b**) deep pits near the interface between WZ and BM of Al0.5CoCrFeNi HEA. Adapted from [27].

In another study regarding the properties of HEAs welded joints, Wu et al. [24] compared the microstructure and mechanical properties of the CoCrFeMnNi alloy welded by GTAW and electron beam welding (EBW). In spite of the low process velocity for the EBW process, 38 mm min<sup>−</sup>1, which was compared to the GTAW process, 25.4 mm min−1, a remarkable difference in the microstructure could be observed in comparison to each case. The GTAW welding zone is at least 2.5 times wider than the EBW welding zone, 3.3 and 1.3 mm respectively which is, obviously, associated with the heat input in each technique. Moreover, both of the welds exhibit yield columnar grains that grows towards the maximum temperature gradient following the solid-liquid interface direction. The mechanical properties of tensile strength and ductility were analyzed, Figure 7a–c. For both GTAW and EBW conditions, the yield strain was increased when compared to the base metal, but only the EBW weld was able to keep similar ultimate tensile strength properties. For both cases, the ductility was decreased, 15% for GTAW, 27% for EBW against 38% for the BM. Furthermore, a compositional mapping of the welding produced revealed a depletion of Mn in the weld zone for the EBW technique, where the atomic percentages in the range from 13–18 at% of Mn were reported. This depletion was associated with the evaporation of Mn due to the high power density of the EBW welding process. Indeed, this should be expected, since Mn is the element with the lowest melting temperature and highest evaporation pressure in the Cantor alloy. On the other hand, the energy input for the GTAW technique did not impact the local composition of the Cantor alloy and the atomic percentage of all elements was kept around 20 at%.

**Figure 7.** Comparison of mechanical properties of CoCrFeMnNi HEAs welded by different techniques (**a**) yield strength, (**b**) ultimate tensile strength, and (**c**) elongation.

Jo et al. [28] also undertook a comparative study on the properties of CrMnFeCoNi HEA welded by FSW and LBW. It was not observed macroscopic defects for both FSW and LBW HEA. The strength and ductility of the welded specimens were comparable with that of the BM. The FSW specimen had relatively higher yield strength (296 MPa) when compared with that of the BM (272 MPa). The loss in ductility in the FSW specimen (9%) was less than that in the LBW specimen (16%) when compared with the BM, was associated with the grain refinement due to dynamic recrystallization by the FSW process. Similarly to what Wu et al. [24] reported for EBW, the LBW technique also addressed a high energy input, which leads to a fluctuation in the composition of the FZ with the emergence of Mn-rich and Fe-rich phases. The tensile fracture tended to occur in the BM away from the weld center in the FSW specimen, while, in the LBW specimen, fracture occurred in the FZ. The fracture surface in both FSW and LBW specimens showed the features of dimpled rupture, which is typical of ductile fracture. Figure 7a–c compare the mechanical properties that were reported for the Cantor alloy joint by these different welding techniques.

Kashaev et al. [35] used LBW with a laser power of 2kW, 300μm of core diameter, 300 mm of focal length, focus position of 0.0 mm, and welding velocity in range of 3–6 m min−<sup>1</sup> to butt joints CoCrFeMnNi alloys. It was observed that the welding process resulted in the precipitation of M7C3 carbides along the fcc matrix. The precipitation of this secondary phase enhanced the hardness from 150 to 205 HV, in the BM and WZ, respectively. The authors also evaluated the influence of such a welding technique in the fatigue behavior of the alloy. No significant difference was observed between the investigated conditions. Moreover an endurance limit of 200 MPa was determined for both conditions. Due to the higher strength of the weld, the failure occurs in the BM, and any possible stress concentrators in the weld as well at the WZ/HAZ or HAZ/BM boundaries do not play any significant role. Figure 8a–d present the results of that investigation. Figure 8a shows the hardness distribution of along the material. Figure 8b the behavior of the material is showed under fatigue cycles. It is important to notice that, for cycles above 107, both of the conditions reached to a plateau which was associated with the endurance limit of the material. Finally, Figure 8c,d show the microstructure of the materials as-sintered (BM) and the LBW material after 107 cycles, respectively. One should observe a high density of dislocation in both case, but, in Figure 8d, it is arrowed the presence of the M7C3 carbides that seem to lock the dislocations around it.

**Figure 8.** (**a**) Hardness distribution along different regions of CoCrFeMnNi HEA, (**b**) fatigue behavior of the this alloy as-sintered and LBW, TEM microstructure of the alloy (**c**) as-sintered, and (**d**) laser beam welding (LBW) after 107 cycles. Adapted from [35].

In some cases, even an enhancement of the properties was obtained in WZ in comparison with the BM. Shaysultanov et al. [36] used FSW to butt joint 2 mm thickness of a modified CoCrFeMnNi alloy. Along with the main elements, 0.9 at% of C was added to the alloy. This resulted in a microstructure of the HEA alloy that consisted of face centered cubic matrix and fine Cr- rich M23C6 carbides. The use of FSW for a butt-jointing of carbon-doped CoCrFeNiMn HEA specimens permitted the formation of a sound weld without any cracks or pores. Moreover, a moderate microstructure refinement was observed as the grain size in the BMl was measured to be 9.2 μm, while being 4.6 μm for the WZ. The grain refinement was claimed to be one of the advantages of the FSW and similar results were reported for the Co16Cr28Fe28Ni28 [29], CoCrFeNiAl0.3 [30] and CoCrFeNiMn [28]. Furthermore, this microstructural change leads to a notable enhancement of mechanical properties, such as yield strength and ultimate tensile strength. By contrast, the ductility was impaired. However, it is also important to notice that the failure occurred in the BM and not in the WZ. Figure 9 shows the grain size in three different regions along the welding joint and also display the tensile strength of both the BM and WZ.

**Figure 9.** Grain size in three different regions of CoCrFeMnNi alloy welded by FSW and mechanical properties. Adapted from [36].

Table 1 summarizes the main parameters, the range, and references of the alloys that were welded by each technique discussed in this short-review.


Welding parameters and techniques reported for high entropy alloys (HEAs).

#### **4. Challenges and Future Perspective**

In an innovative work, Hao et al. [39] suggested the possibility of using a (CoCrFeNi)100-xCux HEA as a filler metal in a hybrid structure between 304 stainless steel and TC4 titanium alloy, as in Figure 10a. Figure 10b shows that the weld reinforcement presented some undercut defects, pores, and slag inclusion. No obvious interface could be seen between the WZ and 304 stainless steel, which suggested that reliable metallurgical bonding was achieved for these materials. However, a thin transition layer was formed between TC4 titanium alloy and the WZ. This transition layer could be divided into two different regions, a Ti-depleted and a Ti-rich layer, as shown in Figure 10c. All of the joints failed through the Ti/Cu transition zone, exhibiting a brittle nature with typical cleavage fracture characteristics. In spite of the unsuccessful attempt of joint this hybrid structure, such an approach with careful evaluation of the microstructural evolution of the BM/WZ interface as well as compositional matching between the materials to be welded and the filler metal could represent an important strategy for hybrid structure welding.

**Figure 10.** (**a**) Schematic illustration of dissimilar welding between TC4 Ti alloy and 304 stainless steel using a HEA as filler metal, (**b**) welding macroscopic aspect, and (**c**) layer formed between the WZ and the TC4 titanium. Adapted from [39].

The use of brazing welding has also been reported along with HEAs [40,41]. Lin et al. [40] investigated the dissimilar infrared brazing of CoCrFeMnNi equiatomic high entropy alloy and 316 stainless steel. In that study, two nickel-based fillers (BNi-2 and MBF601) were investigated as candidates for producing the CoCrFeMnNi / 316 SS joint. As expected, microstructural evolution through the joint tends to occurs and P-rich compounds were observed precipitated in the grain boundaries of the CoCrFeMnNi base metal as well as 316 SS substrate. The highest shear strength was obtained for CoCrFeMnNi / BNi-2 / 316 SS joint, 374 MPa when brazed at 1020 ◦C for 600 s, against 324 MPa that was obtained for the CoCrFeMnNi/MBF601/316 SS brazed at 1080 ◦C for 600 s. Using a different approach, Bridges et al. [41] studied the parameters and effects of using a NiMnFeCoCu HEA as a filler metal for laser brazing Inconel®718 nickel superalloy. The authors were able to achieve a reliable metallurgical bond, Figure 11a. Moreover, it was observed that, if the brazing temperature is way above the Liquidus temperature, the shear strength of the brazing decreased, as shown in Figure 11b.

**Figure 11.** Brazing dissimilar welding between HEA and Inconel 718. (**a**) Metallurgical bond and (**b**) relationship between shear strength and brazing temperature. Adapted from [41].

Abed et al. [37] used the GTAW process to add a hardfacing HEA layer into a carbon steel substrate. This could be understood as another interesting strategy for the welding of HEAs into different substrate, as the HEA filler rod was added layer after layer, as in Figure 12a. In this investigation a Fe49Cr18Mo7B16C4Nb6 was the chosen HEA filler material. The produced microstructure consisted of α-Fe matrix with Mo2FeB2 and NbC precipitated particles that substantially increased the hardness and wear resistance of the material. It was observed high-quality multi-layer deposits free of cracking with an excellent metallurgical bonding to the carbon steel base metal, as in Figure 12b.

**Figure 12.** Schematic illustration of the use of HEA welded layer after layer in a carbon steel substrate and inset of the formed interface. Adapted from [37].

Many interesting techniques, properties, and possibilities have been discussed regarding the welding in HEAs in the present work. Yet, one should be aware that most of the reported papers are based on the application of the Cantor alloy, its variants, or its modification. The CoCrFeMnNi HEA is, by far, the most investigated among all possible alloys due to both its properties at room and cryogenic temperature, compromise between strength and toughness, as well as the synergy of their main elements [42]. Nevertheless, this is just one possibility in a limitless compositional hyper-space. Miracle and Senkov [43] classified the HEAs in seven possible families and suggested that over 500,000 HEAs are possible to be produced if only equiatomic configurations with five main elements be considered. Hence, it is clear that, so far, only a minimum number of the potential application of HEAs is reported and discussed. This is particularly true regarding the welding of such materials, where only the "easiest" cases have been investigated so far.

The welding of HEAs, in which the main elements present significant differences in the melting temperature could be considered as a big challenge. Chen et al. [44] suggested that the evaporation of elements with low melting points lead to a difficult proper control of the chemical composition of the produced alloy. The aforementioned work from Wu et al. [24] proves that hypothesis. In fact, in their work was revealed the depletion of Mn in the WZ due to the high energy input through the EBW welding and the lowest melting temperature of Mn when compared to the other main elements in the Cantor alloy. Moreover, Stepanov et al. [45] was able to produce an AlNbTiV high entropy refractory alloy with over 1 GPa of compression strength and density of approximately 5.6 g/cm3. Despite these remarkable properties, one should be wondering how to weld such material, once the melting temperature for Nb is 247 ◦C and the vaporization temperature for Al is 2470 ◦C. Adapting solid-state welding techniques, such as FSW or RFW, could come up with possible solutions for this case. New processing techniques, such as metal additive manufacturing that are emerging rapidly could be helpful in the manufacturing of HEAs with controllable microstructure and enhanced properties, as well as high complex geometry components and high freedom of design [46]. In this layer-wise fabrication process, the metallic materials are bonded together by sintering or melting using high energy source, such as high power laser, electron beam, or plasma arc. Joseph et al. [47] faced some difficult obtain desired microstructure of an Al0.6CoCrFeNi that is produced by direct laser deposition (laser power of 800 W, beam focus diameter of 4mm, and velocity of 800 mm min<sup>−</sup>1). The reason was associated with the higher cooling rate and much larger thermal gradient than traditional methods, such as arc melting. It is interesting to notice that the parameters are in the same magnitude of LBW, as showed in Table 1. Therefore, this process could be seen as a localized, layer-by-layer welding technique. In fact, most of challenges faced on the welding of HEAs are also observed for additive manufacturing of these alloys. Ocelik et al. [48] focused their study on the effects of laser processing parameters in the manufacturing of an AlCoCrFeNi HEA. Deviation from the original chemical composition, as the concentration of an element would be higher the lower its melting point, and porosity were some of the commonly reported defects. Thus, understanding and optimizing the parameters could be beneficial for both additive manufacturing and LBW in HEAs.

Finally, other important points to be looked into include the geometry of the welded specimens, as well as heat treatment and protection against possible contamination during the welding. In fact, most of the papers about HEAs welding discussed in this short-review limited their investigation to the butt-joint configuration, which is the simplest possible one. Complex configurations, closer to service conditions, such as corner, edge, and T- joints, and how heat input and amount of energy would impact in the quality of the welded joint should be further studied. The investigation on different heat treatment conditions, with or without pre- or post-weld heating associated with the welding technique, should be carried out in order to verify the effects on the mechanical properties and microstructure for each condition of HEAs welding.

#### **5. Conclusions**

The present short-review discussed the recent development in the welding of high entropy alloys (HEAs) and how the different possible welding techniques impact in the microstructure, mechanical properties, and service conditions, such as fatigue and corrosion behavior in this novel class of materials.


• Different approaches were presented for the use of HEAs as filler materials for hybrid structure, as brazing welding dissimilar joints and layer-by-layer welding in comparison to additive manufacturing.

Finally, this specific field of welding still walks in baby steps, as the HEAs presents limitless possibilities of properties and perspectives for near future applications.

**Author Contributions:** Conceptualization, F.C.G.F., S.N.M.; Methodology, F.C.G.F.; Investigation, F.C.G.F., Writing—Original draft preparation, F.C.G.F.; Writing—Review and editing, S.N.M.; Supervision, S.N.M.; Project administration, S.N.M.; Funding acquisition, F.C.G.F., S.N.M. All authors have read and agreed to the published version of the manuscript.

**Funding:** This study was financed in part by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior-Brasil (Capes)-Finance Code 001.

**Acknowledgments:** The authors thank the support to this investigation by the Brazilian agencies: CNPq, CAPES and FAPERJ.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Materials* Editorial Office E-mail: materials@mdpi.com www.mdpi.com/journal/materials

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18

www.mdpi.com ISBN 978-3-0365-3874-7