Numerical Simulation and Experimental Measurement of Residual Stresses in a Thick-Walled Buried-Arc Welded Pipe Structure

Abstract

In this study, a numerical simulation of a single pass welding of two thick-walled pipes with the buried-arc method was performed in order to determine the residual stresses caused by welding. The numerical simulation procedure in the thermal analysis was performed by the element birth and death method while the structural analysis was performed simultaneously, without the application of the element birth and death technique in order to reduce the duration of the numerical simulation. The simulation results were validated by experimental residual stress measurements on the outside surfaces of the welded model using the X-ray diffraction technique. A good agreement between the results of the numerical simulation and experimental measurements was confirmed.

1. Introduction

Butt-welded pipes comprise the majority of welds in various industries such as the petrol and gas industry, shipbuilding, chemical and thermal power plants, etc. Traditionally, due to the low cost of metal active gas (MAG) technology and the ease of its use, it is a very widespread pipe welding process for large wide diameter ranges and wall thicknesses. This process is characterized by the high local input of heat required to melt the electrode which is then rapidly transferred to the base material of the structure. The subsequent rapid cooling of molten materials leads to material shrinkage in the weld zone and its immediate vicinity. It is generally known that this phenomenon leads to the appearance of undesirable effects such as changes in the crystal structure, changes of mechanical properties in the weld zone and the occurrence of permanent plastic deformations and residual stresses [1,2,3,4]. Plastic deformations disrupt the external appearance and cause dimensional deviations which is a problem when assembling pipe structures, while on the other hand, residual stresses contribute to the appearance of cracks and generally shorten the life of the welded structure [5,6]. Although there are effective thermal [7,8] and mechanical [9,10] procedures for the removal of residual stresses and plastic deformations in engineering practice, their application is often limited by the size of the structure, large number of welds, operating conditions and additional financial costs. Nowadays, the magnitude of residual stresses and strains can be predicted using numerical simulations, i.e., preventive measures should already be taken to reduce them in the design phase, thus avoiding the cost of subsequent repairs [11,12,13,14]. These methods primarily include an adequate choice of weld shapes and dimensions, welding sequences, number of passes, heat input and mechanical boundary conditions. In doing so, Teng and Chang [15] were among the first to perform a numerical simulation of two butt-welded pipes of very small wall thickness by using a two-dimensional sequential model. The conducted research provided insight into the temperature distribution during welding as well as the final state of residual stresses. Here, the numerical results obtained were not verified experimentally, but their value was shown in a number of studies that followed later. Due to the attractiveness of the topic, residual stresses in MAG butt-welded pipes have been the subject of research in numerous numerical and experimental studies in the last thirty years. Deng and Murakawa [16] developed a numerical procedure based on the ABAQUS code to simulate a temperature field, residual stress field and phase transformations for the case of multipass welding for two 2.25Cr-1Mo steel pipes. The residual stresses measured by the blind hole-drilling technique of stress relaxation showed good agreement with the numerical calculation. Giri et al. [17] presented research on the effect of weld groove on residual stresses in thick steel pipes used by boiling water systems and steam pipes. They conducted experimental measurements of residual stresses with the blind hole-drilling technique with an accompanying numerical simulation. The obtained numerical results agreed well with the experimental ones. Zhao et al. [18], on the example of thick-walled steel pipes with a large diameter, investigated the distribution of temperatures during welding as well as the influence of temperature fields on the microstructure of the weld metal. The numerically predicted hardness, microstructure and phase composition confirmed good correlation with the results of experimental measurements. Wu et al. [19] numerically and experimentally investigated the bending deformations caused by MAG welding on the example of butt-welded pipes. Liu et al. [20] used a combined shell/3D model in research of a temperature field and residual stresses. The obtained results were in good correlation with the full 3D model as well as experimental measurements.

From the above cited references, it is evident that all pipe models were welded using a conventional MAG technique, which in the case of welding thick pipes requires many electrode passes to fill the weld groove. This reduces production efficiency and the manufacturing process itself is unnecessarily prolonged, which is a significant disadvantage of the conventional MAG process. Therefore, for the needs of the industry, an improved MAG process was developed (in the literature also known as buried-arc welding), characterized by higher welding currents and voltage, higher speeds of welding wire supplies and deeper weld penetrations. This technique enables the butt-weld welding of structures up to approximately 30 mm in thickness in one pass. Compared to the conventional MAG process, the buried-arc process is also characterized by a smaller volume of weld grooves, which saves filler material. In addition, due to the reduced number of weld passes, the consumption of electricity is lower as well as the consumption of shielding gas.

It can be concluded that unlike the conventional MAG procedure, the study of residual stresses and strains in the scientific and professional literature for the case of buried-arc welded structures are very rare. In the authors’ earlier works, numerical and experimental welding was performed on the examples of butt-welded plates [21], T-joint welded plates [22] and a circular disk model [23]. In this paper the research of residual stresses was extended to thick steel pipes welded with the buried-arc procedure, which according to the authors’ knowledge is the first in the literature.

The paper consists of five sections. A brief description of buried-arc welding technology is given in Section 2. Section 3 describes the applied experimental model. Within Section 4, a detailed description of the numerical welding simulation model is given. Section 5 presents the results of numerical simulations compared to experimental measurements. The final conclusions of the paper are provided in the last section.

2. Buried-Arc Welding Technology in Brief

For the welding of thick structures, processes such as submerged arc welding [24], electroslag welding [25] and electrogas arc welding [26] have been applied in industrial production. Nowadays, laser hybrid welding processes [27] are also well known and used along with plasma arc welding [28], especially with “keyhole” technology and in some special cases electron beam welding [29]. However, conventional MAG welding is more flexible and applicable in conventional metal transfer modes, but in areas of high currents, when unstable arc and intensive spattering can occur if the arc control is not at the highest level.

This leads to the development of a new welding process based on MAG called “buried arc”. What is specific for this process is the fact that the arc is generated at a lower position under the surface of the molten pool and the space around it is the liquid metal of the molten pool. The metal transfer in this process is in the “buried” cavity which is supported by the pressure of the arc plasma. To support this kind of arc phenomena, a high current range above 500 A is required, together with a unique high speed wire feed system with a wire buffer. Furthermore, the welding power source, which consists of two digital inverter-controlled units, is necessary to obtain enough current and enable the physics of the process. The buried arc space is very unstable as the geometry of the cavity is changed due to the welding current and instantaneous arc voltage variations. The metal transfer consists of a droplet transfer and rotating arc and the conditions for stable welding are within a very narrow range. Therefore, a precise and fast control of the process is of importance to achieve an applicable process. For a more detailed description of the buried-arc welding technology, readers are referred to [30].

3. Experimental Investigations

Two pipes made of non-alloyed carbon steel S355J2+N steel and dimensions Ø168 × 12 mm, without pipe-end preparation, were buried-arc welded by using a seven-axis robot (Figure 1 and Figure 2). The welding was performed in ISO 6947: PA position. Pipe rotation with constant speed was applied to achieve the required welding velocity. The electrode inclination angle was neutral. In order to maintain I-bevel preparation, tack welds were made from the inner side on metallic backing plates with a 120° disposition. With this solution, full penetration weld could be done in a single pass. The pipe was fixed with a clamping head from one side and a rotary positioner was applied from the other side of the sample to enable rotation without eccentric movements (Figure 1). A return current path was connected through the clamping head sliding contacts. For welding, a wire A: G 42 4 C/M G3Si1 with a diameter of 1.6 mm in accordance with the ISO 14341 standard was selected. During welding, the metal transfer was stable and without significant interruptions or spattering. The main welding parameters used in this experiment are given in

Table 1. Main welding parameters.

Welding Current I	Welding Voltage U	Wire Diameter	Welding Time Duration	Shielding Gas Composition	Shielding Gas Flow
540 A	40 V	1.6 mm	68 s	100 CO2	25 L/min

. Upon completion of welding, the quality of the weld was inspected visually in accordance with the ISO 17637 standard, and no surface imperfections were found.

After welding and cooling the sample to ambient temperature, by using the X-ray diffraction cosα-method [31] using the Pulstec μ-X360 device (Pulstec Industrial Co., Ltd., Shizuoka, Japan), the residual stresses were measured. The measurement of residual stresses on the outer surface of the pipe (line z1-z2, α = 270°, Figure 2) was performed at six locations labeled ML-1 (z = 115 mm), ML-2 (z = 124 mm), ML-3 (z = 167 mm), ML-4 (z = 177 mm), ML-5 (z = 197 mm) and ML-6 (z = 208 mm). Prior to the start of the measurement, the measurement locations were electropolished and the measurement was performed at a depth of 0.015 mm below the upper surface of the pipe. Electropolishing was performed by using the EP-3 device, also a product of Pulstec Industrial Co. Ltd. The experimental setup for measuring welding residual stresses with the X-ray method is shown in Figure 3.

4. Welding Numerical Simulation

The numerical simulation of the welding process was performed by a sequential procedure consisting essentially of two steps. In the first step, a nonlinear heat transfer analysis was performed to obtain the spatio-temporal distribution of temperatures in each time increment. In the second step, the obtained temperature distribution was used as the heat load in the structural analysis. In both analyses, the temperature-dependent thermal and mechanical properties of the materials were applied and are given in Figure 4 and Figure 5, while its chemical composition is provided in

Table 2. Chemical composition of S355J2+N steel (mass%), [23].

C	Si	Mn	P	S	N	Cu	Cr	Ni	Mo	Al	V	Ti	Nb
0.17	0.24	1.25	0.016	0.006	0.008	0.23	0.06	0.1	0.11	0.0032	0.005	0.025	0.033

.

In the thermal numerical simulation, the space-time temperature field distribution was obtained by solving non-linear differential Equation (1) that is derived from Fourier’s law of heat conduction and law of energy conservation:

$$k(\frac{{\partial }^{2}T}{\partial x^2}+\frac{{\partial }^{2}T}{\partial y^2}+\frac{{\partial }^{2}T}{\partial z^2})+Q=\rho C\frac{\partial T}{\partial t}$$

(1)

In Equation (1), k, ϱ and C represent thermal conductivity, density and specific heat capacity, respectively, T is the temperature, while Q represents the external heat introduced to the welded specimen which can be expressed in the form:

$$Q=\frac{\eta UI}{V_H}$$

(2)

In Equation (2), η, U and I represent welding process efficiency, voltage and current, respectively, while V_H denotes the volume of the activated set of finite elements of the weld bead. In this study, a value of 85% was taken for the efficiency of the welding process. The following boundary conditions necessary for the solution of Equation (1) on the external surfaces of the welded pipes were assumed: the temperature-independent convective heat transfer coefficient (h_c = 10 Wm⁻²K⁻¹) and emissivity (ε = 0.9). The element birth and death technique was applied to simulate the movement of the electrode and weld filler addition. For this purpose, the entire 3 mm wide and 12 mm thick weld ring was divided into 68 sets of finite elements. All 68 sets of finite elements representing the volume of the weld were virtually removed in the first step of the numerical simulation (element death). In the following step, individual sets of finite elements (element birth) were reactivated with applied heat flux of Q = 7.08 × 10¹⁰ Jm⁻³s⁻¹ to each individual set of finite elements. Here, a pure volumetric heat flux with uniformly distributed heat input per weld volume was applied [33]. Thus, the activation of the finite element sets simulated the movement of the welding torch and filler metal addition. It is important to note that each of the 68 element sets in the numerical simulation was first added (reactivated) in a very short time interval, e.g., 10⁻⁷ s, and then subjected for 1 s to heat flux. Adding the 68th set of finite elements, the welding process ended and started the 8500-s cooling process. In the thermal analysis, DC3D8 three-dimensional eight-node linear hexahedral finite elements with full integration were used.

In order to reduce the duration of the simulation, the structural analysis was performed in one step, also without the application of the birth and death technique [34]. Here, base metals and weld fillers were considered as full isotropic and elastic-perfect plastic metals that yield in accordance with the von Mises criterion and the associated flow rule [35,36]. In the numerical simulation, due to the lack of accurate thermal and mechanical data for the weld filler, its properties were considered identical to the base material. Due to the very short exposure of the material to high temperatures, creep material was not taken into account. Furthermore, phase transformations of materials due to the low impact on the post-weld residual stress field in the case of welding structures made of carbon steel were neglected [37,38,39]. Taking into account these simplifications, the total strain tensor can be expressed in the form:

$$d{\epsilon }^{total}=\left\{d{\epsilon }^e\right\}+\left\{d{\epsilon }^p\right\}+\left\{d{\epsilon }^{th}\right\}$$

(3)

In the structural simulation, C3D8I finite elements with incompatible nodes were used and the complete sequential procedure of the numerical welding simulation was presented in the authors’ previous study [40]. The finite element mesh consisted of 42,432 elements and is shown in Figure 6. For better visibility, only half of the global model and submodel are shown.

It can be seen that a denser mesh was used in the weld zone where the temperature and stress gradients are very high. Mesh convergence was checked using the global-local submodeling technique [41]. In both the thermal and mechanical analyses, the same finite element mesh was used, only with the conversion of DC3D8 elements from the thermal analysis to C3D8I ones in the structural analysis. The mechanical boundary conditions used in this study are shown in Figure 2. Both numerical simulations were performed by using the ABAQUS software (version 6.14, ABAQUS Inc., Palo Alto, CA, USA).

5. Results and Discussion

A comparison of the axial and hoop residual stress distributions of the global model and submodel along line z1-z2, α = 90°, (Figure 2), is given in Figure 7. The full-field comparison of the global model and submodel is shown in Figure 8. Here it can be seen that the stresses of the global model and submodel match very well, which means that the global model mesh was well designed.

Temperature histories for three randomly selected nodes 15 mm away from the weld centerline (outside surface, z = 141 mm, α = 180°, α = 270° and α = 0°, Figure 2) are shown in Figure 9. It can be seen that the peak temperatures reach the value of 625 °C and that the temperature profile remains stationary and independent at an α angle. The exception to the stationary profile is only near the start/end point of welding. As mentioned in the previous section, the obtained temperature field is the load in the structural analysis, thus the stationarity of the temperature field is directly reflected in the residual stress field. Figure 10 shows the moving of the electrode and the peak process temperature of 3929 °C, resulting from the very high welding current and voltage of the buried-arc process, i.e., large heat input per weld volume. Furthermore, the reason for such a numerically obtained high temperature in the weld pool lies in the neglect of coupling between the solid and liquid phase. This phenomenon is not directly embedded in the ABAQUS software, which can lead to unrealistically high temperatures within the weld pool and is significantly different from the realistic situation. This problem can be partially alleviated by artificially increasing the thermal conductivity at temperatures above the melting point [37]. As the area of extremely high temperatures covers a very small volume of the weld and as this is a single-pass weld, this phenomenon has little impact on post-weld residual stresses and deformations.

Figure 11 shows the axial stress distributions along line z1-z2, α = 270° (Figure 2) on the outside surface of welded pipes with the results of X-ray measurements. The numerically obtained axial residual stresses are compressive in the weld and its vicinity, reaching a minimum value of −360 MPa at a distance of 5 mm from the weld centerline. The maximum tensile stress is 79 MPa at a distance of 51 mm from the weld centerline. It can be confirmed that the numerically calculated results of axial stresses are in good correlation with the experimentally obtained ones. In Figure 12, the axial stress distributions along line z1-z2, α = 270° (Figure 2) on the inside surface of the welded model is presented. It can be seen that unlike the outside surface, tensile residual stresses dominate here and reach a value of 372 MPa, while the minimum compressive stress reach −82 MPa.

Figure 13 shows hoop stress distributions along line z1-z2, α = 270° (Figure 2) on the outside surface of the welded model and it can be noticed that the stress distribution is very uneven here. This phenomenon was also observed in previous research on welded thin pipes [42]. Very sudden changes in stresses are also reflected in differences compared to experimental measurements. The amplitude range of hoop stresses is between −221 and 236 MPa. In contrast, residual stress hoop distribution on the inner side (Figure 14) is steady in a range of −177 to 335 MPa. Generally speaking, axial and hoop stresses exist only in the weld and zones close to it, while towards the ends of the pipe they disappear.

Axial and hoop residual stress distributions through the pipe thickness are shown in Figure 15. If the areas close to the outer and inner surface are neglected, it can be noticed that the change of axial residual stresses from tensile to compressive is almost linear and almost symmetrical with respect to the middle pipe surface, while hoop stresses always remain positive. Cross sections through thickness axial and hoop residual stress distributions (z = 156 mm, Section A-A, Figure 2) are shown in Figure 16 and Figure 17, respectively. Figure 18 and Figure 19 show full field axial and hoop stress distributions, respectively, and for better visibility only the left half of the welded model (Figure 2) is shown.

6. Conclusions

In the frame of this numerical and experimental study, investigations of the temperature field during welding as well as post-weld residual stresses for the case of a single-pass welding of two thick pipes made of non-alloyed S355 carbon steel by using the buried-arc technique were conducted, which according to the authors’ knowledge is the first case in the scientific and professional literature. The conclusions are as follows:

• Temperature histories after the start of welding become steady very quickly along the welding path.
• Axial residual stress at the weld centerline on the outer surface are compressive. As they move away from the welds, they become tensile and disappear towards the ends of the pipe. The comparison of numerically and experimentally obtained stresses showed very good agreement.
• Axial residual stress at the weld centerline on the outside surface are tensile. As they move away from the welds, they become compressive and disappear towards the ends of the pipe.
• Hoop stresses on the outer surface of welded pipes very quickly change the sign of stress and disappear towards the ends of the pipe. The deviation in the measured and numerically calculated results is greater compared to axial residual stresses due to sudden changes in hoop stresses in the weld zone.
• Hoop stresses on the inner surface of welded pipes in the weld zone are tensile and as they move away from the welds become pressured and disappear towards the ends of the pipe.
• Through the thickness change of residual stresses from compressive to tensile, it is almost linear if the effects at the outer or inner surface of the welded model are neglected.
• Hoop stresses through the thickness are always tensile.