ABSTRACT
In the field of Mg/Al dissimilar metal welding, high-temperature interlayer-assisted dual-molten-pool laser welding has been proven effective for metallurgical joining of Mg/Al dissimilar alloys. To explore alternative interlayers for joint optimization, finite element models offer an efficient alternative to extensive process experiments, enabling rapid prediction of interlayer adaptability. However, existing models generally simplify interlayer contact as ideal thermal conduction, failing to reproduce dynamic thermal behavior at dissimilar material interfaces. Taking Ni-interlayered Mg/Al welding as the benchmark, this study proposes a high-precision numerical modeling method based on temperature-dependent contact thermal resistance mapping. The “Indentation/Staining-Binarization Method” is developed to extract real interfacial micro-contact features, which, combined with a USDFLD subroutine, accurately reproduces the dynamic heat transfer mechanism from mechanical contact to metallurgical bonding. The model achieves high-confidence prediction of molten pool morphology and key point thermal cycles, capturing the initial thermal barrier effect and post-melting thermal conduction mechanism. This work establishes a systematic dynamic heat transfer modeling framework for non-ideal interfaces, providing a reliable benchmark and low-cost prediction platform for interlayer optimization and process regulation of dual-molten-pool laser welding.
Keywords:
Dissimilar metal welding; Finite element models; Heat transfer mechanism; Interfacial micro-contact characterization
1. INTRODUCTION
Achieving reliable joining of magnesium alloys and aluminum alloys enables lightweight alloys to replace traditional steel materials in service across multiple fields, supporting energy conservation and emission reduction [1]. However, the welding of Mg/Al dissimilar metals is confronted with a core challenge—Mg/Al alloys tend to form hard and brittle Mg-Al intermetallic compounds (IMCs), resulting in poor joint performance [2]. To effectively regulate this brittle phase, the measures commonly employed include reducing heat input or adding an interlayer.
Reducing heat input controls the interdiffusion of magnesium and aluminum metals, thereby inhibiting the formation process of Mg-Al IMCs. This approach is typically achieved through welding processes such as laser welding and Cold Metal Transfer (CMT) [3, 4]. From a kinetic perspective, intermetallic compound formation requires time and energy. Reducing heat input shortens both the molten pool lifetime and the high-temperature dwell time. This significantly limits the interdiffusion time between magnesium and aluminum atoms, leading to the formation of an exceptionally thin intermetallic compound layer. Consequently, the adverse effects on the mechanical properties of the entire welded joint are minimized. From a thermodynamic perspective, lower heat input implies a relatively lower overall temperature in the weld zone. Since atomic diffusion coefficients increase exponentially with temperature, reducing the temperature directly slows the diffusion rate of atoms, thereby inhibiting intermetallic compound growth. However, while simply reducing heat input can decrease the formation of Mg-Al IMCs and reduce the thickness of the intermetallic compound layer, it cannot alter the inherent hardness and brittleness of the Mg-Al IMCs themselves. Under load, stress becomes highly concentrated at the Mg-Al IMCs layer. Since Mg-Al IMCs themselves exhibit virtually no plastic deformation capacity, microcracks rapidly initiate and propagate, leading to low-stress brittle fracture of the joint. At this point, the strength of the welded joint is determined not by the base metals, but by this weakest IMCs interfacial layer. Magnesium and aluminum alloys exhibit significant differences in their physical properties, including thermal conductivity and specific heat capacity. Consequently, substantial thermal stress tends to develop within the joint during the welding thermal cycle process, which in turn induces defects such as cracks. In addition, the differences in surface tension and wettability between the two materials in the liquid state make it difficult to achieve stable molten pool flow and sound weld formation.
To address this issue, researchers have adopted the approach of adding interlayer materials for Mg/Al alloy welding. The core principle of introducing an interlayer is to alter the reaction pathway and prevent direct contact between magnesium and aluminum atoms. Common types of interlayers are generally categorized into powder, welding wire and foil, with Zn, Ti, Ni, Ag and other metals serving as typical interlayer materials [5,6,7,8]. The interlayer metals react with either aluminum or magnesium, thereby replacing the direct reaction between magnesium and aluminum and forming new intermetallic compounds, whose plasticity and toughness are generally superior to those of the hard and brittle Al-Mg intermetallic compounds. For example, Zn can form solid solutions with relatively good toughness when reacting with Al. In addition, Zn has a certain solubility in Mg, thus exerting a transition and buffering effect. The brittleness of the resulting Mg-Zn or Al-Zn phases is far lower than that of Al-Mg phases [9]. Ag can form various intermetallic compounds with Al, but their brittleness is generally lower than that of Al3Mg2, and Ag is less likely to form a large amount of brittle phases with Mg [10].
Combining the two aforementioned methods for controlling the formation of Mg-Al IMCs, we propose the dual-molten-pool laser welding process. The dual-molten-pool laser welding process employs laser welding—featuring low heat input, high energy density and promising prospects for automated production—as its heat source. A barrier interlayer is used to fully block Mg-Al interdiffusion, and differential heat input is applied to each side of the interlayer to form separate molten pools. Welding is ultimately achieved by wetting of the interlayer by the molten pools on both sides, followed by the formation of metallurgical bonds. Previously, based on this dual-molten-pool laser welding process, we achieved differential heat input control for the two molten pools by using a Ni interlayer as the physical barrier. The fabricated joints exhibited a tensile strength of 83.58 MPa, which is significantly superior to that of joints prepared by traditional processes such as CMT and laser brazing [11].
Against this background, investigating the suitability of other interlayer materials for this process has become an important research direction. However, given the numerous influencing factors of dual-molten-poollaser welding, relying on the traditional trial-and-error method for process parameter optimization of multi-component interlayer materials will lead to high experimental costs and extremely long R&D cycles, making it difficult to achieve efficient process screening. In contrast, the application of welding numerical simulation technology can provide efficient support for the adaptability research of multi-component interlayer materials. In our in-depth research on dual-molten-pool laser welding, we found that intense thermal hydrodynamic interfacial responses are generated during the sequential formation of the two weld seams. The interface exhibits distinct discontinuous heat transfer characteristics, which increases the difficulty of screening suitable welding parameters. Such heat transfer characteristics stem from incomplete thermal conduction between materials. Contact between the rough peaks and valleys on the material surface results in the actual effective contact area being much smaller than the apparent surface area, and the morphological differences caused by surface roughness constitute the physical origin of the resistance to heat transfer [12]. Nevertheless, existing numerical simulation studies on dissimilar metal welding mostly focus on the distribution of macroscopic temperature fields, and when dealing with multi-layer heterogeneous structural interfaces, such contact interfaces are generally simplified into ideal full contact [13,14,15,16,17,18].
As a matter of fact, the significance of the actual contact state of solid interfaces has attracted the attention of numerous researchers. For the elastoplastic spherical contact system, EGUCHI et al. [19] employed polarized white light optical interferometry to visualize the evolution law of the contact area at the interface between a rough rubber hemisphere and glass, confirming that the actual contact area is much smaller than the macroscopically observed area. For metallic materials, OVCHARENKO et al. [20] tracked the evolution law of the actual contact area of copper and stainless steel with glass during the load cycling process via in-situ optical technology. LIANG et al. [21] and LI et al. [22], based on total internal reflection, investigated the actual contact behavior between copper and glass with different surface conditions, and found that the size of the actual contact area is closely correlated with surface conditions and contact pressure. KALIN and JAN [23] took steel and sapphire as the research objects, and clarified the nonlinear regulatory effect of surface roughness and load on the actual contact area via in-situ testing with submicron resolution. However, existing technologies for measuring the actual contact area mostly rely on transparent materials as contact counterparts, and fail to achieve the measurement and characterization of the actual contact state between opaque metals. Simplified treatments that ignore this actual contact state in numerical simulation work will detach from the physical reality of interfacial contact, causing the simulation results to artificially smooth the temperature between dissimilar metals and fail to capture key temperature jumps. Therefore, achieving the quantitative characterization of the actual contact state between opaque metals and embedding it as a key material characteristic into the finite element model constitutes the core prerequisite for further research on the dissimilar metal dual-molten-pool laser welding process.
This study takes the dual-molten-pool laser welding of AZ91/6061 dissimilar metals as the research object and establishes a high-confidence finite element model capable of truly reflecting interfacial dynamic heat transfer. First of all, this study developed an Indentation/Staining-Binarization Method characterization technique to quantitatively obtain the actual contact state at the interface between the Ni interlayer and the base metal under different temperature ranges. Furthermore, the discrete mapping method was employed to convert this contact state into the spatial boundary conditions of the finite element model, and the USDFLD subroutine was utilized to realize the simulation of the nonlinear evolution of the interface during the transition from non-ideal mechanical contact to full contact with metallurgical bonding. Secondly, the numerical simulation results are highly consistent with the weld cross-sectional morphology and temperature variation law at key points obtained from welding experiments, which verifies the correctness of this model. Finally, based on this model, an in-depth investigation was conducted into the regulatory mechanism of non-ideal interfaces on the dual-molten-pool temperature field distribution, providing a scientific basis for the research on thermophysical behaviors in dissimilar metal joining.
2. EXPERIMENTAL AND NUMERICAL SIMULATION METHODS
2.1. Experimental materials
The materials used in both the indentation/staining-binarization experiments and welding experiments in this study were AZ91 magnesium alloy, 6061 aluminum alloy, and industrial pure nickel. The chemical compositions of the three materials are shown in Table 1. The base metal dimensions are 100 mm × 40 mm × 2 mm, and the interlayer dimensions are 100 mm × 4 mm × 2 mm.
2.2. Quantitative characterization of interface contact states: indentation/staining-binarization method
In numerical simulations of dissimilar metal welding with interlayers, accurately defining thermal conduction behavior between materials is central to enhancing model reliability. Due to the unavoidable surface roughness of solid materials, actual contact between solids is discontinuous, with interface loads supported solely by rough micro-asperities between solid materials. The opaque nature of the metallic materials used in the experiment makes direct observation of microscopic contact states at high temperatures extremely challenging. This section proposes a method combining experimental measurement with numerical mapping to digitize the actual contact state of materials.
2.2.1. Experimental principles and platform setup
This work developed an Indentation/Staining-Binarization Method to address measurement challenges in opaque metal contact states, aiming to obtain the effective thermal conductivity area ratio at the solid contact interface under constant load at different temperatures. This method avoids the limitations of optical techniques and transparent media. Its characterization performance is unaffected by coating thickness, and it requires no vacuum environment, complex procedures or expensive in-situ equipment, enabling quantitative measurement of the contact state of opaque solid materials under various conditions. Cubic specimens measuring 4 mm × 2 mm × 2 mm were machined from AZ91, 6061 and Ni plates.The contact surfaces of the Ni interlayer and base metals were pre-treated to surface roughness values of Ra = 3 μm and Ra = 0.5 μm, respectively. In this study, it is verified that under the specified surface roughness conditions, the two types of dissimilar metal interfaces can form uniformly distributed effective thermally conductive contact sites even under a low clamping load, which balances the physical performance of microscopic interfacial contact and the implementability in macroscopic engineering applications. To simulate the thermo-mechanical coupling environment during welding, a test procedure was designed as shown in Figure 1. In this study, a longitudinal preload of 100 N applied by the welding fixture was selected, which corresponds to the 2 N test load used in the microscopic interfacial contact experiments. This preload magnitude not only ensures the assembly stability and positional accuracy of thin-sheet materials during the thermal cycle of high-energy beam welding at the macroscopic level, but also effectively avoids secondary defects at the microscopic physical level, such as macroscopic irreversible plastic deformation of soft base metals and misalignment of mating contact surfaces caused by excessive mechanical loading. Coat the designated contact surface of the Ni specimen with edding 750 series high-temperature resistant oil-based ceramic ink. Apply a constant 2 N normal load to the specimen assembly to simulate fixture preload. The specimens were then divided into three groups and subjected to different temperature treatments: one group was placed directly in a 20°C environment; the other two groups were placed in a temperature-controlled furnace and heated to 200°C and 400°C respectively before being immediately removed and cooled in water. Using an optical microscope, collect data on the indentation and the distribution of stained areas on the base metal surface in contact with Ni. In this study, the two materials establish contact through plastic deformation and elastic deformation induced by a normal load. The plastic deformation zones are characterized by indentation distribution, while the elastic deformation zones exhibit staining characteristics due to contact-induced wetting of high-temperature ink, enabling the separate characterization of the two contact modes. Based on this, the obtained images are processed by performing binary segmentation to separate indentation/stained areas from unstained regions. This enables quantitative characterization of the interface contact state, providing a physical basis for subsequent discrete mapping of the finite element mesh.
Schematic diagram of the experimental process for testing the interface contact area ratio.
2.2.2. Variation in effective thermal conductivity area ratio and mesh mapping
Using an OLYMPUS DSX510 optical microscope to observe the contact surfaces of 6061 and AZ91, the thermal conduction state during contact in the numerical simulation work was determined based on the distribution of indentations and stained areas on the surfaces of 6061 and AZ91. To convert experimentally measured two-dimensional contact images into boundary conditions recognizable by finite element models, this study employs a mesh discretization mapping method. Divide the microscopic image into a regular grid matrix, with each grid cell representing an interface element in the finite element model. Based on the image, areas with dense indentations/staining are determined to be in contact, while unstained areas are determined to be non-contact. Combined with the Otsu automatic thresholding algorithm in ImageJ software, the interfacial contact state distribution was processed by mesh-based localized binarization with a uniform grayscale threshold of 135, and the corresponding binarization and meshing procedures are illustrated in Figure 2.
Five specimens were adopted for each parameter condition, and the corresponding indentation/staining area ratios are shown in Figure 3.
The detailed indentation/staining distribution and grid mapping results are presented in Figure 4.
Microscopic contact state at dissimilar metal interfaces at typical temperatures and finite element mesh mapping: (a) Mg-Ni 20°C; (b) Mg-Ni 200°C; (c) Mg-Ni 400°C; (d) Al-Ni 20°C; (e) Al-Ni 200°C; (f) Al-Ni 400°C; (a1-f1) Grid mapping.
2.2.3. Micro-mechanism analysis
Indentation/Staining-Binarization Method is proposed based on the physical principle that contact on rough surfaces occurs only at the tips of micro-protrusions at the microscopic level. As temperature rises, the material’s yield strength decreases, causing the micro-protrusions to deform under load pressure and significantly increasing the actual contact area. Microscopic observations clearly confirm this: at 20°C, only sparse contact points exist at the Al-Ni/Mg-Ni interface. Upon heating to 200°C and 400°C, the number of contact points significantly increases and connectivity emerges, indicating a substantial increase in effective heat transfer pathways.
To ensure computational stability, minor adjustments were made to the contact points of the material at 200°C and 400°C, guaranteeing topological continuity of the contact state as temperature varies. The adjustment principle is to maintain the proportion of effective thermal conductivity area ratio, and perform nearest-neighbor positional adjustment of the high-temperature contact points based on the low-temperature contact points. The adjusted high-temperature contact points fully cover the positions of the low-temperature contact points, so as to prevent random jumps in the contact area. The adjusted grid mapping positions are shown in Figure 5.
Adjusted finite element mesh discretization mapping: (a) Mg-Ni 20°C; (b) Mg-Ni 200°C; (c) Mg-Ni 400°C; (d) Al-Ni 20°C; (e) Al-Ni 200°C; (f) Al-Ni 400°C.
Through the above steps, we abstract complex microscopic contact physics into a set of discrete spatial distribution data that evolves with temperature, laying the data foundation for subsequent high-precision modeling.
2.3. Finite element model construction
2.3.1. Model construction and mesh generation
All parameters adopted in this study were finally converted to the T·mm·s unit system. Based on the characteristics of dual-molten-pool laser welding, a model of the two base metals and the interlayer was established in ABAQUS. The base metal dimensions are 100 mm × 40 mm × 2 mm, and the interlayer dimensions are 100 mm × 4 mm × 2 mm. The welded joint configuration is a butt joint. Owing to the high energy density of laser welding, the temperature gradient in the weld zone is relatively large. To balance both the computational accuracy and efficiency of numerical simulations, the mesh configuration is designed with dense meshes in the weld zone. Three mesh schemes for the weld zone with element sizes of 0.4 mm, 0.2 mm and 0.1 mm were compared in this study. The results show that when the mesh size decreases from 0.4 mm to 0.2 mm, the isotherms of the molten pool become smoother. However, when the mesh is further refined to 0.1 mm, the computational time increases exponentially due to the further increase in the number of iterative calculations for heat source model loading, and the computational accuracy shows negligible difference compared with that of the 0.2 mm mesh. Considering both accuracy and iterative efficiency comprehensively, the element size of 0.2 mm was finally selected for the core weld zone, i.e., 0.2 mm × 0.2 mm × 0.2 mm, and the mesh gradually becomes sparse away from the weld zone. Additionally, to simulate the contact state between the interlayer and the base metals, a 0.02 mm × 0.2 mm × 0.2 mm mesh layer was placed between the interlayer and each base metal. This constitutes a key feature of the model. The elements in this layer do not represent actual thickness but serve solely as physical carriers for the contact thermal resistance properties. The constructed model is shown in Figure 6.
Three-dimensional finite element model of AZ91/6061 dual-molten-pool laser welding: (a) Three-dimensional component assembly model; (b) Three-dimensional mesh generation and interface element layer.
2.3.2. Thermophysical parameters of materials
The data obtained through calculations using JMatPro software and adjustments based on literature data are as follows: The solid-liquid phase transition enthalpy for 6061 aluminum alloy is 465 J/g; The solid-liquid phase transition enthalpy for AZ91 magnesium alloy is 524 J/g; The solid-liquid phase transition enthalpy for Ni is 297 J/g. Other parameters are shown in Figure 7. To simulate subsequent changes in contact state, the figure also includes the thermophysical parameters of air.
Thermophysical parameters of Ni, 6061, AZ91, and Air: (a) Thermal conductivity; (b) Density; (c) Specific heat capacity.
2.3.3. Handling of contact status
As the weld undergoes melting and solidification, the contact state between the base metal and the interlayer dynamically transitions from non-ideal mechanical contact to metallurgical bonding during the welding process. This work employs a finite element mesh based on the mapping of actual interface contact states. The mesh is repeatedly arranged and mapped across the constructed thin interface layer elements, exhibiting “short-range disorder and long-range order” until the entire contact layer is fully covered. The USDFLD subroutine is used to manipulate the material properties of the thin-layer elements at the interface between the two base metals and the interlayer, enabling the properties of these thin-layer elements to transition from those of air to those of the base metals based on the elements’ own temperatures, thus achieving the dynamic evolution of the contact state. The USDFLD subroutine logic is: The program reads element temperatures and preset contact distribution mapping data in real time. For elements that do not satisfy the contact criteria, air’s thermophysical parameters are assigned. Based on the element’s own temperature, when contact criteria are met—specifically at 200°C, 400°C, and its melting point temperature—a transition occurs in batches from air’s thermophysical parameters to the thermophysical parameters of both base materials. When the interface temperature exceeds the melting point of the base metal, it is determined that melting has occurred in that region, forming a dense metallurgical bond. At this point, regardless of the original contact state, the thermophysical parameters of all elements in this region switch to a fully contacted state, meaning they all adopt the thermophysical properties of the base metal. This simulates the process of liquid metal wetting and filling the gap.
In the actual welding process, the evolution of the interfacial contact state is essentially a gradual process with continuous change as a function of temperature. However, it is difficult to quantitatively characterize the full-process evolution of interfacial contact through experiments, and at the level of finite element numerical simulation, the description using continuous functions incurs extremely high material nonlinear iteration costs. Therefore, this study adopts a stepwise discrete switching strategy for interfacial contact based on characteristic temperature points.
The evolution of thin-layer element properties undergoing parameter switching with increasing temperature is shown in Figure 8: From sparse contact points at low temperatures, the contact area gradually increases as the temperature rises. All elements within the region where the temperature reaches the melting point of the base metal undergo transformation, with the thermophysical parameters of the base metal dominating the thermal conduction process.
Set of contact elements at the interface of dual-molten-pool laser welding: (a) Mg-Ni contact at 20 °C; (b) Al-Ni contact at 20 °C; (c) Newly formed Mg-Ni contact at 200 °C; (d) Newly formed Al-Ni contact at 200 °C; (e) Newly formed Mg-Ni contact at 400 °C; (f) Newly formed Al-Ni contact at 400 °C; (g) Newly formed Mg-Ni contact at 595 °C; (h) Newly formed Al-Ni contact at 660 °C.
It should be noted that in the present model, once a mesh element is designated as a “contact” element based on experimental mapping, its thermal conductivity is directly set to that of the corresponding base metal, i.e., the solid-solid contact thermal resistance at actual micro-asperity contact points is neglected. This treatment is grounded in the following rationale: at the microscale, real contact points are subjected to extremely high local pressure, which induces plastic deformation and forms bonds approaching metallic bonding, with thermal resistance orders of magnitude lower than that of the base material. Hence, assigning the thermal conductivity of the base metal to these contact elements physically represents the high-efficiency heat conduction paths across the interface. For non-contact elements, the thermal conductivity of air is adopted to characterize the barrier effect of interfacial gaps. This “metal/air” binary assignment, based on the statistically measured area ratio and spatial distribution, can equivalently reproduce the overall thermal resistance of rough interfaces at the macroscopic engineering scale, without introducing empirical contact resistance coefficients that are difficult to calibrate.
The main drawback of this simplification is that the binary discretization smooths out contact inhomogeneity at the sub-grid scale, which may slightly underestimate the local temperature gradient near contact/non-contact boundaries. Verified by experiments, the model predictions of molten pool morphology and thermal cycles at key points are highly consistent with experimental measurements, confirming that the error introduced by this simplification for engineering-scale thermal analysis is fully within the acceptable range. Accordingly, the present model is applicable for predicting the global temperature field and heat transfer mechanisms, but not for resolving microscale phenomena such as local stress concentration or microcrack initiation at individual micro-asperities.
2.3.4. Governing equations and boundary conditions
When analyzing the temperature field using the finite element method (FEM), the only variable in the finite element model is the temperature parameter, whose basic thermal conduction differential equation can be expressed as Formula (1):
Among them, ρ is the density of the material; c is the specific heat capacity; T is the temperature; kx, ky, and kz are the thermal conductivity coefficients of the material in the x ,y and z directions, respectively; Q is the internal heat source intensity.
In the numerical simulation, the heat input into the base metal during welding is represented by the welding heat source model. Considering the heat input in the depth direction of laser deep penetration welding, the conical heat source model was adopted in this study to approximate the energy attenuation distribution of laser beams along the thickness direction. Its heat flux density function can be described as Formula (2):
Among them, η is the welding efficiency coefficient, taken as 0.8; Q is the laser power; H is the heat source height, with a value of 4.0 mm; and rs is the heat source radius, with a value of 1.0 mm. The DFLUX heat source subroutine was implemented into the finite element model to perform numerical calculations, and its heat flux distribution is shown in Figure 9 [24].
Based on the ambient temperature of the welding experiment, the ambient temperature in the initial simulation condition was set as 20 °C. In addition to the heat input boundary condition, the heat output boundary condition is divided into two types: convective heat dissipation boundary condition and radiative heat dissipation boundary condition. The equation for the convective heat dissipation boundary condition is expressed as Formula (3):
Among them, h is the convective heat transfer coefficient; T is the material temperature; and T∞ is the ambient temperature. In the experiment conducted in this study, high-purity argon gas with a flow rate of 20 L/min was used as the shielding gas to prevent oxidation of the materials during welding. The values of h were set as follows: h6061 = 50 W/(m2·K), hAZ91 = 40 W/(m2·K), hNi = 60 W/(m2·K) [25].
The equation for the radiative heat dissipation boundary condition is expressed as Formula (4):
Among them, ε is the material surface emissivity; σ is the Stefan–Boltzmann constant; T is the material temperature; and T∞ is the ambient temperature. The values of ε for the three materials are listed in Table 2 [26,27,28,29].
2.4. Welding experiment verification scheme
The correctness of welding finite element models serves as the foundation for predicting temperature field distribution, and the correctness of temperature field distribution constitutes the prerequisite for predicting welding residual stress, deformation, microstructural evolution and other related aspects [30,31,32,33]. In order to ensure the accuracy and reliability of the welding finite element model established in this study, systematic verification was conducted based on the numerical simulation results of the temperature field, and thus a Mg-Al dissimilar metal dual-molten-pool laser welding verification experiment was carried out. The welding process is essentially a complex thermophysical process involving dynamic thermal variations, and this study verified the simulation accuracy from two aspects: the macroscopic cross-sectional morphology of the weld and the thermal processes at key points. This study selected the cross-sections of the weld fusion zone and the thermal cycle curves at key points as validation metrics. Quantitative comparisons were performed between the numerical simulation results and the corresponding experimental measurement data to comprehensively evaluate the predictive capability of the model.
The specifications of the base metals and interlayer used in the welding experiment were consistent with those described earlier (Table 1). The experiment adopted Prima Power’s Papido 3D fiber laser welding machine, and the Thermocouple Temperature Measurement Method (TTMM) with AT4208K type thermocouples was employed for temperature measurement at key points. The dimensions of the materials used, contact surface conditions, welding parameters, and welding sequence are shown in Table 3.
The experimental procedures were as follows:
-
(1)
Remove the oxide films on the surfaces of the aluminum alloy and magnesium alloy.
-
(2)
Place the Mg, Ni and Al specimens in sequence on a parallel-jaw vice and apply a longitudinal clamping force; adjust the clamping force to 100 N using the pressure tester shown in Figure 1.
-
(3)
Fix the temperature-measuring ends of the thermocouples to the selected positions on the Mg side, interlayer and Al side, respectively, and perform temperature acquisition at intervals of 0.5 seconds.
-
(4)
Adopt double-sided protection with high-purity argon gas at a flow rate of 20 L/min.
-
(5)
Adjust the laser spot diameter to 1.0 mm, and set the welding power and welding path.
-
(6)
Weld the Mg side; after welding, let the assembly stand for 5 minutes until the Mg-Ni weld cools down to room temperature, then read the temperature data collected by the thermocouples.
-
(7)
Weld the Al side; after welding, let the assembly stand for 5 minutes until the Al-Ni weld cools down to room temperature, then read the temperature data collected by the thermocouples.
-
(8)
Remove the thermocouples and organize the recorded temperature data. To reduce the error caused by the thermocouple response time, this study performed corresponding translation processing on the temperature data collected from three repetitions according to the simulated heating response time.
-
(9)
Use wire electrical discharge machining (WEDM) to cut metallographic samples perpendicular to the welding path for quantitative observation of the weld cross-sectional morphology.
The welding path and the selected positions for temperature acquisition are illustrated in Figure 10.
Schematic of welding paths and temperature acquisition points in dual-molten-pool laser welding experiment.
3. RESULTS AND DISCUSSION
3.1. Model validation
3.1.1. Validation of weld cross-sectional morphology
In welding numerical simulation, the weld cross-sectional morphology is a direct characterization of heat input, and its matching degree with the actual weld directly reflects the model’s accuracy in reproducing the welding physical process. Therefore, the validation of this characteristic is an essential prerequisite for ensuring the reliability of simulation results and underpinning subsequent analyses.
A comparison between the numerical simulation results and experimental metallographic observations of the Mg-Ni and Al-Ni weld cross-sections is shown in Figure 11. All images were calibrated with the same scale to ensure the objectivity and accuracy of the comparison. From the perspective of macroscopic morphology, the simulated molten pool profiles of the two welds exhibit high geometric consistency with the experimentally measured results, both showing a typical semi-funnel-shaped morphology. To further quantify the accuracy of the simulation results, three key characteristic positions (upper, middle, and lower) on the weld cross-section were selected, and a quantitative analysis was performed on the simulated predicted values and experimentally measured values.
Weld cross-sectional morphology: (a) Numerical simulation of Mg-Ni side; (b) Welding experiment of Mg-Ni side; (c) Welding experiment of Al-Ni side; (d) Numerical simulation of Al-Ni side.
Figure 11(a) shows the Mg-Ni side weld: the simulated predicted widths of the fusion zone at the upper, middle, and lower surfaces were 0.80 mm, 0.62 mm, and 0.43 mm, respectively; the corresponding experimentally measured values were 0.78 mm, 0.58 mm, and 0.41 mm. After calculation, the relative errors at each characteristic position were 2.56%, 6.89%, and 4.88%, respectively, with an arithmetic mean comprehensive error of 4.77%. The error distribution was uniform and at an extremely low level.
Figure 11(b) shows the Al-Ni side weld: the simulated predicted widths of the fusion zone at the upper, middle, and lower surfaces were 0.49 mm, 0.37 mm, and 0.06 mm, respectively; the corresponding experimentally measured values were 0.46 mm, 0.38 mm, and 0.08 mm. The relative errors at each position were 6.52%, 2.63%, and 25.0% in sequence. Specifically, the errors of the upper and middle surface widths were controlled within 7%, while only the error of the lower surface was slightly higher, which was attributed to the solidification shrinkage effect of liquid metal, with an absolute error of merely 0.02 mm.
The validation results confirm that the heat source energy distribution parameters adopted in the model can reasonably reflect the attenuation law of heat along the plate thickness direction. The consistency between the overall shape and key dimensions fully demonstrates that the model reliably captures the core physical characteristics of the welding process, laying a credible foundation for its application in predicting thermal cycles, stress-strain behavior, and microstructure evolution.
3.1.2. Validation of welding thermal cycle curves at key points
Temperature variation characteristics of key points over time serve as a sampling characterization for the spatial distribution and dynamic evolution of the temperature field.
Comparing and validating the simulated curves against the measured data is also a key step to ensure the reliability of the model. In this study, three characteristic points covering key regions were selected for comparative analysis between the curves predicted by the Finite Element Method (FEM) numerical simulation and the curves measured by the Thermocouple Temperature Measurement Method (TTMM). The specific positions are as follows: P1 is located in the AZ91 magnesium alloy base metal region, offset 8 mm along the negative Y-axis from the center of the Step 1 welding path; P2 is located in the Ni interlayer region, offset 8 mm along the positive X-axis from the center of the ends of the two welding paths; P3 is located in the 6061 aluminum alloy base metal region, offset 8 mm along the positive Y-axis from the center of the Step 2 welding path.
The temperature variation curves at key points selected for the experiment are shown in Figure 12. Overall, both exhibit highly consistent temperature change patterns: starting from an ambient temperature of approximately 20 °C, they rapidly rise to peak temperatures during the laser heat source application phase, followed by a sustained thermal diffusion cooling phase as the heat source moves away. No significant deviation from this trend was observed. This matching indicates that the heat source behavior and the formulation of the thermal conduction control equations in the model are well matched to the actual thermophysical processes of welding.
Temperature variation curves of key points: (a) Step 1–P1; (b) Step 1–P2; (c) Step 1–P3; (d) Step 2–P1; (e) Step 2–P2; (f) Step 2–P3.
The differences in thermal cycle responses during the two-step welding process were accurately captured by the model: during Step 1 welding, only P1 and P2 exhibited significant temperature rises, while the temperature change at P3 was relatively weak, which reflects the thermal barrier effect of the Ni interlayer. In contrast, during Step 2 welding, secondary heating peaks appeared at P3, P2, and P1 simultaneously. This phenomenon is highly consistent with the design concept of “zonal heat control and reaction isolation” for the dual-molten-pool welding process.
Through an in-depth analysis of the three points P1, P2, and P3, the reliability of the model can be evaluated from different dimensions:
For point P1, both the simulated and measured thermal cycle curves exhibit the characteristic of “rapid heating – slow cooling”. The peak temperature is 146.24 °C in the FEM simulation and 156.5 °C in the TTMM measurement, with the deviation controlled within 7%. Point P2, located in the interlayer, is the key to evaluating the interfacial heat transfer model. The measured and simulated curves of point P2 are highly coincident in the peak temperature and cooling stages, which verifies the effectiveness of the non-ideal interfacial heat transfer between Mg-Ni and Al-Ni regulated by the USDFLD subroutine in the model, and indicates that the model accurately characterizes the role of the interlayer as a thermal barrier. Compared with point P1, the thermal cycle curve of point P3 exhibits a lower peak temperature and a wider curve profile. This is attributed to the higher thermal conductivity of the aluminum alloy, which enables faster heat transfer to distant regions. The comparison results show that the simulation error of point P3 at the thermal equilibrium stage is 4.2%. The aforementioned deviation is attributed to the experimental error induced by slight laser beam deviation and the random error associated with the actual contact condition.
Based on the comprehensive comparison results above, it can be concluded that the model not only reproduces the key characteristics of the welding thermal cycle, including heating rate, peak temperature range, and cooling trend, but also captures the core physical features such as regional temperature variation differences caused by discrepancies in material thermophysical properties, thermal cycle effects of multi-pass welding, and the switching between thermal barrier and conduction dominated by interfacial contact conditions. Meanwhile, the detailed deviations are within the range of comprehensive errors in welding thermal measurement and numerical simulation. This result further verifies that the finite element numerical model incorporating the stepwise discrete switching approach for interfacial contact states correctly predicts the evolution law of the welding temperature field.
3.2. Dynamic evolution analysis of temperature field
Unlike the continuum heat transfer in single-molten-pool welding, the dual-molten-pool laser welding process involves dynamic heat exchange between multi-layer heterogeneous structures. The locally magnified evolution characteristics of the temperature field during the dual-molten-pool laser welding of AZ91/6061 dissimilar metals are shown in Figure 13. Overall, the temperature field exhibits significant asymmetric distribution characteristics, which are essentially different from the continuum heat transfer behavior in traditional single-material welding.
Contour plots of temperature field evolution at different time instants during AZ91/6061 dual-molten-pool laser welding: (a) Step 1–0.3 s; (b) Step 1–2.0 s; (c) Step 1–4.0 s; (d) Step 1–4.5 s; (e) Step 2–0.48 s; (f) Step 2–3.2 s; (g) Step 2–6.4 s; (h) Step 2–7.2 s.
During the Step 1 stage, as the laser heat source moves along the Mg-Ni interface toward the positive X-direction, the temperature field exhibits a typical comet-shaped unsteady distribution. At the front of the heat source, the isotherms are densely distributed, indicating an extremely large temperature gradient in this region, with a larger temperature gradient observed on the Mg alloy side. In contrast, at the rear of the heat source, the isotherms gradually become sparse due to the thermal diffusion effect of the solidified metal in the trailing zone. Notably, since the Al-Ni interface is in a state of mechanical contact in the initial stage, the interfacial air gap hinders the heat transfer between the two materials. Consequently, the propagation of heat flux perpendicular to the weld direction presents serrated isotherms at the Al-Ni interface, showing a strong passive cutoff characteristic. The simulation results show that the high-temperature region is mainly confined to the AZ91 side, and heat cannot be effectively transferred across the Ni interlayer to the 6061 side. This heat flux zoning not only facilitates the rapid formation and stabilization of the molten pool on the magnesium alloy side, but also prevents the premature heating of the aluminum alloy base metal during the first welding pass, which would otherwise lead to unintended softening or melting.
During the Step 2 stage, a significant mechanistic reversal occurs in the temperature field evolution behavior. When the laser beam acts on the Al-Ni side, the higher thermal diffusivity of the aluminum alloy enables rapid heat dissipation to the surroundings, resulting in wider isotherms compared to Step 1. At this point, the Mg-Ni interface has achieved metallurgical bonding in Step 1, and the contact thermal resistance is reduced to zero in the simulation, converting it into an efficient thermal conduction channel. However, the Al-Ni interface remains in a to-be-welded non-ideal contact state at the front of the molten pool. Thus, the temperature field exhibits a dynamic on-off characteristic again at the Al-Ni interface: in the laser spot and molten pool-covered areas, interface wetting induces a sudden change in thermal conductivity; in contrast, in areas not covered by the spot, contact thermal resistance still exists, restricting heat transfer across the Al-Ni interface.
It can be concluded that the temperature field evolution of the dual-molten-pool laser welding process is not a simple linear superposition, but rather a complex nonlinear process dominated by the dynamic switching of interfacial contact states. The present model accurately captures the heat flux path reconstruction induced by the transition from mechanical contact to metallurgical bonding, verifies the physical isolation effect of the Ni interlayer in dual-molten-pool welding, and plays a positive role in the precise regulation of heat input during dissimilar metal welding.
3.3. Topological evolution of interfacial contact states
To further clarify the physical essence of the transition of contact states between materials from mechanical contact to metallurgical bonding in dual-molten-pool laser welding, the dynamic evolution contour plots of field variables of thin-layer contact elements at the Mg-Ni interface in Step 1 and the Al-Ni interface in Step 2 were extracted, as shown in Figure 14. In the plots, the red regions represent the fully thermally conductive zones where melting and wetting have occurred, corresponding to the transition of the field variable from FV = 0 to FV = 1, while the blue regions denote the zones where material properties remain unchanged. It should be noted that the blue elements remaining in the large-area color-transition region are initial contact elements, which retain the base metal properties and thus do not participate in the FV transition.
Contour plots of dynamic evolution of interfacial contact states at Al-Ni and Mg-Ni interfaces: (a-a1) Mg-Ni; (b-b1) Al-Ni.
As shown in Figure 14(a), in front of the laser heat source (positive X-direction) during Step 1, the regions with FV = 1 at the Mg-Ni interface exhibit a discrete spot-like distribution, which accurately reflects the microscopically rough contact state mapped by the “Indentation/Staining-Binarization Method”. As the heat source continues to act, when the interfacial temperature exceeds the melting point, the field variable subroutine triggers the welding criterion; the originally discrete contact spots rapidly connect and coalesce, forming a continuous red banded region behind the heat source. This topological mutation from discreteness to continuity intuitively verifies that the model successfully captures the transient physical process of liquid metal filling the microscopic air gaps.
By analogy, Figure 14(b) illustrates the welding process at the Al-Ni interface during Step 2. After the completion of the first weld pass, the Al-Ni interface, though affected by residual heat, still retains its mechanical contact characteristics. It is not until the scanning of the second laser pass and the formation of the molten pool on the aluminum side that the interface truly achieves continuity of physical properties. The accurate simulation of such interfacial states is the key reason why the present model can precisely predict the asymmetric temperature field in dual-molten-pool welding.
3.4. Analysis of regulation mechanism of interfacial contact states on temperature field in dual-molten-pool welding
During the dual-molten-pool laser welding process, thermal conduction is not a simple continuum physical behavior, but a process strictly regulated by discrete non-ideal contact interfaces. At 2.0 s during the Step 1 welding process, the contour plots of the locally magnified transient temperature field, the distribution of selected paths, and the path temperature evolution curves are shown in Figure 15. Path 1 and Path 2 are located on either side of the Mg-Ni weld respectively, Path 3 and Path 4 are located on either side of the Al-Ni weld respectively, and Path 5 is situated on the aluminum plate far from the Al-Ni weld. In-depth analysis of the spatial temperature curves of the selected paths reveals the unique heat transfer phenomenon based on interfacial contact in dual-molten-pool laser welding. During the Step 1 process, the temperature curves of Path 1 and Path 2 exhibit significant spatial differences. In the section where the laser heat source has traversed, the Mg-Ni interface undergoes fusion, and the temperature curves of Path 1 and Path 2 almost completely overlap in the contour plots. From a thermodynamic perspective, this is because the liquid metal completely eliminates the residual air between the microscopic interfaces at high temperatures, establishing atomic-level metallurgical bonding. At this point, the interfacial thermal resistance approaches zero. Such an overlap phenomenon not only verifies the accuracy of the subroutine in the model for switching the welding state, but also objectively reflects the extremely high heat exchange efficiency between the interlayer and the base metal in the dual-molten-pool process.
Local contour plots of transient temperature field and temperature distribution curves of characteristic paths at Step 1–0.2 s: (a) Local temperature field contour plots and characteristic path temperature distribution; (b) Path temperature distribution curves.
In contrast, in the unfused zone at the front end where the laser heat source has not yet reached, the temperature curves of Path 1 and Path 2 exhibit a significant separation characteristic. The temperature at the end of Path 1 is significantly higher than that at the end of Path 2 on the nickel side, and the degree of temperature difference between Path 1 and Path 2 increases as the distance from the center of the welding heat source grows. The essence of this trend stems from the fact that the area at the front of the heat source remains in the initial point-contact mechanical assembly state. In this state, heat transfer is affected by the bottleneck effect caused by microscopic roughness. This high degree of temperature difference at the front of the heat source plays a positive role in the stable spreading of the subsequent molten pool, because it physically preheats the to-be-welded zone and reduces the critical threshold for laser-induced deep penetration.
Under the macroscopic continuum hypothesis, isotherms typically exhibit a smooth transition. However, at the unfused Al-Ni interface, the temperature field presents remarkably inhomogeneous characteristics. During Step 1, the Al-Ni interface is not directly exposed to the laser heat source, and its temperature is entirely driven by indirect heat transfer across the Ni interlayer. The observed corrugated curves here are attributed to the inhomogeneous localization of heat flux induced by microscopic contact, which serves as the most compelling evidence for the physical realism reproduced by the mapping model in this study.
Path 3 is located in the region of the interlayer adjacent to the aluminum alloy side, and its temperature distribution exhibits distinct spatial fluctuation characteristics. Among these features, the valley regions correspond to the experimentally mapped effective contact points at the Al-Ni interface. At these points, heat can be rapidly conducted through metallic contact to the aluminum alloy matrix with extremely high thermal conductivity. Owing to the heat extraction effect of the aluminum alloy, heat at these locations on Path 3 is rapidly transferred, manifesting as local low temperatures. In contrast, the peak regions correspond to the air gaps at the interface. Given the extremely low thermal conductivity of air, heat flux is blocked in these regions; heat is confined within the Ni interlayer without timely outward transfer, leading to heat accumulation at these locations.
Path 4 is located on the aluminum alloy side, and its fluctuation pattern is exactly opposite to that of Path 3, exhibiting a complementary relationship. Its peak regions correspond to the contact spots, where heat is smoothly transferred from the Ni interlayer to the aluminum alloy through the contact points, enabling these locations on Path 4 to receive more heat transfer and thus manifesting as high temperatures. The valley regions correspond to the air gap regions. As the heat transfer path is truncated, these locations on Path 4 receive almost no effective heat supply; heat is mainly transferred from the contact points over a long distance, keeping the temperature at a relatively low level. Such corrugated distribution profoundly reveals the spatial modulation effect of interfacial contact heterogeneity on microscale heat flux. Figure 16 shows the contour plots of the temperature field characterized by heat flux localization at the Al-Ni interface during Step 1, which clearly illustrates the reason for the occurrence of temperature oscillation waveforms highly correlated with the surface roughness distribution in Path 3 and Path 4 of Figure 15. The low-temperature regions accurately correspond to the microscopic contact spots, where heat can be efficiently transferred; in contrast, the high-temperature regions correspond to the air gaps, manifesting as local thermal blockage. This phenomenon reveals that non-ideal contact interfaces impede the uniform thermal conduction, and discretize the heat flux into several microscale channels for transmission. Traditional continuum models fail to simulate such local temperature difference fluctuations due to the assumption of an isotropic and uniform thermal conductivity field in the interfacial region, whereas the present model accurately reproduces the thermal conduction array composed of contact points through discrete mapping.
As the observation line crosses the Al-Ni interface and translates toward the positive Y-axis direction to Path 5, the temperature curve that fluctuated sharply at the interface returns to smoothness again. From the perspective of heat transfer principles, this phenomenon is attributed to the heat flux diffusion and self-equilibrium effect. The extremely high heat transfer performance of the aluminum alloy exerts a filtering effect. When the heat flux enters the aluminum alloy from discrete contact points, lateral diffusion occurs within an extremely short time, eliminating the temperature extremes caused by the interfacial inhomogeneous contact. The smoothness of the temperature curve of Path 5 demonstrates that the significant influence of the interfacial contact state on the molten pool near-field is gradually diluted by the material’s intrinsic high thermal conductivity. It can be concluded that by capturing the heat transfer process from microscopic fluctuation to far-field uniformity, the present model fully reflects the high-confidence physical characteristics of the thermal process in dual-molten-pool welding.
3.5. Response characteristics of transient temperature field based on interfacial state variation
Figure 17 shows the locally enlarged contour plots of the transient temperature field at 2.0 s during Step 1 of the welding process, together with the Y-directional path distribution at intervals of 2 mm and the corresponding path temperature evolution curves. To investigate the transient evolution mechanism of the interfacial heat transfer mechanism with heat source scanning, it can be clearly observed that the dynamic evolution of interfacial thermal conduction during dual-molten-pool laser welding is supported by thermodynamic evidence, through the analysis of temperature evolution of the transverse paths (Path 1–Path 8) along the weld direction in Figure 17.
Transverse path distribution perpendicular to the weld and corresponding temperature evolution curves at 2 s during step 1: (a) Path distribution: Path 1–Path 8; (b) Temperature evolution curves.
In the to-be-welded zone not yet scanned by the heat source, i.e., the location of Path 1 and Path 2, the Mg-Ni interface exhibits intense temperature discontinuity. The simulation data quantitatively show that at Path 1, the temperature jumps instantaneously from 152.4°C on the Ni interlayer side to 184.1°C on the magnesium alloy side. A stepwise temperature difference of 31.7°C and a temperature jump ratio of 20.8% at Path 1 confirm the high thermal resistance characteristic of the initial mechanical contact interface. At this point, the asperities are only partially softened, and contact thermal resistance dominates the local heat flux truncation effect. This thermal blocking mechanism prevents the efficient transfer of heat from the magnesium alloy side to the Ni interlayer, resulting in local heat accumulation, which makes the temperature of Path 1 on the magnesium alloy side significantly higher than that on the Ni interlayer. The temperature gradient on the magnesium alloy side at the front of the molten pool is higher than that on the Ni interlayer side. This indicates that under the action of the heat source, the heat transfer from the magnesium alloy side to the Ni interlayer is incomplete, resulting in heat accumulation on the magnesium alloy side. This phenomenon plays a key role in heat input regulation during the dual-molten-pool laser welding process. The accumulated energy on the magnesium alloy side further enables the base metal to reach the melting threshold more rapidly, which demonstrates the positive effect of contact thermal resistance in assisting molten pool formation.
As the laser heat source moves, Path 3 is passing through the heat source center. The temperature distribution characteristics of the Mg-Ni interface on Path 3 and Paths 4–8 (scanned by the heat source) have undergone qualitative changes compared with those on Paths 1 and 2: the original temperature mutation disappears completely, and the curves exhibit a smooth transition. This is because as the interfacial temperature crosses the solidus line of the AZ91 magnesium alloy, the liquid metal completely fills the original contact gaps under the combined action of vapor pressure and capillary force, thus eliminating the air gap thermal resistance. Such an evolution from discrete contact point contact to continuous liquid-solid wetting metallurgical bonding results in a sharp change in contact thermal resistance. The continuous temperature distribution curves on Paths 4–8 not only verify the accuracy of the welding criterion in the field variable subroutine, but also objectively reflect the thermal equilibrium behavior that after the melting process occurs, a complete metallurgical connection is established between the materials, and heat begins to be efficiently transferred from the weld zone to the interlayer.
Unlike the state transition occurring at the Mg-Ni interface, all Paths 1–8 maintained a distinct temperature mutation characteristic at the Al-Ni interface during Step 1. This is because the second laser heat source was not activated yet, and the Al-Ni interface (Distance = 3 mm) remained in a solid-state mechanical contact state throughout this period. This finding holds significant process value: it reveals that the Ni interlayer not only acted as a chemical barrier during Step 1, but also functioned as a thermal insulator physically, which effectively reduced the unordered energy leakage toward the aluminum alloy side. By capturing the evolution of the temperature difference across the interface, the present model achieves an improvement in the accuracy of heat transfer characterization for multi-layer heterogeneous structures.
Figure 18 shows the locally enlarged contour plots of the transient temperature field at 3.2 s during Step 2 of the welding process, together with the Y-directional path distribution at intervals of 2 mm and the corresponding path temperature evolution curves. In the Step 2 welding process, the laser heat source acts on the Al-Ni interface. By observing the instantaneous temperature field during welding, significantly smoother isotherms can be clearly observed compared with those in the Step 1 process. Since the laser has completed the welding of Mg-Ni in Step 1, the Mg-Ni interface (Distance = 7 mm) has been completely transformed from the initial mechanical contact state to the full metallurgical bonding state. That is, at the beginning of the Step 2 welding process, the Mg-Ni interface maintains the metallurgical bonding formed under the action of the Step 1 heat source. The original microscopic air gaps are filled with liquid metal and solidified to form intergranular bonding, which reduces the simulated interfacial thermal resistance to zero. At this point, when observing the temperature distribution law of Paths 1–8 from right to left, all paths do not exhibit any temperature gradient jumps when crossing the Mg-Ni interface, and the curves present extremely high continuity and smoothness. This phenomenon powerfully demonstrates that in the dual-molten-pool process, the preceding weld provides a stable thermal conduction carrier for the subsequent temperature field distribution, enabling the residual heat backflow to follow the bulk thermal conduction law.
Transverse path distribution perpendicular to the weld and corresponding temperature evolution curves at 3.2 s during Step 2: (a) Path distribution: Path 1–Path 8; (b) Temperature evolution curves.
On the Al-Ni interface side, the curves exhibit two heat transfer modes that dynamically switch with the position of the heat source. On Paths 1 and 2 in the to-be-welded zone ahead of the laser scanning path, the temperature jumps instantaneously from 131.3 °C on the Ni interlayer side to 156.3 °C on the aluminum alloy side when crossing the Al-Ni interface, resulting in a temperature jump of 25 °C and a temperature jump ratio of 19%. The temperature jump ratios along the same paths for the Mg-Ni interface during Step 1 and the Al-Ni interface during Step 2 are comparable. Although the entire component has been thermally affected by the first weld at this point, the Al-Ni interface in the regions of Paths 1 and 2 still maintains the original mechanical contact state. This temperature difference reflects the blocking effect of interfacial asperities on low heat flux density under low load, which is manifested as incomplete thermal conduction. It also indirectly verifies the sensitivity of the high-confidence model in capturing the characteristics of micro-scale interfacial thermal resistance.
It should be noted that prior to the initiation of Step 2, as the Al–Ni interface is affected by heat conduction from the Step 1 laser, the proportion of effective thermal conductivity area ratio at this interface should fall between the values presented in Figure 5(d) and Figure 5(e). In the modeling of this study, this minor irreversible plastic deformation is neglected and slight oxidation of the interface is not accounted for; that is, at the initial stage of Step 2, the proportion of effective thermal conductivity area ratio at the Al–Ni interface is assumed to remain in the 20 °C state.As a consequence, the model will overestimate the contact thermal resistance of the Al–Ni interface at the laser leading edge (positive X-direction) in Step 2, meaning the simulated “thermal barrier” effect will be slightly stronger than the actual situation. Considering both the neglected minor irreversible plastic deformation and unaccounted interfacial oxidation, the resulting deviation in the interfacial temperature jump caused by this treatment will not exceed 5%.
Accurately predicting the proportion of effective thermal conductivity area ratio at the Al–Ni interface and incorporating it into the model is extremely challenging. Furthermore, as indicated by Figure 5(d) and Figure 5(e), the proportion of effective thermal conductivity area ratio at the Al–Ni interface exhibits a slight increase over the low temperature range of 20 °C to 200 °C. Therefore, this process is simplified in the model construction of this study.
Observations of Path 3 passing through the heat source center and Paths 4–8 in the rear zone scanned by the heat source reveal that the temperature mutation phenomenon at the Al-Ni interface also disappears. As the laser energy acts directly on the aluminum side, the temperature of the aluminum alloy rises to the melting point, melts and wets the surface of the Ni layer, achieving interfacial continuity. The transition of the curves from a step-like shape to a smooth profile records in real time the process in which the air gap thermal resistance is instantaneously eliminated during the advancement of the liquid molten pool front.
Figure 19 presents the locally enlarged contour plots of the simulated temperature field under the assumption that the interface between the base metal and the interlayer is in an ideal contact state. The results show that the isotherms at the front of the laser heat source exhibit smooth and continuous distribution throughout the two-step welding process, i.e., no temperature jump characteristics are observed at the interface between the interlayer and the base metal. This temperature distribution pattern is inconsistent with the physical essence of actual welding and cannot support the accurate simulation of the subsequent evolution of stress and strain fields.
Locally enlarged contour plots of the simulated temperature field under ideal contact state: (a) Step 1–2 s; (b) Step 2–3.2 s.
Comparative analysis of Figure 19(a) and Figure 17(a) reveals that the assumption of ideal interfacial contact significantly weakens the thermophysical isolation effect of the Ni interlayer, which is specifically manifested as unordered heat leakage from the heat source to the aluminum alloy side during the Step 1 welding process. These comparative findings demonstrate that incorporating the interfacial contact state into the establishment of the finite element model for dual-molten-pool laser welding is a necessary prerequisite for obtaining correct simulation results. The temperature field redistribution induced by interfacial state switching is the core rationale for the high-confidence simulation to accurately reproduce the thermal history, and this method provides a more reliable simulation approach for dissimilar metal welding research.
4. CONCLUSIONS
This study focuses on the dual-molten-pool laser welding process of AZ91/6061 dissimilar metals. Starting from the fundamental interfacial contact state, it establishes a set of high-confidence modeling methods integrating experimental characterization and numerical calculation, which has been scientifically validated based on the temperature field. Furthermore, relying on this temperature field, an in-depth analysis is conducted on the transformation of interfacial heat transfer mechanisms under the action of the heat source. The main conclusions are as follows:
-
(1)
Interfacial Behavior Characterization: This study characterizes the discrete contact behavior of Mg-Ni and Al-Ni interfaces at different temperatures under constant load, confirming the variation law that the effective contact area increases with rising temperature.
-
(2)
Numerical Modeling: An Indentation/Staining-Binarization Method is developed to achieve the quantitative characterization of interfacial contact states, and the obtained results are accurately mapped into the finite element model of dual-molten-pool laser welding.
-
(3)
Model Reliability: The numerical simulation results are highly consistent with the welding experiments in terms of weld cross-sectional morphology and thermal cycle curves at key points. This indicates that the model can accurately capture and verify the heat input and the spatial distribution of the temperature field, yielding a high-confidence finite element model for the dual-molten-pool laser welding process.
-
(4)
Heat Transfer Mechanism: The simulation results reveal the heat transfer process. In the mechanical contact stage where no metallurgical bonding is formed at the interface, local air gaps lead to a significant “thermal barrier” effect, whereas this “thermal barrier” effect disappears after the interface forms metallurgical bonding with the scanning of the heat source.
5. ACKNOWLEDGEMENTS
This work was supported by the S&T Program of Hebei (No. 25361001D) and Major Special Project of National Key Laboratory of Tianjin (No. 25ZXZSSS00140).
6. DATA AVAILABILITY
The raw data are not publicly available due to confidentiality and intellectual property restrictions, but may be obtained from the corresponding author upon reasonable request.
7. BIBLIOGRAPHY
-
[1] PENG, H., CHEN, D.L., GUO, S.F., et al., “Dissimilar ultrasonic spot welding of rare-earth containing ZEK100 magnesium-to-aluminum alloy with a zinc interlayer: Microstructural evolution and mechanical properties”, Journal of Manufacturing Processes, v. 92, pp. 422–434, 2023. doi: https://doi.org/10.1016/j.jmapro.2023.03.014.
» https://doi.org/10.1016/j.jmapro.2023.03.014 -
[2] CUI, M., ZHENG, Z., DU, P., et al., “Experimental and numerical simulation study of magnesium/aluminium dissimilar metal swing laser welding”, Materials Today. Communications, v. 39, pp. 108813, 2024. doi: https://doi.org/10.1016/j.mtcomm.2024.108813.
» https://doi.org/10.1016/j.mtcomm.2024.108813 -
[3] WU, Z., WU, X., ZHANG, Y., et al., “Effect of nanosecond pulse laser power on welding interface and mechanical properties of AZ31B Mg/6061 Al.”, Optics & Laser Technology, v. 175, pp. 110848, 2024. doi: https://doi.org/10.1016/j.optlastec.2024.110848.
» https://doi.org/10.1016/j.optlastec.2024.110848 -
[4] SHANG, J., WANG, K., ZHOU, Q., et al., “Microstructure characteristics and mechanical properties of cold metal transfer welding Mg/Al dissimilar metals”, Materials & Design, v. 34, pp. 559–565, 2012. doi: https://doi.org/10.1016/j.matdes.2011.05.008.
» https://doi.org/10.1016/j.matdes.2011.05.008 -
[5] MENG, Y., FU, J., GONG, M., et al., “Laser dissimilar welding of Al/Mg lap-joint with Ti interlayer through optimized 8-shaped beam oscillation”, Optics & Laser Technology, v. 162, pp. 109304, 2023. doi: https://doi.org/10.1016/j.optlastec.2023.109304.
» https://doi.org/10.1016/j.optlastec.2023.109304 -
[6] ZHOU, J., JIA, J., HE, L., et al., “Numerical study of heat and mass transfer in dissimilar laser welding of Mg/Al lap-joint with Ti interlayer”, International Journal of Thermal Sciences, v. 199, pp. 108932, 2024. doi: https://doi.org/10.1016/j.ijthermalsci.2024.108932.
» https://doi.org/10.1016/j.ijthermalsci.2024.108932 -
[7] KUMAR, S., WU, C., “Eliminating intermetallic compounds via Ni interlayer during friction stir welding of dissimilar Mg/Al alloys”, Journal of Materials Research and Technology, v. 15, pp. 4353–4369, 2021. doi: https://doi.org/10.1016/j.jmrt.2021.10.065.
» https://doi.org/10.1016/j.jmrt.2021.10.065 -
[8] WANG, H., LIU, L., LIU, F., “The characterization investigation of laser-arc-adhesive hybrid welding of Mg to Al joint using Ni interlayer”, Materials & Design, v. 50, pp. 463–466, 2013. doi: https://doi.org/10.1016/j.matdes.2013.02.085.
» https://doi.org/10.1016/j.matdes.2013.02.085 -
[9] ABDOLLAHZADEH, A., SHOKUHFAR, A., CABRERA, J.M., et al., “The effect of changing chemical composition on dissimilar Mg/Al friction stir welded butt joints using zinc interlayer”, Journal of Manufacturing Processes, v. 34, pp. 18–30, 2018. doi: https://doi.org/10.1016/j.jmapro.2018.05.029.
» https://doi.org/10.1016/j.jmapro.2018.05.029 -
[10] PENG, H., CHEN, D.L., BAI, X.F., et al., “Microstructure and mechanical properties of Mg-to-Al dissimilar welded joints with an Ag interlayer using ultrasonic spot welding”, Journal of Magnesium and Alloys, v. 8, n. 2, pp. 552–563, 2020. doi: https://doi.org/10.1016/j.jma.2020.04.001.
» https://doi.org/10.1016/j.jma.2020.04.001 -
[11] CUI, M., ZHENG, Z., DU, P., et al., “A double molten pool laser welding process for joining magnesium alloys and aluminum alloys: The process, microstructure and mechanical properties of joints”, Materials Today. Communications, v. 35, pp. 106116, 2023. doi: https://doi.org/10.1016/j.mtcomm.2023.106116.
» https://doi.org/10.1016/j.mtcomm.2023.106116 -
[12] SHI, X., LIN, J., XIONG, H., et al., “Practical estimation of the thermal conductivity of granular soils considering grading and relative density using three physical parameters”, Acta Geotechnica, v. 22, n. 2, pp. 1–21, 2025. doi: https://doi.org/10.1007/s11440-024-02431-8.
» https://doi.org/10.1007/s11440-024-02431-8 -
[13] ZHANG, X., SHI, L., LUO, H., et al., “Numerical simulation of heat and mass transfer in novel stepped structure friction stir welding of Ti/Al dissimilar alloys”, International Journal of Heat and Mass Transfer, v. 242, pp. 126803, 2025. doi: https://doi.org/10.1016/j.ijheatmasstransfer.2025.126803.
» https://doi.org/10.1016/j.ijheatmasstransfer.2025.126803 -
[14] YU, H., JIANG, X., ZHANG, M., “An innovative coil for magnetic pulse welding of dissimilar sheet metals: Numerical simulation and experiments”, Journal of Materials Processing Technology, v. 324, pp. 118230, 2024. doi: https://doi.org/10.1016/j.jmatprotec.2023.118230.
» https://doi.org/10.1016/j.jmatprotec.2023.118230 -
[15] GAO, K., CHEN, Q., XUAN, L., et al., “Numerical simulation and experimental investigation on thermal characteristics during induction diffusion welding of Cu/Al dissimilar metals”, International Journal of Thermal Sciences, v. 220, pp. 110335, 2026. doi: https://doi.org/10.1016/j.ijthermalsci.2025.110335.
» https://doi.org/10.1016/j.ijthermalsci.2025.110335 -
[16] YUE, J., ZHOU, H., XIE, C., et al., “Temperature field simulation and asymmetric heat transfer distribution of dissimilar steel welded with external transverse magnetic field”, Materials Today. Communications, v. 37, pp. 107141, 2023. doi: https://doi.org/10.1016/j.mtcomm.2023.107141.
» https://doi.org/10.1016/j.mtcomm.2023.107141 -
[17] XU, W., LI, P., LIU, H., et al., “Numerical simulation of molten pool formation during laser transmission welding between PET and SUS304”, International Communications in Heat and Mass Transfer, v. 131, pp. 105860, 2022. doi: https://doi.org/10.1016/j.icheatmasstransfer.2021.105860.
» https://doi.org/10.1016/j.icheatmasstransfer.2021.105860 -
[18] LI, T., XIE, X., XU, J., et al., “Research on AZ31 Mg alloy/22MnB5 steel pinless friction stir spot welding process and interfacial temperature field simulation”, Journal of Materials Research and Technology, v. 26, pp. 3710–3725, 2023. doi: https://doi.org/10.1016/j.jmrt.2023.08.169.
» https://doi.org/10.1016/j.jmrt.2023.08.169 -
[19] EGUCHI, M., SHIBAMIYA, T., YAMAMOTO, T., “Measurement of real contact area and analysis of stick/slip region”, Tribology International, v. 42, n. 11–12, pp. 1781–1791, 2009. doi: https://doi.org/10.1016/j.triboint.2009.04.046.
» https://doi.org/10.1016/j.triboint.2009.04.046 -
[20] OVCHARENKO, A., HALPERIN, G., VERBERNE, G., et al., “In situ investigation of the contact area in elastic-plastic spherical contact during loading-unloading”, Tribology Letters, v. 25, n. 2, pp. 153–160, 2007. doi: https://doi.org/10.1007/s11249-006-9164-y.
» https://doi.org/10.1007/s11249-006-9164-y -
[21] LIANG, X.M., XING, Y.Z., LI, L.T., et al., “An experimental study on the relation between friction force and real contact area”, Scientific Reports, v. 11, n. 1, pp. 20366, 2021. doi: https://doi.org/10.1038/s41598-021-99909-2. PubMed PMID: 34645959.
» https://doi.org/10.1038/s41598-021-99909-2 -
[22] LI, L.T., LIANG, X.M., XING, Y.Z., et al., “Measurement of real contact area for rough metal surfaces and the distinction of contribution from elasticity and plasticity”, Journal of Tribology, v. 143, n. 7, pp. 071501, 2021. doi: https://doi.org/10.1115/1.4048728.
» https://doi.org/10.1115/1.4048728 -
[23] KALIN, M., JAN, P., “Understanding the friction laws of Amontons and Coulomb by evaluating the real contact area”, Friction, v. 13, n. 1, pp. 9440986, 2025. doi: https://doi.org/10.26599/FRICT.2025.9440986.
» https://doi.org/10.26599/FRICT.2025.9440986 -
[24] LI, L., GONG, J., XIA, H., et al., “Influence of scan paths on flow dynamics and weld formations during oscillating laser welding of 5A06 aluminum alloy”, Journal of Materials Research and Technology, v. 11, pp. 19–32, 2021. doi: https://doi.org/10.1016/j.jmrt.2020.12.102.
» https://doi.org/10.1016/j.jmrt.2020.12.102 -
[25] ZHANG, X., LI, L., CHEN, Y., et al., “Numerical simulation analysis of dual-beam laser welding of tailored blanks with different thicknesses”, Metals, v. 9, n. 2, pp. 135, 2019. doi: https://doi.org/10.3390/met9020135.
» https://doi.org/10.3390/met9020135 -
[26] WHITNEY, M.A., JIN, H., WELLS, M.A., et al., “Thermal analysis and characterization of cast 6061 aluminum alloy microstructures with elevated iron content”, Thermochimica Acta, v. 746, pp. 179951, 2025. doi: https://doi.org/10.1016/j.tca.2025.179951.
» https://doi.org/10.1016/j.tca.2025.179951 -
[27] CHAUDHARY, B., JAIN, N.K., MURUGESAN, J., et al., “Friction stir powder additive manufacturing of Al 6061 alloy: Enhancing microstructure and mechanical properties by reducing thermal gradient”, Journal of Materials Research and Technology, v. 26, pp. 1168–1184, 2023. doi: https://doi.org/10.1016/j.jmrt.2023.07.270.
» https://doi.org/10.1016/j.jmrt.2023.07.270 -
[28] JIANG, L., JIANG, Y., YU, L., et al., “Thermal analysis for brake disks of SiC/6061 Al alloy co-continuous composite for CRH3 during emergency braking considering airflow cooling”, Transactions of Nonferrous Metals Society of China, v. 22, n. 11, pp. 2783–2791, 2012. doi: https://doi.org/10.1016/S1003-6326(11)61533-1.
» https://doi.org/10.1016/S1003-6326(11)61533-1 -
[29] PRASAD, B.L., NEELAIAH, G., KRISHNA, M.G., et al., “Joining of AZ91 Mg alloy and Al6063 alloy sheets by friction stir welding”, Journal of Magnesium and Alloys, v. 6, n. 1, pp. 71–76, 2018. doi: https://doi.org/10.1016/j.jma.2017.12.004.
» https://doi.org/10.1016/j.jma.2017.12.004 -
[30] SUN, J., DILGER, K., “Influence of thermal conductivity on the predicted temperature fields in welding of steels”, International Journal of Thermal Sciences, v. 202, pp. 109066, 2024. doi: https://doi.org/10.1016/j.ijthermalsci.2024.109066.
» https://doi.org/10.1016/j.ijthermalsci.2024.109066 -
[31] LIANG, G., QIN, G., CAO, P., et al., “Evolutions of temperature field and stress field in narrow gap oscillating laser welding process based on equivalent heat source”, Journal of Materials Research and Technology, v. 28, pp. 154–167, 2024. doi: https://doi.org/10.1016/j.jmrt.2023.11.262.
» https://doi.org/10.1016/j.jmrt.2023.11.262 -
[32] KONG, W., HUANG, W., LIU, X., et al., “Experimental and numerical study on welding temperature field by double-sided submerged arc welding for orthotropic steel deck”, Structures v. 56, pp. 104943, 2023. doi: https://doi.org/10.1016/j.istruc.2023.104943.
» https://doi.org/10.1016/j.istruc.2023.104943 -
[33] GIUDICE, F., SILI, A., “A theoretical approach to the residual stress assessment based on thermal field evaluation in laser beam welding”, International Journal of Advanced Manufacturing Technology, v. 123, n. 7–8, pp. 2793–2808, 2022. doi: https://doi.org/10.1007/s00170-022-10247-7.
» https://doi.org/10.1007/s00170-022-10247-7






































