**1. Introduction**

The laser welding process is used widely in international industry. It involves using a laser to assemble various metals (carbon steels, stainless steels, aluminum, and titanium). A laser is a concentrated heat source that can provide a tight and deep weld bead. This process has different advantages over other welding processes: the absence of chamfers and filler material, a high welding speed, a controlled welding depth, a small thermal affected zone, and a low deformation of welded parts. Recently, additive manufacturing processes have begun to occupy an increasingly important place in the industrial world, especially those based on the melting of a powder bed. They consist of producing parts through layer by layer deposition and selective melting of powder layers using a laser. During the melting of a powder bed, different materials can be used: metals, polymers, and ceramics. Metals are by far the most used materials in different industrial environments such as medical, aerospace, automotive, and power.

The laser welding and especially the melting of a powder bed require continuous studying to improve their performance. The experimental approach is often used to study and analyze these thermomechanical processes [1–8]. These studies can be very expensive considering that the processes parameters are numerous [9,10]. In this context, numerical simulation can be an alternative solution. It facilitates the optimization of these parameters and predicting the final characteristics of manufactured parts for a reasonable cost and time.

During the numerical simulation of these thermomechanical processes, several physical phenomena must be considered. These phenomena include the formation of the molten pool, heat transfers, metallurgical transformation, residual stresses, and distortions. Depending on the aims targeted by the simulation, some physical phenomena can be neglected [11]. For example, during numerical modeling that focuses on the formation of the molten pool (thermo-fluid simulation), the mechanical computation (residual stresses and distortions computation) is often overlooked [12–15]. Thermomechanical simulation is generally used to estimate residual stresses and distortions, thus providing very useful information for fatigue lifetime predictions. The thermal computation of this type of simulation is often simplified (the formation of the molten pool is neglected) to reduce the computation time [16–24]. Some studies have attempted to consider most of these physical phenomena. For example, Saadlaoui et al. [25] developed a new strategy to simulate the interaction between the fluid flow in the molten pool and the deformations in the base metal. Therefore, their strategy enables the provision, at the same time, of thermal, fluid, and mechanical results (thermal cycle, molten pool morphology, fluid flow, residual stresses, and deformations). As mentioned above, the numerical simulation can be a useful solution in studying these processes for a reasonable cost and time. However, the numerical simulation is based on different assumptions [11]. For this, a step of validating the numerical models may be necessary. It involves testing the efficiency and reliability of the simulation using experimental or numerical benchmarks. Frequently, researchers prefer to validate their modeling using numerical benchmarks. That is mostly because of the reasonable cost and the accessibility of these benchmarks (contrarily to experimental benchmarks). Here, the problem is that these numerical benchmarks are often simple academic simulations. For example, the sloshing problem was used by Saadlaoui et al. [15] to validate the fluid formulation of the molten pool formation during the laser welding process. It involves tracking the oscillations of a liquid in a container [26]. Therefore, these benchmarks cannot be directly used to validate the results of a laser welding simulation. In this context, several experimental measurements were developed to study and to build a database (experimental benchmarks) of thermomechanical processes [27–32]. These experimental benchmarks can clearly be used to improve and validate numerical simulations.

Some experimental studies have examined the formation of a molten pool during laser welding and the melting of powder bed processes [10,30,33–35]. They have enabled the measuring of the molten pool morphology. This may assist in validating the thermo-fluid simulation of these processes. Gao et al. [36] studied the formation of a molten pool and the keyhole during ND: YAG laser welding of stainless steel. They applied a coaxial visual sensing system to record images of the molten pool. The same system was used by Kim et al. [29] to observe the morphology of the molten pool and keyhole during a remote laser welding. The effect of the welding speed on the molten pool morphology during a high-power laser welding was studied by Li et al. [37]. Lei et al. [32] used a high-speed camera synchronized with an auxiliary laser light to study the effect of laser welding parameters on the molten pool morphology. The same system was used by Trapp et al. [10] to follow the formation of the molten pool during the melting of a powder bed. These authors and others have studied only the formation of the molten pool. In addition, the molten pool images of these studies (especially for the melting of a powder bed) are not always of sufficient quality to determine the molten pool morphology [10,38–42]. Therefore, these studies cannot be used to validate a complete simulation (thermo-fluid-mechanical simulation) of thermomechanical processes. Indeed, information on the temperature field and the residual stresses distribution may be necessary for such validation. These physical quantities were measured separately by some studies [27,31]. Finding an experimental study that measured the molten pool morphology, the temperature field, the residual stresses, and the distortions at the same time is very rare.

In this paper, an experimental approach is proposed to study laser welding. It involves instrumenting this process to measure different physical quantities (molten pool morphology, temperature field, residual stresses, and distortions). These quantities can be used to build a database (experimental benchmark) to validate the numerical simulation of this process. In addition, the formation of the molten pool during the melting of a powder bed process was also studied using the same experimental approach. Here, the aim was to analyze and understand this recent process better.

The paper is organized as follows:


#### **2. Experiment Setup**

#### *2.1. Proposed Approach*

A specific experimental setup (Figure 1) was designed and realized using different instrumentation tools (high-speed camera, infrared camera, etc.).

For laser welding by conduction, the numerical simulation is well advanced and controlled. It enables the consideration of the physical phenomena during this process [11,14,15,43]. For this, the main objective of welding measurements is to acquire the necessary results to validate the numerical simulation of this process. However, obtaining results of all physical quantities using the experimental approach is not always easy. Hence, we will focus on the weld pool morphology, the temperature field, residual stresses, and distortions.

Contrary to laser welding, the numerical simulation of the melting of a powder bed is not sufficiently advanced [44–47]. Thus, the first interest of the experimental study was to understand the physical phenomena involved better. This can help to improve the numerical simulation of this process. We note that the same experimental setup was used to study both processes (laser welding and melting of a powder bed).

**Figure 1.** Experiment setup: 1—High-speed camera; 2—Infrared camera; 3—Laser fiber; 4—Laser head; 5—Pyrometer; 6—Auxiliary laser light; 7—Samples support.

#### *2.2. Experimental Equipment and Samples*

#### 2.2.1. Machine and Laser

A TRUMPF LASMA 1054 CNC machine (3 axis) was used (Figure 1). A PRECITEC YC52 laser head was powered by a LASERLINE LDM 3000-60 diode laser with a theoretical power of 3 kW and a wavelength of 930 nm. A real laser power equal to 2.5 kW was measured using a wattmeter. The beam analysis of this laser was carried out for the maximum power (2.5 kW) and the results are given in Figure 2.

This analysis enables us to position the sample in the focal plane. Initially, we decided that the upper surface of the sample is merged with this plane (Figure 3a). This configuration causes the formation of a keyhole that is not often considered during numerical simulations [15]. For this reason, the laser beam was defocused by 5 mm (Figure 3b) to conduct a laser welding by conduction. This second configuration give a laser beam diameter of 2.4 mm. The energy distributions of the laser beam of these two configurations are given by Figure 3c,d. The formation of the molten pool during the two welding configurations is presented in Figure 4.

#### 2.2.2. High-Speed Camera and Auxiliary Laser Light

A Phantom VEO 710 high-speed camera with a resolution of 1280 × 800 pixels was used. With a maximal rate of 7400 *f ps*, this camera model is perfect to track the formation of the molten pool. Phantom Camera Control (PCC) software was used to configure, control, process, and download images. A filter with a wavelength of 810 nm was installed on the camera lens (NAVITAR ZM 6000-II, 12 mm FF). The camera was also synchronized with an auxiliary laser light (CAVILUX diode laser with a wavelength of 810 nm and a maximal power of 500 W). This facilitated the capturing of clear images of the molten pool surface (without the light of the laser power). As shown in Figure 1, the camera was attached to the laser head to monitor the molten pool movement. The working distance (the distance between the camera lens and the tip of the power laser) was approximately 100 mm and the tilt angle of the camera from the vertical was approximately 30◦. We note that a reference sample was used to determine the scale of the camera images.

**Figure 2.** Laser beam analysis for maximal power of 2.5 kW.

**Figure 3.** (**a**) Focused laser beam; (**b**) Defocused laser beam of 5 mm; (**c**) Energy distribution of focused laser beam; (**d**) Energy distribution of defocused laser beam.

**Figure 4.** Two welding configuration (images of high-speed camera): (**a**) welding by keyhole; (**b**) welding by conduction (P = 2.5 KW, v = 600 mm.min−1).

#### 2.2.3. Infrared Camera and Pyrometer

A FLIR Phoenix MWIR infrared camera with a resolution of 320 × 256 pixels was used to determine the temperature distribution in the weld bead. As Figure 1 shows, the camera was attached to the machine table. Its working distance and tilt angle were in the order of 695 mm and 29◦, respectively. With this configuration, the camera facilitated the capturing of the temperature distribution throughout the weld bead. An attenuator filter was installed on the camera lens to capture temperatures approximately 1600 ◦C. An integration time of 0.01 ms and acquisition time of 10 μs were chosen.

We note that the infrared camera was not calibrated, so it could not provide temperature values. For this, a KLEIBER LVO 25 S-7 monochromatic pyrometer was used to measure the temperature in the weld bead. It can measure temperatures between 350 and 3500 ◦C. A specific mounting was used to install the pyrometer on the laser head (Figure 1). This specific mounting expedited moving the pyrometer and measuring the temperature in four positions (P1, P2, P3, and P4) on the weld bead (Figure 5). A working distance of approximately 123 mm and a length of the pyrometer body of 61 mm enabled the measuring of the temperature with a spot diameter of 0.9 mm. We note that the emissivity of the material was assumed to be 1. Therefore, the temperature given by the pyrometer was the brightness temperature.

**Figure 5.** Measurement points of the pyrometer.

#### 2.2.4. Optical Microscope, DRC Machine, and 3D Measuring Arm

The length and width of the weld pool can be measured using the images of the high-speed camera. However, to determine its depth and height, the use of another approach is necessary. It involves making samples by transversely cutting the welded samples. These samples must then undergo a polishing step. Subsequently, a chemical attack must be conducted to distinguish the different zones: weld bead, heat affected zone (HAZ), and the base metal. The final step is to visualize the samples using an optical microscope.

A PROTO DRX machine (with a collimator diameter of 1 mm) was used to measure residual stresses. The detector uses two orientations, 0◦ and 90◦, to determine the axial and longitudinal residual stresses. For each of these two orientations, the scanned angles are −25◦; −16.66◦; −8.33◦; 0◦; 8.33◦; 16.66◦, and 25◦.

A FARO GAGE 3D measuring arm (with a precision of 0.025 mm) was used to measure the flatness of the samples after welding. This measuring enabled us to determine the vertical displacements *Uz* of the welded samples (Figure 6).

**Figure 6.** Vertical displacement *Uz* of welded sample: transverse plane.

#### 2.2.5. Samples Preparation

For the laser welding, 316L stainless steel plates (austenitic steel) from hot rolling were machined to obtain nine samples with dimensions of 100 mm × 100 mm × 10 mm. The chemical composition of this material is given in Table 1. These samples were subjected to a relaxation heat treatment at 610 ◦C for 2.5 h with slow cooling in a neutral atmosphere (nitrogen). This heat treatment facilitates the reduction of the residual stresses related to the machining and rolling processes. Thus, this enables avoiding considering the initial residual stresses during the numerical simulation of the laser welding process. The first measurement of stresses (after machining of samples and before the heat treatment) indicated values in the order of 700 MPa while values of approximately 100 MPa were observed after this heat treatment. Samples in 15CDV6 low alloy steel (bainitic steel) were also machined (with the same dimensions). The aim here was to study the formation of the weld pool of two materials that do not have the same thermophysical properties.

**Table 1.** Chemical composition of 316L stainless steel in percent.


The melting of a powder bed involved filling a hollow 316L substrate of dimensions 60 mm × 60 mm × 10 mm (Figure 7) by a layer of 316L powder. Subsequently, the power laser facilitated the formation of a bead by melting a zone of this layer (thickness of 2 mm). Two types of powder were examined to study the influence of the size and granulometry of the powder on the formation of the molten pool: a spherical powder with a diameter between 20 and 50 μm, and another irregular one with a size between 100 and 150 μm (Figure 8).

**Figure 7.** Preparation of the powder layer.

**Figure 8.** Granulometry of 316L powder: (**a**) spherical powder (20–50 μm); (**b**) irregular powder (100–150 μm).
