Open Access

The numerical simulations of high power laser keyhole welding at different welding positions are performed by using Volume-Of-Fluid (VOF) method. The main material is SS400. The multi-physics phenomenon is considered using several models, such as the heat flux of Gaussian heat source, the recoil pressure with Clausisus-Clapeyron equation, the Marangoni flow considering temperature gradient, the buoyancy force with Boussinesq approximation, the additional shear stress and heat source due to metallic vapor ejected through keyhole entrance, the bubble formation assumed as adiabatic bubble, and the multiple-reflection by solving proper discriminant, are used. To analyze the fluid flow pattern, the concept of streamline formed by reconstructing the value of the velocity vector is applied. Partial and full penetration cases at different welding positions are considered. The welding position seems to have only a minor influence on bead formation characteristics in both cases. This is probably due to the fact that the recoil pressure has a major influence when compared to other driving forces. The flow characteristics and fluid velocity in weld pool are analyzed to compare the gravity direction effect at different welding positions. It is observed that the clockwise flow pattern is mainly formed by the recoil pressure on the keyhole surface in the case of partial penetration. The laser energy can't maintain the whole weld pool when the weld pool size becomes too large. And then the solidification starts from the middle part of weld pool and a necked weld pool shape is formed. In the full penetration welding, the weld pool flow patterns are affected by the leakage of laser power through the full penetration keyhole and also by surface tension. Furthermore, the numerical simulation of full penetration welding with AISI316L is also performed to analyze the effect of material properties. The weld bead shapes obtained by simulations were compared with the corresponding experimental results to confirm the validity of the process models adopted and the CFD simulation tool.

Electron beam plasma measurement was realised by means of DIABEAM system invented by ISF RWTH Aachen. The Langmuir probe method is used for measurement. The relative simplicity of the method and the possibility of dispersion of high power on the probe allow its application for the investigation of high-power electron beams. The key element of the method is a rotating thin tungsten wire, which intersects the beam transversely on its axis and collects part of the current by itself. The signals, which are registered in the DIABEAM as a voltage, were taken in the form of amplitude. The conversion of the probe current into the distribution along the beam radius was realised using the Abel's method. A voltage-current characteristic was built for the beam current. The local electron density as well as the electron temperature, the floating potential and the plasma potential were measured and calculated by means of this characteristic.

The phenomenon of solidification cracking has been the subject of numerous research projects over the years. Great efforts have been made to understand the fundamentals of hot cracking. It is generally agreed that solidification cracks form in the solidification range between the liquidus and solidus temperatures under the combination of thermal, metallurgical and mechanical factors. There is still a need to determine the time-resolved strain distribution in the cracksensitive region in order to analyse the local critical conditions for solidification cracking phenomena. This was a strong motivation for the development of a measurement system used in this study to estimate the local strains and strain rates in the zone where the solidification crack is expected to occur. The laser beam welding experiments were conducted using the Controlled-Tensile-Weldability test (CTW test) to apply an external strain condition during welding to generate solidification cracks. The CTW test is a test method for investigating the suscep�tibility of laser-welded joints to solidification cracking, in which the sample can be subjected to a defined strain at a defined strain rate during welding. In combination with experimental investigations, numerical simulations provide spatially detailed and time-dependent information about the strain development during the welding process, especially regarding the critical conditions for solidification cracking. Therefore, this tool was also used in the present study to evaluate the accuracy of measurement methods and to estimate experimentally derived values and their concrete influence on the formation of solidification cracks. By integrating experimental methods and numerical simulations, this study investigates the spatially resolved and temporally changing development of strain during welding, with a particular focus on the critical conditions that lead to the formation of solidification cracks.

AbstractLaser beam welding has become widely applied in many industrial fields in recent years. Solidification cracks remain one of the most common welding faults that can prevent a safe welded joint. In civil engineering, convolutional neural networks (CNNs) have been successfully used to detect cracks in roads and buildings by analysing images of the constructed objects. These cracks are found in static objects, whereas the generation of a welding crack is a dynamic process. Detecting the formation of cracks as early as possible is greatly important to ensure high welding quality. In this study, two end-to-end models based on long short-term memory and three-dimensional convolutional networks (3D-CNN) are proposed for automatic crack formation detection. To achieve maximum accuracy with minimal computational complexity, we progressively modify the model to find the optimal structure. The controlled tensile weldability test is conducted to generate long videos used for training and testing. The performance of the proposed models is compared with the classical neural network ResNet-18, which has been proven to be a good transfer learning model for crack detection. The results show that our models can detect the start time of crack formation earlier, while ResNet-18 only detects cracks during the propagation stage.