Ingenieurwissenschaften und zugeordnete Tätigkeiten
Filtern
Erscheinungsjahr
Dokumenttyp
- Zeitschriftenartikel (168)
- Beitrag zu einem Tagungsband (108)
- Vortrag (41)
- Forschungsbericht (6)
- Buchkapitel (1)
- Tagungsband (Herausgeberschaft für den kompletten Band) (1)
- Posterpräsentation (1)
- Forschungsdatensatz (1)
Sprache
- Englisch (236)
- Deutsch (83)
- Russisch (6)
- Italienisch (1)
- Mehrsprachig (1)
Schlagworte
- Laser beam welding (68)
- Additive manufacturing (19)
- Solidification cracking (19)
- Laser metal deposition (18)
- Mechanical properties (13)
- Microstructure (13)
- Numerical simulation (13)
- Additive Manufacturing (11)
- Laser welding (11)
- Laserstrahlschweißen (11)
Organisationseinheit der BAM
- 9 Komponentensicherheit (240)
- 9.3 Schweißtechnische Fertigungsverfahren (236)
- 7 Bauwerkssicherheit (6)
- 7.2 Ingenieurbau (4)
- 9.6 Additive Fertigung metallischer Komponenten (3)
- 5 Werkstofftechnik (2)
- 5.2 Metallische Hochtemperaturwerkstoffe (2)
- 7.7 Modellierung und Simulation (2)
- 1 Analytische Chemie; Referenzmaterialien (1)
- 1.9 Chemische und optische Sensorik (1)
Eingeladener Vortrag
- nein (41)
Numerical analysis of ultrasonic vibration enhanced friction stir welding of dissimilar Al/Mg alloys
(2023)
The ultrasonic vibration enhanced friction stir welding (UVeFSW) process has unique advantages in joining dissimilar Al/Mg alloys. While there are complex coupling mechanisms of multi-fields in the process, it is of great significance to model this process, to reveal the influence mechanism of ultrasonic vibration on the formation of Al/Mg joints.
In this study, the acoustic-plastic constitutive equation was established by considering the influence of both ultrasonic softening and residual hardening on the flow stress at different temperatures and strain rates. And the ultrasonic induced friction reduction (UiFR) effect on friction coefficient in different relative directions at the FSW tool-workpiece interface was quantitatively calculated and analyzed.
The Al/Mg UVeFSW process model was developed through introducing the above acoustic effects into the model of Al/Mg friction stir welding (FSW). The ultrasonic energy is stronger on the aluminum alloy side. In the stirred zone, there is the pattern distribution of ultrasonic sound pressure and energy. The heat generation at the tool-workpiece contact interface and viscous dissipation were reduced after applying ultrasonic vibra-tion. Due to the UiFR effect, the projection of friction coefficient and heat flux distributions at the tool-workpiece interface present a "deformed" butterfly shape. The calculated results show that ultrasonic vibra-tion enhanced the material flow and promoted the mixing of dissimilar materials.
The weld quality and the possible defect formation are directly determined by the weld pool shape and the thermo-fluid dynamics therein. In this paper, an untypical weld pool profile, i.e., elongated at its top and bottom but narrowed at the middle, is found experimentally and numerically in the wire feed laser beam welding. The detrimental influence of the weld pool narrowing on the element transport is analyzed and discussed. A magnetohydrodynamic technique is utilized to suppress the narrowing, aiming at a more homogenous element distribution. It is found that a low-temperature region is formed in the middle of the weld pool due to the interaction of the two dominant circulations from the top and bottom regions. The weld pool is significantly narrowed due to the untypical growth of the mushy zone in the low-temperature region, which results in a direct blocking effect on the downward flow and the premature solidification in the middle region. The Lorentz force produced by a transverse oscillating magnetic field shows the potential to change the flow pattern into a single-circulation type and the low-temperature-gradient region is mitigated.
Therefore, the downward transfer channel is widened, and its premature solidification is prevented. The numerical results are well validated by experimental measurements of metal/glass observation and X-ray fluorescence element mapping.
Through experimental observation and auxiliary numerical simulation, this investigation studies the different types of grain refinement of 5754 aluminum alloy laser beam welding by applying a transverse oscillating magnetic field. Scanning electron microscope results have proved that the application of a magnetic field can reduce the average crystal branch width and increase its number. The interaction between the induced eddy current generated by the Seebeck effect and the applied external magnetic field produces a Lorentz force, which is important for the increase of the number of crystal branches. Based on the theory of dendrite fragmentation and the magnetic field-induced branches increment, the grain size reduction caused by the magnetic field is studied. Furthermore, the effects of the magnetic field are analyzed by combining a phase field method model and simulations of nucleation and grain growth. The grain distribution and average grain size after welding verifies the reliability of the model. In addition, the introduction of a magnetic field can increase the number of periodic three-dimensional solidification patterns. In the intersection of two periods of solidification patterns, the metal can be re-melted and then re-solidified, which prevents the grains that have been solidified and formed previously from further growth and generates some small cellular grains in the new fusion line. The magnetic field increases the building frequency of these solidification structures and thus promotes this kind of grain refinement.
The effect of the oscillating metal vapor plume on the keyhole and molten pool behavior during the laser beam welding of AlMg3 aluminum alloys is investigated by the experimental and numerical method. The real-time height of the metal vapor plume is measured by high-speed camera observation. The obtained experimental results are used to evaluate the additional heating source and laser beam attenuation caused by the scattering and absorption based on the Beer-Lambert theory. Furthermore, the dynamic behavior of the metal vapor plume is incorporated into a 3D transient heat transfer and fluid flow model, coupled with the ray tracing method, for the laser beam welding of the AlMg3 alloy. It is found that the additional heating resulting from the scattered and absorbed laser beam energy by the metal vapor plume significantly expands the shape of the molten pool on the top region. Moreover, the oscillating metal vapor plume caused the fluctuation of the molten pool shape. The probability of keyhole collapse at the bottom increases significantly to 72% due to the oscillating laser power induced by the laser beam attenuation. The internal interplay between the metal vapor plume, molten pool shape, and the keyhole collapse are obtained. The developed model has been validated by the experiments, which shows a good agreement.
The strain field can reflect the initiation time of solidification cracks during the welding process. The traditional strain measurement is to first obtain the displacement field through digital image correlation (DIC) or optical flow and then calculate the strain field. The main disadvantage is that the calculation takes a long time, limiting its suitability to real-time applications. Recently, convolutional neural networks (CNNs) have made impressive achievements in computer vision.
To build a good prediction model, the network structure and dataset are two key factors. In this paper, we first create the training and test sets containing welding cracks using the controlled tensile weldability (CTW) test and obtain the real strain fields through the Lucas–Kanade algorithm. Then, two new networks using ResNet and DenseNet as encoders are developed for strain prediction, called StrainNetR and StrainNetD. The results show that the average endpoint error (AEE) of the two networks on our test set is about 0.04, close to the real strain value. The computation time could be reduced to the millisecond level, which would greatly improve efficiency.
A technique for calculating the main solidification parameters for a two-dimensional columnar crystal growth during complete penetration laser beam welding of thin steel sheets was developed. Given that the weld pool interface is described by Lamé curves (superellipses) within the horizontal plane of growth, general analytical solutions were derived for the geometry of the crystal axis and the corresponding growth rate and cross-sectional area of the crystal. A dimensionless analysis was performed to provide insights on the dependence of the solidification parameters on the shape and dimensions of the rear part of the weld pool boundary. The derived solutions were applied for the case of complete penetration laser beam keyhole welding of 2 mm thick 316L austenitic chromium-nickel steel sheets. It was shown that the reconstruction of the weld pool boundary with Lamé curves provides higher accuracy and flexibility compared to results obtained with elliptical functions. The validity of the proposed technique and the derived analytical solutions was backed up by a comparison of the obtained solutions to known analytical solutions and experimentally determined shapes and sizes of the crystals on the top surface of the sheet. The dimensions of the calculated crystal axis correlated well with the experimentally obtained results.
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.
New developments in nickel-based superalloys and production methods, such as the use of additive manufacturing (AM), can result in innovative designs for turbines. It is crucial to understand how the material behaves during the AM process to advance industrial use of these techniques. An analytical model based on reaction-diffusion formalism is developed to better explain the solidification behavior of the material during laser metal deposition (LMD). The well-known Scheil-Gulliver theory has some drawbacks, such as the assumption of equilibrium at the solid-liquid interface, which is addressed by this method. The solidified fractions under the Scheil model and the pure equilibrium model are calculated using CALPHAD simulations. Differential scanning calorimeter is used to measure the heat flow during the solid-liquid phase transformation, the result of which is further converted to solidified fractions. The analytical model is compared with all the other models for validation.
Durch pulverbasiertes Directed-Energy Deposition lassen sich Gradierungen fertigen, um diskrete Materialübergänge zu vermeiden und die Lebensdauer von Hartschichten zu erhöhen. Die Kombination aus Stahl als Basiswerkstoff und einer verschleiß- und korrosionsbeständigen Co-Cr Legierung verspricht durch Vermeiden von Spannungskonzentrationen das Verhindern von Abplatzungen und Rissen in der Schutzschicht. Um die Qualität des gefertigten Bauteils zu beurteilen, liegen für solche Funktional Gradierten Materialien (FGM) wenig Erkenntnisse vor. Daher wird im Rahmen dieser Studie eine Methodik erarbeitet, um die relative Dichte eines Funktional Gradierten Materials auf Stahl und Co-Cr Basis mittels Maschinendaten zu bestimmen. Anschließend wird unter Einsatz eines künstlichen neuronalen Netzes anhand von Sensordaten die relative Dichte vorhergesagt. Das trainierte Netz erreicht eine Vorhersagegenauigkeiten von 99,83%. Abschließend wird eine Anwendung anhand von einem Demonstrator gezeigt.
Eine typische Automobilkarosserie kann bis zu 5000 Widerstandspunktschweißverbindungen aufweisen, welche hohen Qualitätsanforderungen genügen müssen. Daher ist eine durchgehende Prozessüberwachung unerlässlich. Die Transformation zur E-Mobilität in der Automobilindustrie und die damit einhergehende Reichweitenproblematik treiben die Entwicklung und Einführung neuer hochfester Stähle an. Dies resultiert in einem gesteigerten Fertigungsaufwand hinsichtlich einer stabilen Prozess-führung in der Fügetechnik. Um diesen Anstieg an Komplexität zu bewältigen, sind die Methoden der künstlichen Intelligenz ein geeignetes Mittel. Mit ihnen kann, durch Auswertung der Prozessparameter und -signale, die individuelle Schweißpunktqualität sichergesellt werden. Die Vorhersagegenauigkeit von neuen Daten, also das extrapolieren, stellt für die meisten Algorithmen eine große Herausforderung dar.
In dieser Arbeit wird ein künstliches neuronales Netz zur Vorhersage des Punktdurchmessers von Widerstandspunktschweißungen anhand von Prozessparametern implementiert. Die Vorhersagegenauigkeit und Extrapolationsfähigkeit des Modells wird durch die Auswertung des dynamischen Widerstandssignals verbessert.
Um die Extrapolationsfähigkeit zu untersuchen, wird die Vorhersagegenauigkeit des Modells mit Daten getestet, die sich in Bezug auf den Werkstoff und der Beschichtungszusammensetzung deutlich von den Trainingsdaten unterscheiden. Dazu wurden mehrere Schweißexperimente mit Werkstoffen verschiedener Hersteller durchgeführt und nur ein Teil der Daten in das Training einbezogen. Die Ergebnisse dieser Arbeit verdeutlichen den positiven Einfluss der Prozesssignale auf die Robustheit des Modells und die Skalierbarkeit der Algorithmen
künstlicher neuronaler Netze auf Daten außerhalb des Trainingsraums.