TY - CONF A1 - Neumann, B. A1 - Gottschalk, G.-F. A1 - Biegler, M. A1 - Goecke, S.-F. A1 - Rethmeier, Michael T1 - Detektion von Bindefehlern mittels neuronaler Netze bei DED-Arc/M von Aluminium anhand von Echtzeit- Schweißstromquellendaten N2 - Mit der Anwendung des Schutzgasschweißens für additive Strukturen wird aufbauend auf Machine-Learning-Modellen, welche bereits zum Überwachen beim Verbindungsschweißen erforscht wurden, ein tiefes neuronales Netz (DNN) zum Monitoring beim DED-Arc/M von Aluminium vorgestellt. Ziel des Machine Learning Modells ist das Identifizieren von Bindefehlern in den aufgebauten Volumina mit prozessbegleitend gemessenen Schweißstromquellensignalen als Input. Es werden durch Algorithmen Merkmalsvariablen in der Vorverarbeitung der Daten extrahiert und die Korrelation zwischen den Merkmalsvariablen und den Bindefehlern analysiert. Durch den vorgestellten Algorithmus werden diese automatisiert als Input an ein DNN übergeben. Diese Arbeit untersucht die Genauigkeit der Klassifizierung von verschiedenen DNN-Architekturen mit jeweils 4 verdeckten Schichten. Als Trainings- und Testsatz werden randomisiert extrahierte Merkmale von defektfreien und fehlerhaften Proben verwendet, wobei Bindefehler zum Teil absichtlich provoziert werden. Das entwickelte neuronale Netz erkennt anhand signifikanter Merkmale aus den Strom- und Spannungsdaten Bindefehler mit einer Genauigkeit von ca. 90%. T2 - 45. Assistentenseminar Füge- und Schweißtechnik CY - Niederaudorf, Germany DA - 09.11.2024 KW - DED-Arc/M KW - Aluminiumschweißen KW - Prozessüberwachung KW - Machine Learning KW - Deep Neural Network KW - Bindefehler PY - 2025 SP - 139 EP - 144 PB - DVS Media GmbH AN - OPUS4-64838 LA - deu AD - Bundesanstalt fuer Materialforschung und -pruefung (BAM), Berlin, Germany ER - TY - JOUR A1 - Neumann, Benedikt A1 - Biegler, Max A1 - Goecke, Sven-Frithjof A1 - Rethmeier, Michael T1 - Detektion von Bindefehlern beim DED-Arc T1 - Detection of Lack of Fusion during DED-Arc N2 - Based on artificial intelligence (AI) developed for monitoring arc welding, this article presents a deep neural network for monitoring lack of fusion defects in wire arc additive manufacturing of aluminium. The aim is to detect defects in built-up volumes on the basis of weld source data. These can be successfully processed by the algorithm presented and a trained AI. The achieved accuracy of the network is 90 percent. N2 - In dem Beitrag wird aufbauend auf Machine-Learning-Modellen, welche bereits zum Überwachen des Schutzgasschweißen erforscht wurden, ein tiefes neuronales Netz (DNN) zum Monitoring beim DED-Arc von Aluminium vorgestellt. Ziel ist die Detektion von Bindefehlern in den aufgebauten Volumina auf Grundlage von in Echtzeit gemessenen Schweißstromquellensignalen. Es werden Merkmalsvariablen durch Vorverarbeitung extrahiert sowie die Korrelation zwischen den Merkmalsvariablen und den Defekten analysiert. Durch den vorgestellten Algorithmus werden diese automatisiert als Input an ein DNN übergeben. Das entwickelte und trainierte neuronale Netz erkennt anhand signifikanter Merkmale aus den Strom- und Spannungsdaten Bindefehler mit einer Genauigkeit von ca. 90 Prozent. KW - DED-Arc KW - Aluminiumschweißen KW - Bindefehler KW - Prozessüberwachung KW - Maschinelles Lernen KW - Deep Neural Network PY - 2024 DO - https://doi.org/10.1515/zwf-2024-1107 VL - 8 SP - 577 EP - 583 PB - Walter de Gruyter GmbH AN - OPUS4-61630 LA - deu AD - Bundesanstalt fuer Materialforschung und -pruefung (BAM), Berlin, Germany ER - TY - JOUR A1 - Huo, Wenjie A1 - Bakir, Nasim A1 - Gumenyuk, Andrey A1 - Rethmeier, Michael A1 - Wolter, Katinka T1 - Detection of solidification crack formation in laser beam welding videos of sheet metal using neural networks N2 - 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. KW - Artificial Intelligence KW - Software PY - 2023 UR - https://nbn-resolving.org/urn:nbn:de:kobv:b43-586116 DO - https://doi.org/10.1007/s00521-023-09004-y SN - 0941-0643 VL - 35 IS - 34 SP - 24315 EP - 24332 PB - Springer Science and Business Media LLC AN - OPUS4-58611 LA - eng AD - Bundesanstalt fuer Materialforschung und -pruefung (BAM), Berlin, Germany ER - TY - JOUR A1 - Fabry, Çağtay A1 - Pittner, Andreas A1 - Rethmeier, Michael T1 - Design of neural network arc sensor for gap width detection in automated narrow gap GMAW N2 - An approach to develop an arc sensor for gap width estimation during automated NG-GMAW with a weaving electrode motion is introduced by combining arc sensor readings with optical measurements of the groove shape to allow precise analyses of the process. The two test specimen welded for this study were designed to feature a variable groove geometry in order to maximize efficiency of the conducted experimental efforts, resulting in 1696 individual weaving cycle records with associated arc sensor measurements, process parameters and groove shape information. Gap width was varied from 18 mm to 25 mm and wire feed rates in the range of 9 m/min to 13 m/min were used in the course of this study. Artificial neural networks were applied as a modelling tool to derive an arc sensor for estimation of gap width suitable for online process control that can adapt to changes in process parameters as well as changes in the weaving motion of the electrode. Wire feed rate, weaving current, sidewall dwell currents and angles were defined as inputs to calculate the gap width. The evaluation of the proposed arc sensor model shows very good estimation capabilities for parameters sufficiently covered during the experiments. KW - GMA welding KW - Narrow gap welding KW - Sensor KW - Neural networks KW - Adaptive control PY - 2018 DO - https://doi.org/10.1007/s40194-018-0584-8 SN - 1878-6669 SN - 0043-2288 VL - 62 IS - 4 SP - 819 EP - 830 PB - Springer CY - Berlin AN - OPUS4-45234 LA - eng AD - Bundesanstalt fuer Materialforschung und -pruefung (BAM), Berlin, Germany ER - TY - CONF A1 - Putra, Stephen Nugraha A1 - Meng, Xiangmeng A1 - Bachmann, Marcel A1 - Rethmeier, Michael T1 - Der Einfluss verschiedener räumlicher Diskretisierungsansätze des Ray-Tracing-Verfahrens bei der Simulation des Laserstrahltiefschweißen N2 - Die Wärmeverteilung des Lasers beim Laserstrahltiefschweißen ist für die Formgebung der Dampfkapillare und für die Schweißbaddynamik entscheidend. In dieser Arbeit werden die Laserwärmeverteilung und deren Einflüsse auf die Schweißbadtiefe sowie -breite numerisch anhand des Ray-Tracing-Verfahrens analysiert. Hierbei wird der La-serstrahl in mehreren Strahlenbündeln bzw. Subrays unterteilt. Diesbezüglich soll der Pfad der Subrays präzis be-rechnet werden, um die Dynamik der Dampfkapillare und des Schweißbades eines realen Schweißprozesses rich-tig abzubilden. Zu diesem Zweck beschäftigt sich die vorliegende Arbeit mit der Genauigkeitsverbesserung der Kontaktposition und der Reflexionsrichtung der Subrays auf der freien Oberfläche anhand der Level-Set-Methode. Um die Güte dieses Simulationsansatzes zu gewährleisten, wurde eine Gegenüberstellung mit den zwei klassischen Ray-Tracing-Verfahren mittels drei verschiedenen Benchmark-Testreihen durchgeführt. Anschließend wurden die Versuchsergebnisse zur Validierung der implementierten numerischen Ansätze verwendet. Im Rahmen dieser Arbeit kann es gezeigt werden, dass unterschiedliche Wärmeverteilung aufgrund der verschiedenen Ray-Tracing-Verfahren deutlich zu erkennen ist, welche wiederum die Schweißbaddynamik sowie die lokalisierte Dampfkapil-lardynamik stark beeinflussen. Ferner wurde es bestätigt, dass die implementierte Level-Set-Methode zu einer ge-naueren Ermittlung der Kontaktposition und der Reflexionsrichtung der Subrays und somit zu einer Verbesserung der simulierten Schmelzkontur führt. T2 - 43. Assistentenseminar Füge- und Schweißtechnik CY - Schwarzenberg/Erzgebirge, Germany DA - 27.09.2022 KW - Laser beam welding KW - Ray tracing method KW - Weld pool dynamics KW - Numerical modeling PY - 2023 UR - https://www.dvs-media.eu/de/neuerscheinungen/4584/43.-assistentenseminar-fuegetechnik SN - 978-3-96144-212-6 VL - 386 SP - 91 EP - 102 PB - DVS Media GmbH CY - Düsseldorf AN - OPUS4-59119 LA - deu AD - Bundesanstalt fuer Materialforschung und -pruefung (BAM), Berlin, Germany ER - TY - CONF A1 - Rethmeier, Michael T1 - Datenbasierte Regelalgorithmen für das automatisierte MSG-Engspaltschweißen auf Basis vernetzter Sensorik N2 - Digital vernetzte Sensorik für Erfassung von Prozessgrößen - Datenbankstruktur zur Integration sämtlicher Messgrößen - In-situ Korrelation zwischen Prozessparametern und Schweißergebnis auf Basis künstlicher Intelligenz - Sukzessiver Aufbau an Prozesswissen selbstlernendes System - Prozessregelung MSG-Engspaltschweißen -Adaptiver Pendelwinkelkeine Vorgabe durch Anwender notwendig gleichmäßiger Flankeneinbrand -Füllgradregelung Vorgabe einer Zielaufbauhöhe durch Anwender autarke Regelung der Prozessgrößenfür gleichmäßige Füllhöhe bei Variation der Spaltbreite T2 - Schweißen und Schneiden 2017 CY - Düsseldorf, Germany DA - 25.09.2017 KW - Datenbankstruktur KW - Prozessparametern KW - selbstlernendes PY - 2017 AN - OPUS4-43531 LA - deu AD - Bundesanstalt fuer Materialforschung und -pruefung (BAM), Berlin, Germany ER - TY - JOUR A1 - El-Batahgy, A.M. A1 - Gumenyuk, Andrey A1 - Gook, S. A1 - Rethmeier, Michael T1 - Comparison between GTA and laser beam welding of 9%Ni steel for critical cryogenic applications N2 - IncomparisonwithGTAweldedjoints,highertensilestrengthcomparabletothatofthebasemetalwasobtained for laser beam welded joints due to fine martensitic microstructure. Impact fracture toughness values with much lower mismatching were obtained for laser beam welded joints due to similarity in the microstructures of its weld metal and HAZ. In this case, the lower impact fracture toughness obtained (1.37J/mm2) was much higher than that of the GTA welded joints (0.78J/mm2), which was very close to the specified minimum value (≥0.75J/mm2). In contrast to other research works, the overall tensile and impact properties are influenced not only by the fusion zone microstructure but also by the size of its hardened area as well as the degree of its mechanical mismatching, as a function of the welding process. A better combination of tensile strength and impact toughness of the concerned steel welded joints is assured by autogenous laser beam welding process. KW - Impact absorbed energy KW - 9%Ni steel KW - GTAW KW - Laser beam welding KW - Fusion zone size KW - Microstructure Tensile strength PY - 2018 DO - https://doi.org/10.1016/j.jmatprotec.2018.05.023 SN - 0924-0136 IS - 261 SP - 193 EP - 201 PB - Elsevier AN - OPUS4-45776 LA - eng AD - Bundesanstalt fuer Materialforschung und -pruefung (BAM), Berlin, Germany ER - TY - CONF A1 - Gumenyuk, Andrey A1 - Rethmeier, Michael A1 - Bakir, Nasim T1 - Comparative study of hot cracking susceptibility for laser welded joints by means of a self- restraint and an externally loaded hot cracking tests N2 - Over the past decade, laser beam welding has significantly evolved and established itself as an efficient tool in the industry. Solidification cracking and the weldability of materials have been highly contentious issues for many years. Today, there are many self and externally loaded tests to investigate the hot cracking resistance of steels. The purpose of this paper is to compare the susceptibility of three stainless steel grades to hot cracking by using an externally loaded hot cracking test (CTW) and a self-restraint test in accordance with SEP-220-3. The repeatability and effectiveness of the results are discussed. The experimental results are widely dispersed, implying a low predictive value for the self-restraint test. On the other hand, the results from the externally loaded test exhibit excellent repeatability and provide a quantitative characterization of the susceptibility of steels to hot cracking. T2 - The 5th International Conference on Steels in Cars and Trucks CY - Amsterdam-Schiphol, Netherlands DA - 19.06.2017 KW - Externally loaded test KW - Hot cracking test KW - SEP-1220-3 KW - CTW test KW - Self-restraint test PY - 2017 SP - 1 EP - 8 AN - OPUS4-41174 LA - eng AD - Bundesanstalt fuer Materialforschung und -pruefung (BAM), Berlin, Germany ER - TY - JOUR A1 - Savitsky, Viktor A1 - Schmies, Lennart A1 - Gumenyuk, Andrey A1 - Rethmeier, Michael T1 - Comparative performance of DIC and optical flow algorithms for displacement and strain analysis in laser beam welding N2 - The measurement of strain and displacement in the context of the welding process represents a significant challenge. Optical methods, such as digital image correlation (DIC) or optical flow algorithms, have demonstrated their efficacy in robust and reliable data acquisition in harsh environments, including those encountered in welding processes. Concurrently, a trade-off between the accuracy of the measurement and the computational resources required for the associated calculations must be evaluated on a case-by-case basis. The application of filters to initial images represents a technique that serves to enhance the quality and accuracy of the strain and displacement prediction. In the present study, the estimated error of two algorithms, namely the Lucas-Kanade (LK) and the inverse compositional Gauss-Newton (IC-GN), is compared on the basis of both synthetic and real welding images. The displacement field is evaluated for different zones in the laser weld seam with varying contrast performance. Based on the aforementioned results, a strain calculation was conducted for both methods, which yielded comparable results for the LK and IC-GN algorithms. KW - Laser speckle KW - DIC KW - Optical flow KW - Error estimation KW - Strain measurement KW - Laser beam welding PY - 2025 UR - https://nbn-resolving.org/urn:nbn:de:kobv:b43-625919 DO - https://doi.org/10.1016/j.optlaseng.2025.108870 SN - 1873-0302 VL - 187 SP - 1 EP - 15 PB - Elsevier Ltd. AN - OPUS4-62591 LA - eng AD - Bundesanstalt fuer Materialforschung und -pruefung (BAM), Berlin, Germany ER - TY - CONF A1 - Graf, B. A1 - Schuch, M. A1 - Petrat, T. A1 - Gumenyuk, Andrey A1 - Rethmeier, Michael T1 - Combined laser additive manufacturing with powderbed and powder nozzle for turbine parts N2 - Metal additive manufacturing is often based on laser beam processes like Laser Metal Fusion (LMF) or Laser Metal Deposition (LMD). The LMF process is in particular suitable for very complex geometries. However build rate, part volume and material flexibility are limited in LMF. In contrast, LMD achieves higher deposition rates, less restricted part sizes and the possibility to change the material composition during the build-up process. On the other hand, due to the lower spatial precision of the material deposition process, the complexity of geometries is limited. Therefore, combined manufacturing with both LMF and LMD has the potential to utilize the respective advantages of both technologies. In this paper, combined additive manufacturing with LMF and LMD is described for Ti-6Al-4V and Inconel 718. First, lattice structures with different wall thickness and void sizes are built with LMF. The influence of LMD material deposition on these LMF-structures is examined regarding metallurgical impact and distortion. Cross-sections, x-ray computer tomography and 3D-scanning results are shown. For the titanium alloy specimen, oxygen and Nitrogen content in the deposited material are analysed to evaluate the LMD shielding gas atmosphere. The results are used to develop guidelines for a LMD build-up strategy on LMF substrates. With these findings, a gas turbine burner is manufactured as reality test for the combined approach. T2 - 6th International Conference on Additive Technologies iCAT 2016 CY - Nürnberg, Germany DA - 29.11.2016 KW - Ti-6Al-4V KW - Inconel 718 KW - Combined laser manufacturing PY - 2016 SN - 978-691-285-537-6 SP - 317 EP - 323 AN - OPUS4-38702 LA - eng AD - Bundesanstalt fuer Materialforschung und -pruefung (BAM), Berlin, Germany ER -