TY - CONF A1 - Zia, Ghezal Ahmad Jan A1 - Torres, Benjami Moreno A1 - Skrotzki, Birgit A1 - Bayerlein, Bernd T1 - Quantitative Precipitate Analysis of an Age-Hardenable Aluminium Alloy Using a Deep Learning Approach N2 - Mechanical properties of metals and their alloys are strongly governed by their microstructure. The nanometer-sized precipitates in hardenable wrought aluminium alloys, which can be controlled by heat treatment, act as obstacles to dislocation movement within the material and are critical to the mechanical performance of the component, in this case a radial compressor wheel of a ships’ engine. TEM-based image analysis is essential for the study to investigate the microstructural changes (precipitation coarsening) that occur as a result of ageing at elevated temperatures. T2 - MSE 2022 CY - Darmstadt, Germany DA - 27.09.2022 KW - Aluminium Alloy KW - Deep Learning KW - TEM Image PY - 2022 UR - https://dgm.de/mse/2022/ AN - OPUS4-55949 LA - eng AD - Bundesanstalt fuer Materialforschung und -pruefung (BAM), Berlin, Germany ER - TY - JOUR A1 - Schmies, Lennart A1 - Botsch, B. A1 - Le, Q.-H. A1 - Yarysh, A. A1 - Sonntag, U. A1 - Hemmleb, M. A1 - Bettge, Dirk T1 - Classification of fracture characteristics and fracture mechanisms using deep learning and topography data T1 - Klassifikation von Bruchmerkmalen und Bruchmechanismen mittels Deep Learning und unter Verwendung von Topographiedaten N2 - In failure analysis, micro-fractographic analysis of fracture surfaces is usually performed based on practical knowledge which is gained from available studies, own comparative tests, from the literature, as well as online databases. Based on comparisons with already existing images, fracture mechanisms are determined qualitatively. These images are mostly two-dimensional and obtained by light optical and scanning electron imaging techniques. So far, quantitative assessments have been limited to macrocopically determined percentages of fracture types or to the manual measurement of fatigue striations, for example. Recently, more and more approaches relying on computer algorithms have been taken, with algorithms capable of finding and classifying differently structured fracture characteristics. For the Industrial Collective Research (Industrielle Gemeinschaftsforschung, IGF) project “iFrakto” presented in this paper, electron-optical images are obtained, from which topographic information is calculated. This topographic information is analyzed together with the conventional 2D images. Analytical algorithms and deep learning are used to analyze and evaluate fracture characteristics and are linked to information from a fractography database. The most important aim is to provide software aiding in the application of fractography for failure analysis. This paper will present some first results of the project. N2 - Die mikrofraktographische Analyse von Bruchflächen wird in der Schadensanalyse meist auf der Basis von Erfahrungswissen vorgenommen, welches aus vorliegenden Untersuchungen, eigenen Vergleichsversuchen und aus der Literatur und online Datenbanken stammt. Durch Vergleiche mit bereits vorliegenden Bildern werden qualitativ Bruchmechanismen ermittelt. Grundlage dafür sind zumeist zweidimensionale Aufnahmen aus licht- und elektronenoptischen Verfahren. Quantitative Aussagen beschränken sich bislang beispielsweise auf makroskopische Anteile von Bruchmechanismen oder die manuelle Ausmessung von Schwingstreifen. In jüngerer Zeit gibt es vermehrt Ansätze, Computer-Algorithmen einzusetzen, die in der Lage sind, unterschiedlich strukturierte Bruchmerkmale zu finden und zu klassifizieren. Im hier vorgestellten IGF-Vorhaben „iFrakto“ werden elektronenoptische Aufnahmen erzeugt und daraus Topographie-Informationen berechnet. Diese gewonnenen Topographie-Informationen werden zusammen mit den klassischen 2D-Bildern ausgewertet. Analytische Algorithmen und Deep Learning werden eingesetzt, um Bruchmerkmale zu analysieren, zu bewerten und mit Informationen aus einer fraktographischen Datenbank zu verknüpfen. Wichtigstes Ziel ist die Bereitstellung von Software zur Unterstützung der Fraktographie in der Schadensanalyse. In diesem Beitrag werden erste Ergebnisse des Vorhabens vorgestellt. KW - Fractography KW - Deep Learning KW - Classification PY - 2023 U6 - https://doi.org/10.1515/pm-2022-1008 SN - 0032-678X VL - 60 IS - 2 SP - 76 EP - 92 PB - De Gruyter AN - OPUS4-56962 LA - mul AD - Bundesanstalt fuer Materialforschung und -pruefung (BAM), Berlin, Germany ER - TY - JOUR A1 - Schmies, Lennart A1 - Hemmleb, Matthias A1 - Bettge, Dirk T1 - Relevant input data for crack feature segmentation with deep learning on SEM imagery and topography data N2 - Fractography plays a critical role in failure analysis of engineering components and has a considerable importance for safety investigations. Usually, the interpretation of fracture surfaces is done by experts with the help of literature and experimental data, that requires a lot of experience. The use of deep learning (DL) with neural networks in failure analysis becomes more and more relevant with the rapidly developing possibilities. Especially, the modern network architectures can assist fractographers in determining various fracture features on SEM images of the fracture surfaces. The basis for the best possible evaluation is the understanding of the influence of the input data used for training deep neural networks (DNN). Therefore, this study discusses the influence of the selection of the input data used for the prediction quality of these networks in order to take this into account for future data acquisition. Specimens of various metallic materials were subjected to fatigue cracking experiment under laboratory conditions. The fractured surfaces were then imaged using various modes or detectors (such as SE, BSE and topography) in SEM, and those captured images were used to create a training data set. The relevance of the individual data for the quality of the prediction is determined by a specific combination of the different detector data. For the training, the well-established architecture of a UNet-ResNet34 with a fixed set of hyperparameters is used. It has been found in this present study that the combination of all input data significantly increases the prediction accuracy, whereby even the combination of SE and BSE data provides considerable advantages over the exclusive use of SE images. KW - Fractography KW - Machine Learning KW - Deep Learning KW - KI PY - 2023 U6 - https://doi.org/10.1016/j.engfailanal.2023.107814 VL - 156 SP - 1 EP - 8 PB - Elsevier AN - OPUS4-58918 LA - eng AD - Bundesanstalt fuer Materialforschung und -pruefung (BAM), Berlin, Germany ER -