Identification of Martensite Bands in Dual-Phase Steels: A Deep Learning Object Detection Approach Using Faster Region-Based-Convolutional Neural Network

被引:6
作者
Fehlemann, Niklas [1 ]
Aguilera, Ana Lia Suarez [1 ]
Sandfeld, Stefan [2 ]
Bexter, Felix [1 ]
Neite, Maximilian [1 ]
Lenz, David [1 ]
Koenemann, Markus [1 ]
Muenstermann, Sebastian [1 ]
机构
[1] Rhein Westfal TH Aachen, Integr Mat & Struct, Intzestr 1, D-52072 Aachen, Germany
[2] FZ Julich, Inst Adv Simulat Mat Data Sci & Informat IAS 9, Wilhelm Johnen Str, D-52428 Julich, Germany
关键词
dual-phase steels; Faster region-based-convolutional neural networks; machine learning; martensite banding; representative volume elements; transfer learning; REPRESENTATIVE VOLUME; STRAIN; CLASSIFICATION; GENERATION; EVOLUTION; STRESS; DAMAGE;
D O I
10.1002/srin.202200836
中图分类号
TF [冶金工业];
学科分类号
0806 ;
摘要
Martensite banding in dual-phase steels is an important research topic in the field of materials design, since it affects the local damage properties of the material largely. Therefore, it is necessary to quantify the amount and the geometrical details of the bands in a specific microstructure, for example, for simulative approaches. A convolutional neural network is trained on manually labeled scanning electron microscopy images of DP800 steel and a subsequent effort is made to transfer these results to statistical quantities for the generation of representative volume elements (RVE). As exact geometric definitions of martensite bands in 2D are difficult, the influence of different band definitions is investigated. The result of the training shows good prediction accuracy but is strongly dependent on the chosen band definition and the underlying human bias from the labeling process. A statistical analysis using cross-validation shows that reliable results can already be achieved with only small datasets of around 50-100 training images due to the transfer learning approach. This is an important outcome as it eliminates the need to generate a large dataset which can only be obtained from time-consuming microscopy work and manual labeling of the images.
引用
收藏
页数:13
相关论文
共 45 条
[1]   Advanced Steel Microstructural Classification by Deep Learning Methods [J].
Azimi, Seyed Majid ;
Britz, Dominik ;
Engstler, Michael ;
Fritz, Mario ;
Muecklich, Frank .
SCIENTIFIC REPORTS, 2018, 8
[2]   Generation of 3D representative volume elements for heterogeneous materials: A review [J].
Bargmann, Swantje ;
Klusemann, Benjamin ;
Markmann, Juergen ;
Schnabel, Jan Eike ;
Schneider, Konrad ;
Soyarslan, Celal ;
Wilmers, Jana .
PROGRESS IN MATERIALS SCIENCE, 2018, 96 :322-384
[3]   Albumentations: Fast and Flexible Image Augmentations [J].
Buslaev, Alexander ;
Iglovikov, Vladimir I. ;
Khvedchenya, Eugene ;
Parinov, Alex ;
Druzhinin, Mikhail ;
Kalinin, Alexandr A. .
INFORMATION, 2020, 11 (02)
[4]  
Deep Learning, 2017, ADAPTIVE COMPUTATION
[5]  
E04 Committee, PRACT ASS DEGR BAND
[6]   Influence of synthetically generated inclusions on the stress accumulation and concentration in X65 pipeline steel [J].
Fehlemann, N. ;
Sparrer, Y. ;
Puetz, F. ;
Koenemann, M. ;
Muenstermann, S. .
INTERNATIONAL DEEP-DRAWING RESEARCH GROUP CONFERENCE (IDDRG 2021), 2021, 1157
[7]   Numerical study of inclusion parameters and their influence on fatigue lifetime [J].
Gillner, Karl ;
Henrich, Manuel ;
Muenstermann, Sebastian .
INTERNATIONAL JOURNAL OF FATIGUE, 2018, 111 :70-80
[8]  
Girshick R., 2013, Rich feature hierarchies for accurate object detection and semantic segmentation
[9]  
Girshick R, 2015, Arxiv, DOI arXiv:1504.08083
[10]   Representative volume: Existence and size determination [J].
Gitman, I. M. ;
Askes, H. ;
Sluys, L. J. .
ENGINEERING FRACTURE MECHANICS, 2007, 74 (16) :2518-2534