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

Abstract

The drive for evermore efficient construction, design and layout of components, buildings, and vehicles also affects materials science to a large extent. For some time now, research has focused on the adjustment and tailoring of microstructure in the emerging research field of process-(micro)structureproperty-performance relations. Increasingly, machine learning methods are also being used here, for example, to accelerate material design, [1] to link grain sizes and mechanical properties, [2] or to describe dislocation networks. [3] The specific microstructure has a big influence on basic parameters such as the toughness or the yield stress but even more on local properties on the microscale like the damage initiation and accumulation [4] or the fatigue properties. [5] For well-known steels such as dual-phase (DP) steels, the relations between microstructure and damage properties have already been studied in a vast number of publications, see, for example, Pütz et al. [6] for investigations about the damage tolerance of DP800 and DP1000, Tasan et al. [7] regarding simulative experimental studies about stress and strain partioning in DP800, or Heibel et al. [8] for studies on the edge crack sensitivity of different DP and complex phase steels. Experimentally, however, it is almost impossible to quantify the specific effects of individual microstructure parameters (e.g., grain size or phase ratio), since the modification of a particular parameter normally also modifies some or all of the other parameters. [9] In order to overcome this problem, simulation-based approaches using virtual representations of the microstructure are suitable. A commonly used tool for this kind of analysis is the so called representative volume element (RVE). These are sections of the volume, whose specific properties are of a nature that is representative of the behavior of the aggregate material. For this, the following scale separation must apply: L micro ( L rve ( L macro . [10] Since the microscopic characteristic length depends strongly on the material, the required size of the RVE also depends on the properties of the material. [11] RVE are distinguished from real structure models of microstructure. These kinds of models can be generated directly from microstructure images obtained, for example, from light optical microscopy (LOM) or electron backscatter diffraction (EBSD). They are for instance used by Tasan et al. [12] . In this case, the LOM/EBSD image is directly used as the 2D model to simulate the underlying mechanical properties of the experimentally observed microstructure. In contrast to that, the different parameters of the microstructure (grain size, orientations, phase ratios) can be represented by some kinds of distribution functions

Prediction of the Fatigue Strength of Steel Based on Interpretable Machine Learning

“…However, the variations and combinations of these factors result in a vast search space, making it impractical to study them individually through manual analysis. Machine learning, on the other hand, explores the hidden relationships between features and target variables, establishes quantitative models between them, and identifies the influence of features on the target variables, thereby accelerating the efficiency of materials' research and development [11][12][13][14]. In this study, a prediction model for FS is established considering the factors of composition, processing, and microstructure.…”

Research on lemon quality grading detection based on improved YOLOv5s

“…Multi-Task Learning (MTL) refers to the use of a single network model to simultaneously perform two or more detection tasks [5] , sharing parameters and features between different tasks can reduce computational complexity [6] . Traditional object detection algorithms can only perform defect detection at the box level, marking the position and type of defects [7] . In this experiment, in addition to detecting the types of lemon defects, it is also necessary to judge the quality level of the lemon at the image level.…”