Michael Engstler

Advanced Steel Microstructure Classification by Deep Learning Methods

The inner structure of a material is called microstructure. It stores the genesis of a material and determines all its physical and chemical properties. While microstructural characterization is widely spread and well known, the microstructural classification is mostly done manually by human experts, which gives rise to uncertainties due to subjectivity. Since the microstructure could be a combination of different phases or constituents with complex substructures its automatic classification is very challenging and only a few prior studies exist. Prior works focused on designed and engineered features by experts and classified microstructures separately from the feature extraction step. Recently, Deep Learning methods have shown strong performance in vision applications by learning the features from data together with the classification step. In this work, we propose a Deep Learning method for microstructural classification in the examples of certain microstructural constituents of low carbon steel. This novel method employs pixel-wise segmentation via Fully Convolutional Neural Network (FCNN) accompanied by a max-voting scheme. Our system achieves 93.94% classification accuracy, drastically outperforming the state-of-the-art method of 48.89% accuracy. Beyond the strong performance of our method, this line of research offers a more robust and first of all objective way for the difficult task of steel quality appreciation.

Homogeneity Quantification Method and its Application to Microstructure Assessment

Abstract In this work, a method to precisely define and quantify homogeneity of microstructures is proposed. Said method is based on an interpretation of the Gini index and it was developed specifically for microstructure homogeneity assessments. Object homogeneity and region homogeneity are examples of parameters developed to describe different aspects of homogeneity in micrographs. Using these parameters, the homogeneities of Al-Si casting alloys and cast iron have been evaluated. The method can be adapted and combined with other techniques to fit the requirements of a given study case. Further applications of the method are discussed.

Feature Adaptive Sampling for Scanning Electron Microscopy.

A new method for the image acquisition in scanning electron microscopy (SEM) was introduced. The method used adaptively increased pixel-dwell times to improve the signal-to-noise ratio (SNR) in areas of high detail. In areas of low detail, the electron dose was reduced on a per pixel basis, and a-posteriori image processing techniques were applied to remove the resulting noise. The technique was realized by scanning the sample twice. The first, quick scan used small pixel-dwell times to generate a first, noisy image using a low electron dose. This image was analyzed automatically, and a software algorithm generated a sparse pattern of regions of the image that require additional sampling. A second scan generated a sparse image of only these regions, but using a highly increased electron dose. By applying a selective low-pass filter and combining both datasets, a single image was generated. The resulting image exhibited a factor of ≈3 better SNR than an image acquired with uniform sampling on a Cartesian grid and the same total acquisition time. This result implies that the required electron dose (or acquisition time) for the adaptive scanning method is a factor of ten lower than for uniform scanning.

3D Microstructural Study of AlSi8Mg5 Alloy by FIB-Tomography

Abstract In the present work, a foundry AlSi8Mg5 alloy is investigated by FIB-Tomography in the as-cast condition and after semi-solid heat treatment. The different microstrutural features are characterized quantitatively using morphological shape factors. The 3D-shape of Mg2Si Chinese script as well as Si and impurity Fe containing phases are revealed in high detail. A comparative study with respect to 2D structures is presented as well.

Electron beam characterization techniques for the study of wear in sliding contacts

In this study different methods for studying phenomena of adhesive wear in sliding electrical contacts using electron beam characterization methods within a dual beam workstation are presented. The interface between the substrate and transferred material is complex exhibiting small grain size as well as mechanical deformation. For this purpose, a dual beam workstation is used in order to perform high resolution imaging, structural investigation by electron backscatter diffraction as well as FIB cross sectioning. Only a combination of different imaging techniques with different contrast mechanisms can provide a full understanding of the wear mechanism. During wear a prow is formed on the slider. The structure implies a multi-generation history. The deformation texture resembles a simple shear texture, which is in agreement with the presumed deformation mode. The study mainly gives a guideline for future work in the field of wear in sliding contacts.

Why We Need All Dimensions to Solve Both Very Old and Very New Questions in Materials at the Micro-, Nano- and Atomic Scales

In der technologischen Werkstoffentwicklung und der routinemäßigen Qualitätssicherung des Werkstoffgefüges werden auch heute noch alle wesentlichen Entscheidungen zum räumlichen Gefüge am zweidimensionalen mikroskopischen Abbild, also dem metallographischen bzw. materialographischen Anschliff getroffen, so wie dies seit langem bewährte Tradition ist. Waren bislang oft noch Richtreihen und subjektive Expertenmeinungen der häufigste und ausreichende Maßstab, so erzwingen die für anspruchsvolle Anwendungsfälle maßgeschneiderten Hochleistungswerkstoffe immer engeren ToleranzReceived: July 06, 2015 Accepted: July 06, 2015 Translation: Phil Tate

Serial Sectioning Techniques – a Versatile Method for Three-Dimensional Microstructural Imaging: The working group “Serial Sectioning Tomography” is headed by Prof. Dr.-Ing. Frank Mücklich.

Der Begriff Tomographie (altgriechisch: tome = Schnitt, graphein = schreiben) fasst bildgebende Verfahren zusammen, welche die räumliche Struktur eines Objektes ermitteln. Um dies zu erreichen, wird in den Serienschnitt-Verfahren die Probenoberfläche kontrolliert abgetragen und in idealerweise konstanten Abständen werden mikroskopische Aufnahmen der Probe angefertigt. Den einzelnen Pixeln der mikroskopischen 2D-Aufnahmen wird dabei die Dickeninformation des Abtrags zugeordnet und es entsteht der dreidimensionale Voxel (volumetric pixel). Die Ausdehnung des Voxels entspricht in xund Received: May 24, 2018 Accepted: May 30, 2018 Translation: E. Engert

From Correlative Microscopy to 3D Understanding of Material Microstructures

The investigation of the origin and formation of microstructures and the effect that microstructure has on the properties of materials are questions of ever-increasing importance in the strongly growing variety of tailor-made and high performing structural as well as functional materials. The individual ordered regions, the phases and defects, with all their chemical and structural variety, form the complex set of components of the microstructure of a material. The properties of a material are determined not only by the kinds and fractions of components present, but also by details of their geometry; and geometrical analysis can very often prove the key to quantitative understanding the formation of microstructure as well as resulting material properties [1].

“Smart Microscopy”: Feature Based Adaptive Sampling for Focused Ion Beam Scanning Electron Microscopy

A new method for the image acquisition in scanning electron microscopy (SEM) was introduced. The method used adaptively increased pixel-dwell times to improve the signal-to-noise ratio (SNR) in areas of high detail. In areas of low detail, the electron dose was reduced on a per pixel basis, and a-posteriori image processing techniques were applied to remove the resulting noise. The technique was realized by scanning the sample twice. The first, quick scan used small pixel-dwell times to generate a first, noisy image using a low electron dose. This image was analyzed automatically, and a software algorithm generated a sparse pattern of regions of the image that require additional sampling. A second scan generated a sparse image of only these regions, but using a highly increased electron dose. By applying a selective low-pass filter and combining both datasets, a single image was generated. The resulting image exhibited a factor of ≈3 better SNR than an image acquired with uniform sampling on a Cartesian grid and the same total acquisition time. This result implies that the required electron dose (or acquisition time) for the adaptive scanning method is a factor of ten lower than for uniform scanning.

3D Microstructure Characterization and Analysis of Al-Si Foundry Alloys at Different Length Scales

The term microstructure refers to the complete internal structure of a material on the micro, nano and atomic scales. On one hand, it records the entire history of a material’s processing and structuring (casting, forming, heat treatments, but also crystal growth, etc.) through its phase composition, defect structure and morphology. On the other hand, all properties (stiffness, formability, conductivity, all structural and functional properties) are determined by the microstructure. Thus, the microstructure can be seen as an intrinsic multi-scale memory from which we can read at each relevant scale the precise information about all microstructure-building processes as well as predict the final material properties. However, it could not be fully exploited so far due to the lack of adequate 3D characterization techniques. Today new emerging tomographic possibilities provide for the very first time all complementary 3D information in micro, nano and atomic scales.