The strain-rate sensitivity exponent and the activation volume provide a fingerprint of the rate controlling mechanisms during thermally activated deformation. Over many years, nanoindentation has been used for assessing local mechanical properties such as hardness and modulus. To shed light into temperature and rate dependent deformation behavior, indentation creep tests were suggested,1 and Mayo and Nix2 reported superplastic deformation behavior of Pb and Sn during indentation. Especially for ultrafine-grained (UFG) and nanocrystalline (NC) materials, nanoindentation became a frequently used technique to determine strain-rate sensitivity (SRS). Schwaiger et al.3 published in 2003 a systematic comparison of nanoindentation and tensile testing on NC-Ni, where both techniques found a strongly grain size dependent SRS. Within the last decade, many studies on different materials classes, microstructures and environmental influences were reported in literature. Due to the limited amount of required material, especially the severe plastic deformation (SPD)4 and thin film5,6 communities used nanoindentation to gain information on the microstructure dependent deformation behavior of many face-centered cubic (FCC),7,8,9,10,11,12,13,14,15,16 body-centered cubic (BCC),17,18,19,20,21 and some hexagonal closed packed (HCP)22,23,24,25 metals. Also the SRS of nanocomposites such as Cu-Nb,26 nanoporous Cu and Au,27,28 or nowadays high-entropy alloys (HEA)29,30 were successfully investigated by nanoindentation. Besides the successful determination of positive SRS, few studies report experiments and results of negative SRS.31,32 Moreover, many other investigations dealing with various materials systems, such as glasses32,33,34 or intermetallic phases,35 can also be found.
FULL Advance Steel 2006 Activation
In the following, two independent advanced nanoindentation methodologies for a reliable characterization of a strain-rate sensitive deformation behavior will be discussed on the example of a highly deformed austenitic steel (A220). Furthermore, several examples will demonstrate the significance of microstructure, crystal structure, and solute content on the time dependent deformation behavior and the underlying thermally activated deformation processes.
To overcome these issues, Maier et al.57 and Alkorta et al.19 came up with an advanced indentation protocol, where abrupt strain-rate changes are applied during one single indentation segment.57 These kind of test are state-of-the-art in many macroscopic compression or tension tests, and they were successfully adjusted to the small-scale testing world. Within nanoindentation strain-rate jump tests, the applied strain-rate/indentation depth protocol can be adjusted individually, allowing variable indentation depths for the abrupt changes and different indentation strain-rates. Every parameter can be chosen to the requirement of the application and material.32 Additionally, several drawbacks of the cSR-method can be overcome with this one indentation method. For example, the total indentation time is significantly reduced because the initial indentation depth is performed at large strain-rates and segments with lower indentation strain-rates are performed at larger depths, effectively reducing testing times. Moreover, strain-rate jump tests also allow probing the mechanical properties at one single location, thus, e.g., the determination of SRS of individual phases or microstructural heterogeneities.57 For stable microstructural conditions, the evaluated strain-rate sensitivity is independent of the individual applied sequences of strain-rates and indentations depths. Nevertheless, in the case of unstable microstructures or during varying plastic strain conditions, the experimental sequence might slightly influence the evaluated materials properties. Therefore, the SRS should be always discussed as a function of hardness, since the rate sensitivity, the microstrcutural length scales and the hardness are closely related to each other.
This motivation leads us to use Deep Learning methods which are recently grabbing the attention of scientists due to their strong ability to learn high-level features from raw input data. Recently, these methods have been applied very successfully to computer vision problems8,9. They are based on artificial neural networks such as Convolutional Neural Networks (CNNs)9. They can be trained for recognition and semantic pixel-wise segmentation tasks. Unlike traditional methods in which feature extraction and classification are learnt separately, in Deep Learning methods, these parts are learnt jointly. The trained models have shown successful mappings from raw unprocessed input to semantic meaningful output. As an example, Masci et al.10 used CNNs to find defects in steel. In this work, we show that Deep Learning can be successfully applied to identify microstructural patterns. Our method uses a segmentation-based approach based on Fully Convolutional Neural Networks (FCNNs) which is an extension of CNNs accompanied by a max-voting scheme to classify microstructures. Our experimental results show that the proposed method considerably increases the classification accuracy compared to state of the art. It also shows the effectiveness of pixel-based approaches compared to object-based ones in microstructural classification.
Based on the instrument used for imaging, we can categorize the related works into Light Optical Microscopy (LOM) and Scanning Electron Microscopy (SEM) imaging. High-resolution SEM imaging is very expensive compared with LOM imaging in terms of time and operating costs. However, low-resolution LOM imaging makes distinguishing microstructures based on their substructures even more difficult. Nowadays, the task of microstructural classification is performed by observing a sample image by an expert and assigning one of the microstructure classes to it. As experts are different in their level of expertise, one can assume that sometimes there are different opinions from different experts. However, thanks to highly professional human experts, this task has been accomplished so far with low error which is appreciated. Regarding automatic microstructural classification, microstructures are typically defined by the means of standard procedures in metallography. Vander Voort11 used Light Optical Microscopy (LOM) microscopy, but without any sort of learning the microstructural features which is actually still the state of the art in material science for classification of microstructres in most institutes as well as in industry. His method defined only procedures with which one expert can decide on the class of the microstructure. Moreover, additional chemical etching12 made it possible to distinguish second phases using different contrasts, however etching is constrained to empirical methods and can not be used in distinguishing various phases in steel with more than two phases. Nowadays, different techniques and approaches made morphological or crystallographic properties accessible4,13,14,15,16,17,18. Any approach for identification of phases in multiphase steel relies on these methods and aims at the development of advanced metallographic methods for morphological analysis purposes using the common characterization techniques and were accompanied with pixel- and context-based image analysis steps.
Activation Function usually follows a pooling or fully connected layer and introduces a nonlinear activation operation like a sigmoid or rectified linear unit (ReLU). The ReLU function \(relu(x)=\,\max \,\mathrm(0,x)\) are most common as the gradient is piece-wise constant and does not vanish for high activations (in contrast to e.g., sigmoid).
Transgenic animals represent another technology for obtaining fully human mAbs (Fig. 3c). This technology was introduced in 1994 by the publication of two transgenic mouse lines, the HuMabMouse [35] and the XenoMouse [36]. The lines were genetically modified such that human immunoglobulin (Ig) genes were inserted into the genome, replacing the endogenous Ig genes and making these animals capable of synthesizing fully human antibodies upon immunization [35, 37]. The first human antibody generated in a transgenic mouse to anti-epidermal growth factor receptor (EGFR), panitumumab, was approved by the US FDA in 2006 (Fig. 1) [38, 39]. The number of fully human antibodies made from transgenic mice has increased rapidly, with the number of approved drugs currently at 19 (Table 5). Depending on the immunization protocol, high-affinity human antibodies can be obtained through further selection of hybridoma clones generated from immunized transgenic mice. Using a theoretically similar approach, the generation of neutralizing human antibodies from human B cells has also yielded promising results for infectious disease therapeutics.
The recent development of bispecific antibodies offers attractive new opportunities for the design of novel protein therapeutics. A bispecific antibody can be generated by utilizing protein engineering techniques to link two antigen binding domains (such as Fabs or scFvs), allowing a single antibody to simultaneously bind different antigens. Thus, bispecific antibodies may be engineered to exhibit novel functions, which do not exist in mixtures of the two parental antibodies. Most bispecific antibodies are designed to recruit cytotoxic effector cells of the immune system to target pathogenic cells [40]. The first approved bispecific antibody was catumaxomab in Europe in 2009 [41]. Catumaxomab targets CD3 and EpCAM to treat solid tumors in patients with malignant ascites. However, this drug was withdrawn from the market in 2017 for commercial reasons. Currently, two bispecific antibodies have obtained US FDA approval and are on the market. First, blinatumomab is a bispecific T-cell engager (BiTE) that targets CD3 and CD19 for treatment of B-cell precursor acute lymphoblastic leukemia (ALL) [42]. Second, emicizumab is a full-size bispecific IgG with natural architecture, which binds to activated coagulation factors IX and X for the treatment of haemophilia A [43]. To date, there are more than 85 bispecific antibodies in clinical trials, about 86% of which are under evaluation as cancer therapies [40]. The concepts and platforms driving the development of bispecific antibodies continue to advance rapidly, creating many new opportunities to make major therapeutic breakthroughs. 2ff7e9595c
Comments