Francisco Garcia-Sevilla

Characterizing the microstructures of steel subjected to heat treatments is crucial in the metallurgical industry for understanding and controlling their mechanical properties. In this study, we present a novel approach for generating images of steel microstructures that mimic those obtained with optical microscopy, using the deep learning technique of generative adversarial networks (GAN). The experiments were conducted using different hyperparameter configurations, evaluating the effect of these variations on the quality and fidelity of the generated images. The obtained results show that the images generated by artificial intelligence achieved a resolution of 512 × 512 pixels and closely resemble real microstructures observed through conventional microscopy techniques. A precise visual representation of the main microconstituents, such as pearlite and ferrite in annealed steels, was achieved. However, the performance of GANs in generating images of quenched steels with martensitic microstructures was less satisfactory, with the synthetic images not fully replicating the complex, needle-like features characteristic of martensite. This approach offers a promising tool for generating steel microstructure images, facilitating the visualization and analysis of metallurgical samples with high fidelity and efficiency.

The utilization of convolutional neural networks (CNNs) for semantic segmentation has proven to be successful in various applications, such as autonomous vehicle environment analysis, medical imaging, and satellite imagery. In this study, we investigate the application of different segmentation networks, including Deeplabv3+, U-Net, and SegNet, each recognized for their effectiveness in semantic segmentation tasks. Additionally, in the case of Deeplabv3+, we leverage the use of pre-trained ResNet50, ResNet18 and MobileNetv2 as feature extractors for a comprehensive analysis of steel microstructures. Our specific focus is on distinguishing perlite and ferrite phases in micrographs of low-carbon steel specimens subjected to annealing heat treatment. The micrographs obtained using an optical microscope are manually segmented. Preprocessing techniques are then applied to create a dataset for building a supervised learning model. In the results section, we discuss in detail the performance of the obtained models and the metrics used. The models achieve a remarkable 95% to 98% accuracy in correctly labeling pixels for each phase. This underscores the effectiveness of our approach in differentiating perlite and ferrite phases within steel microstructures.

This work aims to compare the effectiveness of different machine-learning techniques for the image classification of steel microstructures. For this, we use a set of samples of hypoeutectoid steels subjected to three heat treatments: annealing, quenching and quenching with tempering. Logically, the samples contain the typical constituents expected, and these are different for each treatment. Images are obtained by optical microscopy at 400× magnification and from different low-carbon steels to generate the data with some heterogeneity. Learning models are created with an image dataset for classification into three classes based on the respective heat treatments. Likewise, we develop two kinds of models by using, on the one hand, classical machine-learning methods based on the “bag of features” technique and, on the other hand, convolutional neural networks (CNN) with a transfer-learning approach by using GoogLeNet and ResNet50. We demonstrate the superiority of deep-learning techniques (CNN) over classical machine-learning methods.

Single point incremental forming (SPIF) is a cheap and flexible sheet metal forming process for rapid manufacturing of complex geometries. Additionally, it is important for engineers to measure the surface finish of work pieces to assess their quality and performance. In this paper, a predictive model based on machine learning and computer vision was developed to estimate arithmetic mean surface roughness (Ra) and maximum peak to valley height (Rz) of Ti6Al4V parts obtained by SPIF. An image database was prepared to train different classification algorithms in accordance with a supervised learning approach. A speeded up robust feature (SURF) detector was used to obtain visual vocabulary so that the classifiers are able to group the photographs into classes. The experimental results indicated that the proposed predictive method shows great potential to determine the surface quality, as classifiers based on a support vector machine with a polynomial kernel are suitable for this purpose.

Determining the intrinsic indices of sheet metals under compression states at high temperatures is vital to accurately predict the behavior of the material in warm/hot forming processes. Nevertheless, the literature contains little previous experimental data in this regard due to the difficulty of carrying out specific test methodologies in sheet metals. The authors of the present manuscript previously developed an approach to evaluate the in-plane compression behavior under a wide range of test conditions, which was applied here to characterize pure titanium and Ti6Al4V alloy until 750 °C. This procedure allowed us to quantify the asymmetric and anisotropic tension–compression (T-C) response of the materials involved and their evolution with temperature and strain rate. The asymmetry detected at room temperature showed a higher compression response in all cases, mostly reaching differences of around 10%. For the lowest strain rate studied, the typical assumed symmetric T-C behavior was observed from 300 and 450 °C onwards, for the rolling and transverse direction, respectively. In addition, stepped compression tests led us to deduce the anisotropy indices, which were different from those found under tension, in contrast to the r-values applied by most authors. Using the experimental results, a factor related to the asymmetry found was proposed to formulate an extended constitutive model. The asymmetry and anisotropy data supplied for compression under warm/hot conditions are the main novelty of this research.

An Experience on Learning Objects Reutilization based on Educational Resources Developed Miguel Latorre, Manuel Blazquez, Elio Sancristobal, Sergio Martín, Francisco Garcia-Sevilla, Gabriel Díaz, Catalina Martinez-Mediano and Manuel Castro Abstract High quality content is not easy to find. Search engines are not suited for this task because they only follow hyperlinks. This technique is often ineffective at finding e-learning resources. Learning objects are the answer to sharing, reusing and locating educational materials. Additional descriptions represented in form of metadata are required to achieve this goal, among other requirements. The information stored there is very important not only for institutions but for individual instructors or students as well. However, it takes time and experience for authors to create valuable metadata. Users -teachers and students- do not need to know learning specifications when they use standard compliant software (knowledge of the collection of TCP/IP protocols is not essential for a simple user), but it is essential for developers. Instructors cannot become expert metadata editors. Thus, a solution with the lesser intervention of the former is necessary. Here is explained a short automatic metadata generation process for different courses about Electronics engineering. The results are collections of circuit simulations and tutorials directly available from a learning repository. Development Electronic circuits almost have all the desired characteristics that learning objects need1. They are complete unto themselves to allow a student to easily apply them. Diagrams can be rather complex, or assessments of minimum size appropriate for their use. Objects may be as short as a single tutorial or a complete lesson. Language is not a barrier because schematics use a symbolic representation. Furthermore, cost is nearly zero with the help of free simulators or demos. Nevertheless, teachers and students need detailed descriptions of their contents to quickly locate them in a network. The Learning Object Metadata (LOM) Standard represents an important step towards fostering the construction of a new generation of artificial intelligence-based Web Learning systems2. Learning object metadata cannot be distributed without using a common format. IEEE LOM3 is the basis of most metadata standards. This metadata element set is encoded in a binding, such as XML or RDF for computer processing4. Institutions select some elements or extend the default set to adapt the standard for their local education systems. Some of these profiles are CanCORE or LOM-es. Once all previous conditions are taken in mind, we can start to create our own metadata. Till recent years, digital media production was not relevant on education. Most authoring software did not provide any information about user-created content. Technical requirements,