Articles — University

Dieses Dokument enthält Notizen aus dem Kurs Mathematik 2 des Studienbefähigungslehrgangs 2014/15 der FH Joanneum Graz.

This master thesis aims to investigate and to assess the fatigue life of a component that undergoes cyclic thermal loads. This study will consider different approaches through the use of different material models, providing a final comparison of the assessed fatigue life. The case study is a cylindrical copper mold used for continuous casting of steel; in order to work at high temperature the mold is properly cooled, but nevertheless high stresses arise due to the huge thermal flux and its uneven distribution. The beginning of this work is focused on understanding theories and models available in literature in order to implement data in a finite element software. A comparison of two software (Ansys and Marc Mentat) is carried out, with the purpose of verifying experimental data coming from another work [3]. Subsequently we use Marc Mentat to define a coupled thermal-mechanical analysis, and we perform it using a combined model. A huge computational time is required to get the solution, so other models were exploited in order to save time, and the variation of the obtained solution was evaluated. Several simulations are performed for reaching stabilized conditions, from where strain data will be used to define fatigue life through the number of cycles to failure. The subsequent chapter takes care of this last concept, giving a brief explanation of fatigue and using data coming from previous simulations for assessing fatigue life. A final comparison is provided with the purpose of finding a trade off between the solution’s reliability and the time employed. In the appendix the macro used for the computed simulations are attached.

Considering the growing impact of ideas nowadays, via quaternary sector, the plagiarism detection has become constant in texts, songs, as well as source codes. This work proposes the creation of the tool \nameOfProgram \ for plagiarism detection in simple texts with GNU GPL license. \nameOfProgram \ was designed to allow your extension for plagiarism detection in source codes. The tool was tested and results are presented in this paper.

En este documento se muestra un acercamiento identificación de especies arbóreas mediante Histogramas De Gradientes Orientados y maquinas de soporte Vectorial.

There are several voice communication systems that are used nowadays which are capable of maintain voice calls between two users in real time. Telephones are widely used all around the world in an unlimited kind of situations. All of these situations expose the microphone (or microphones) of the phones to different and unpredictable noises, as street noise, sea noise, rain noise, wind noise, unwanted voices, car motors, etc. As microphones capture all the sounds around it, including the wanted voice and the unwanted noises, it is necessary to implement digital real time filters capable of attenuate as much as possible all the surrounding noises. It exists a large quantity of noise reduction methods that have been used in the calling algorithms of phones. Even if these methods have had, in general, a good performance, there is still a research being done in this area in order to improve the current results. Because of this, the multichannel methods were created (using multiple microphones) as well as new algorithms that pretend to have a better noise reduction than the single channel methods. Most of these methods require a speech presence probability (SPP) method to achieve the noise reduction. The following document presents a research about different SPP methods as well as a comparison between these. This includes an explanation on how theses algorithm work, a Matlab implementation using real voice and noise recordings and objective tests of the filter.

Trabalho de Grupo Cefet

In this work we apply machine learning as a conducive method towards identifying previously unstudied patterns in chromosome interaction data sets. We rst use supervised learning to show that patterns identi ed by a user can be learned by tensor ow models, and then transition into unsupervised methods to delve even more deeply into the possibilities of discovery without human intervention.

Gearbox Progress Report

Este trabalho tem a missão de explicar alguns temas de sistemas distribuídos, estes sistemas são fundamentais para a comunicação, integração e processamento de dados via internet ou rede local. Ao longo do trabalho abordaremos os principais tópicos tais como: WebServices, Padrões de Sockets entre outros.