OPUS 4 | Universitätspublikationen

Zellulare Nichtlineare Netzwerke - Optimierungsverfahren und Anwendungen (2012)

Gunter Geis

Zellulare Nichtlineare Netzwerke bzw. Zellulare Neuronale Netzwerke, sogenannte CNN, wurden 1988 von L.O. Chua und L.Yang eingeführt und seither intensiv untersucht. Diese sind als Simulations-Software und als schaltungstechnische Realisierungen, in Hardware, verfügbar. Als analog arbeitende Hardware Schaltungen können diese Netzwerke erhebliche Rechenleistungen erzielen. Durch ihren Aufbau ermöglichen sie eine parallele Daten- und Signalverarbeitung. Eine Einführung in CNN wird gegeben und das EyeRIS 1.1 Systems des Unternehmens ANAFOCUS Ltd. vorgestellt. Das EyeRIS 1.1 System ist mit einem analog arbeitenden Focal Plane Prozessor (FPP) und einem digitalen Prozessor ausgestattet, wobei der Focal Plane Prozessor auch als Kamera zur Aufnahme von Bildern und Bildsequenzen benutzt werden kann. Dies ermöglicht es, analoge CNN-Algorithmen zusammen mit digitalen Algorithmen auf einem System zu implementieren und so die Vorteile beider Ansätze zu nutzen. Der Datenaustausch zwischen dem analogen und digitalem Teil des EyeRIS 1.1 Systems geschieht mittels digital/analog und analog/digital Wandlung. Es werden Algorithmen auf dem EyeRIS 1.1 System untersucht und mit Ergebnissen die mittels Simulationen erzeugt wurden verglichen. In Voruntersuchungen werden die Darstellungsgenauigkeit von Werten im analogen Teil des EyeRIS 1.1 Systems und die Verarbeitungsgeschwindigkeiten des EyeRIS 1.1 Systems untersucht. Im Weiteren wird besonderes Augenmerk auf medizinische und technische Anwendungsgebiete gelegt werden. Im medizinischen Anwendungsbereich wird die Implementierung von Algorithmen zur Vorhersage epileptischer Anfälle untersucht. Hierfür wird ein evolutionär motiviertes Optimierungsverfahren entwicklet und auf dem EyeRIS 1.1-System implementiert. Hierbei werden Simulationen durchgeführt und mit Ergebnissen, die mittels Verwendung des EyeRIS 1.1 Systems erlangt wurden, verglichen. Ein zweites Verfahren geht die Signalanalyse für die Vorhersage auf dem EyeRIS 1.1-System mittels Mustererkennung an. Das Mustererkennungsverfahren wird eingehend beschrieben sowie die hierbei zu beachtenden Randbedingungen erläutert. Die Ergebnisse zeigen, daß Algorithmen zur Vorhersage von epileptischen Anfällen auf schaltungstechnichen Realisierungen von CNN implementiert werden können. Im technischen Bereich wird die Anwendbarkeit auf die Problemstellung der Bildverarbeitung gelegt und die Möglichkeit von CNN basierten Algorithmen zur Erkennung von Prozessparametern bei Laserschweißverfahren untersucht. Ein solcher Prozessparameter ist das sogenannte Key-Hole, welches in Bildsequenzen von Laserschweißprozessen als ein Maß für die zu erwartende Qualität einer Schweißnaht herangezogen werden kann. Ein CNN basierter Algorithmus für die Erkennung solcher Key-Holes wird in dieser Arbeit vorgestellt und untersucht. Für die Überwachung eines Laserschweißverfahrens wird der entwickelte Algorithmius und seine Funktionsweise beschrieben. Dieser wird in Teilalgorithmen auf die analog bzw. digital arbeitenden Komponenten des EyeRIS 1.1 Systems verteilt. Die Teilalgorithmen und die möglichen Aufteilungen und deren Laufzeitverhalten werden beschrieben und untersucht. Die Ergebnisse der Untersuchung zeigen, daß eine Prozessüberwachung mittels CNN möglich ist und heben die Vorteile hervor, welche die Bildaufnahme und -verarbeitung mittels analoger CNN-Hardware bietet. Eine Untersuchung des Laufzeitverhaltens auf Grafikkarten Prozessoren (GPU's) wird im Anhang vorgestellt.

Gas system, gas quality monitor and detector control of the ALICE Transition Radiation Detector and studies for a pre-trigger data read-out system (2012)

Nora Pitz

The main purpose of the Transition Radiation Detector (TRD) located in the central barrel of ALICE (A Large Ion Collider Experiment) is electron identification for separation from pions at momenta pt > 1 GeV/c, since in this momentum range the measurements of the specific energy loss (dE/dx) of the Time Projection Chamber (TPC) is no longer sufficient. Furthermore, it provides a fast trigger for high transverse momentum charged particles (pt > 3 GeV/c) and makes a significant contribution to the optimization of the tracking of reaction products in heavy-ion collisions. Its whole setup comprises 18 supermodules out of which 13 are presently operational and mounted cylindrically around the beam axis of the Large Hadron Collider (LHC). A supermodule contains either 30 or 24 chambers, each consisting of a radiator for transition radiation creation, a drift and an amplifying region followed by the read-out electronics. In total, the TRD is an array of 522 chambers operated with about 28 m3 of a Xe-CO2 [85-15%] gas mixture. During the work of this thesis, the testing, commissioning, operation and maintenance of detector parts, the gas system and its online quality monitor, improvements on the detector control user-interface and studies about a new pre-trigger module for data read-out have been accomplished. The TRD gas system mixes, distributes and circulates the operational gas mixture through the detector. Its overall optimization has been achieved by minimizing gas leakage, surveying, controlling, maintaining and continuously improving it as well as designing and carrying out upgrades. Gas quality monitors of the type \GOOFIE" (Gas prOportional cOunter For drIfting Electrons) can be used in gaseous detectors as on-line monitors of the electron drift velocity, gain and gas properties. One of these devices has been implemented within the TRD gas system, while another one surveys the gas of the TPC. Both devices had to be adapted to the specific needs of the detectors, were under constant surveillance and control, and needed to be further developed on both hardware and software side. To improve the operation of the TRD, modifications on its DCS software (Detector Control System) used for monitoring, controlling, operating, regulating and configuring of hardware and computing devices have been carried out. The DCS is designed to enable an operator to interact with equipment through user interfaces that display the information from the system. The main focus of this work was laid on the optimization of the usability and design of the user interface. The front-end electronics of the TRD require an early start signal (\pre-trigger") from the fast forward detectors or the Time-Of-Flight detector during the running periods. The realization of a new hardware concept for the read-out of the TRD pre-trigger system has been studied and first tests were performed. This new module called PIMDDL (Pre-trigger Interface Module Detector Data Link) is meant to acquire all data necessary to simulate and predict the full pre-trigger functionality, and to verify its proper operation. Furthermore, it shall provide all functionalities of the so-called Control Box Bottom as well as keep the functionalities of the already existing PIM (Pre-trigger Interface Module) in order to combine and replace these two modules in the future.

Directed deposition of silicon nanowires using neopentasilane as precursor and gold as catalyst (2012)

Britta Kämpken Verena Wulf Norbert Auner Marcel Winhold Michael Huth Daniel Rhinow Andreas Terfort

In this work the applicability of neopentasilane (Si(SiH3)4) as a precursor for the formation of silicon nanowires by using gold nanoparticles as a catalyst has been explored. The growth proceeds via the formation of liquid gold/silicon alloy droplets, which excrete the silicon nanowires upon continued decomposition of the precursor. This mechanism determines the diameter of the Si nanowires. Different sources for the gold nanoparticles have been tested: the spontaneous dewetting of gold films, thermally annealed gold films, deposition of preformed gold nanoparticles, and the use of “liquid bright gold”, a material historically used for the gilding of porcelain and glass. The latter does not only form gold nanoparticles when deposited as a thin film and thermally annealed, but can also be patterned by using UV irradiation, providing access to laterally structured layers of silicon nanowires.

A demonstrator for the Micro-Vertex-Detector of the CBM experiment (2010)

Christoph Schrader Samir Amar-Youcef Norbert Bialas Michael Deveaux Ingo Fröhlich Christian Müntz Selim Seddiki Tobias Tischler Joachim Stroth

CMOS sensors are the most promising candidates for the Micro-Vertex-Detector (MVD) of the CBM experiment at GSI, as they provide an unprecedented compromise between spatial resolution, low material budget, adequate radiation tolerance and readout speed. To study the integration of these sensors into a detector module, a so-called MVD-demonstrator has been developed. The demonstrator and its in-beam performance will be presented and discussed in this work.

A versatile method for simulating pp -> ppe+e- and dp -> pne+e-p_spec reactions (2010)

Frank Dohrmann Ingo Fröhlich Tetyana Galatyuk Romain Holzmann Philipp K. Kählitz Burkhard Kämpfer Emilie Moriniere Yvonne C. Pachmayer Beatrice Ramstein Piotr Salabura Joachim Stroth Radoslaw Trebacz Jacques Van de Wiele

We have developed a versatile software package for the simulation of di-electron production in pp and dp collisions at moderate beam kinetic energies (1-2GeV). Particular attention has been paid to incorporate different descriptions of the Dalitz decay Δ rightarrow Ne + e - via a common interface. In addition, suitable parameterizations for the virtual bremsstrahlung process NN rightarrow NNe + e - based on one-boson exchange models have been implemented. Such simulation tools with high flexibility of the framework are important for the interpretation of the di-electron data taken with the HADES spectrometer and demonstrates the wide applicability within the field of nuclear and hadronic physics.

Human myocardial protein pattern reveals cardiac diseases (2012)

Jonas Bergquist Gökhan Baykut Maria Bergquist Matthias Witt Franz-Josef Mayer Doan Baykut

Proteomic profiles of myocardial tissue in two different etiologies of heart failure were investigated using high performance liquid chromatography (HPLC)/Fourier transform ion cyclotron resonance mass spectrometry (FT-ICR MS). Right atrial appendages from 10 patients with hemodynamically significant isolated aortic valve disease and from 10 patients with isolated symptomatic coronary heart disease were collected during elective cardiac surgery. As presented in an earlier study by our group (Baykut et al., 2006), both disease forms showed clearly different pattern distribution characteristics. Interesting enough, the classification patterns could be used for correctly sorting unknown test samples in their correct categories. However, in order to fully exploit and also validate these findings there is a definite need for unambiguous identification of the differences between different etiologies at molecular level. In this study, samples representative for the aortic valve disease and coronary heart disease were prepared, tryptically digested, and analyzed using an FT-ICR MS that allowed collision-induced dissociation (CID) of selected classifier masses. By using the fragment spectra, proteins were identified by database searches. For comparison and further validation, classifier masses were also fragmented and analyzed using HPLC-/Matrix-assisted laser desorption ionization (MALDI) time-of-flight/time-of-flight (TOF/TOF) mass spectrometry. Desmin and lumican precursor were examples of proteins found in aortic samples at higher abundances than in coronary samples. Similarly, adenylate kinase isoenzyme was found in coronary samples at a higher abundance. The described methodology could also be feasible in search for specific biomarkers in plasma or serum for diagnostic purposes.

Status of the Micro Vertex Detector of the Compressed Baryonic Matter Experiment (2010)

Michael Deveaux Samir Amar-Youcef Norbert Bialas Dennis Doering Melissa Domachowski Horst Düring Ingo Fröhlich Tetyana Galatyuk Jan Michael Christian Müntz Sarah Ottersbach Paul Scharrer Christoph Schrader Joachim Stroth Tobias Tischler Christian Trageser Bernhard Wiedemann Jérome Baudot Grégory Bertolone Nathalie Chon-Sen Claude Colledani Rita De Masi Andrei Dorokhov Wojchiech Dulinski Jean-Charles Fontaine Mathieu Goffe Abdelkader Himmi Christine Hu Kimmo Jaaskelainen Michal Koziel Frédéric Morel Fouad Rami Mathieu Specht Isabelle Valin Marc Winter Christina Dritsa Selim Seddiki Franz M. Wagner

The CBM experiment will investigate heavy-ion collisions at beam energies from 8 to 45 AGeV at the future accelerator facility FAIR. The goal of the experiment is to study the QCD phase diagram in the vincinity of the QCD critical point. To do so, CBM aims at measuring rare probes among them open charm. In order to identify those rare and short lived particles despite the rich combinatorial background generated in heavy ion collisions, a micro vertex detector (MVD) providing an unprecedented combination of high rate capability and radiation hardness, very light material budget and excellent granularity is required. In this work, we will discuss the concept of this detector and summarize the status of the R&D.

Fragmentation of exotic oxygen isotopes (2003)

Armin Leistenschneider Thomas Aumann Konstanze Boretzky Luiz Felipe Canto Brett Vern Carlson Dolores Cortina Ushasi Datta Pramanik Thomas W. Elze Hans Emling Hans Geissel Alexander Grünschloss Kerttuli Helariutta Margareta Hellström Mahir S. Hussein Stamenko Ilievski Jens-Volker Kratz Reinhard Kulessa Le Hong Khiem Edvard Lubkiewicz Gottfried Münzenberg Rudrayoti Palit Peter Reiter Christoph Scheidenberger Karl-Heinz Schmidt Haik Simon Klaus Sümmerer Eugeniusz Wajda Wladek Walús

Abrasion-ablation models and the empirical EPAX parametrization of projectile fragmentation are described. Their cross section predictions are compared to recent data of the fragmentation of secondary beams of neutron-rich, unstable 19,20,21O isotopes at beam energies near 600 MeV/nucleon as well as data for stable 17,18O beams.

Measuring shear viscosity using correlations (2007)

Sean Gavin Mohamed Abdel-Aziz

Measurements of transverse momentum fluctuations can be used to determine the shear viscosity [1]. We use current data to estimate the viscosity-to-entropy ratio in the range from 0.08 to 0.3, and discuss how future measurements can reduce this uncertainty.

Ion beam analysis based on cellular nonlinear networks (2011)

Vanessa Senger Ronald Tetzlaff Hermine Reichau Ulrich Ratzinger

The development of a non- destructive measurement method for ion beam parameters has been treated in various projects. Although results are promising, the high complexity of beam dynamics has made it impossible to implement a real time process control up to now. In this paper we will propose analysing methods based on the dynamics of Cellular Nonlinear Networks (CNN) that can be implemented on pixel parallel CNN based architectures and yield satisfying results even at low resolutions.

Universitätspublikationen

Refine

Author

Year of publication

Document Type

Language

Has Fulltext

Keywords

Institute

1344 search hits