Mechatronics
Refine
Year of publication
- 2020 (6)
Document Type
- Master's Thesis (6)
Has Fulltext
- yes (6)
Is part of the Bibliography
- no (6)
Keywords
- AI on the Edge (1)
- CNN for Embedded System (1)
- Computational-intelligence (1)
- Diffuser (1)
- Finite-element-methods (1)
- Jade (1)
- KI für Embedded System (1)
- Machine learning (1)
- Optimisation-algorithms (1)
- Partial-differential-equations (1)
In recent years, much research has been done on medical laser applications inside the human body, as they are minimally invasive and therefore have fewer side effects and are less expensive than conventional therapies. In order to bring the laser light into the human body, a glass fibre with a diffuser is needed. The goal of this master thesis is the characterization and production of fibre optic diffusers that can be used for the three therapeutic applications: photodynamic therapy, laser-induced thermotherapy and endovenous laser therapy. For this purpose the following goals have to be achieved:
- Optimization of the efficiency and homogeneity of internally structured diffusers
- Examine damage thresholds of the diffusers in the tissue using a crash test
- Achieving a better understanding of the decouple mechanism with a simulation
Using an ultra-short pulse laser, modifications could be introduced into the fibre in this way that the radiation profile is homogeneous and the decoupling efficiency is 68.3 %. It was discovered that the radiation profile depends on the wavelength. Attempts have been made to improve the decoupling efficiency by mirroring the distal end of the fibre. The mirror reflects the remaining light back into the fibre, so that it is also decoupled lateral on the modifications. Vapor-deposited aluminum with physical vapor deposition is a promising approach. However, the adhesion of the coating must be improved or the coating must be protected by a mechanical cover, otherwise it will flake off too quickly.
In a crash test, it was shown that the glass fibre diffusers can withstand 20 W laser power for 300 s without visible change. In an ex vivo test, the coagulation zone in the tissue was examined and it was showed that the diffusers radiate radially homogeneously. Using a ray trace simulation, the course of the light rays in the fibre was examined and the correlation of modification width and length with the decoupling efficiency was investigated. It was discovered that there are helical light rays in the fibre, which cannot be decoupled by modifications in the fibre centre.
This master thesis investigates a Computational Intelligence-based method for solving PDEs. The proposed strategy formulates the residual of a PDE as a fitness function. The solution is approximated by a finite sum of Gauss kernels. An appropriate optimisation technique, in this case JADE, is deployed that searches for the best fitting parameters for these kernels. This field is fairly young, a comprehensive literature research reveals several past papers that investigate similar techniques.
To evaluate the performance of the solver, a comprehensive testbed is defined. It consists of 11 different Poisson equations. The solving time, the memory consumption and the approximation quality are compared to the state of the art open-source Finite Element solver NGSolve. The first experiment tests a serial JADE. The results are not as good as comparable work in the literature. Further, a strange behaviour is observed, where the fitness and the quality do not match. The second experiment implements a parallel JADE, which allows to make use of parallel hardware. This significantly speeds up the solving time. The third experiment implements a parallel JADE with adaptive kernels. It starts with one kernel and introduce more kernels along the solving process. A significant improvement is observed on one PDE, that is purposely built to be solvable. On all other testbed PDEs the quality-difference is not conclusive. The last experiment investigates the discrepancy between the fitness and the quality. Therefore, a new kernel is defined. This kernel inherits all features of the Gauss kernel and extends it with a sine function. As a result, the observed inconsistency between fitness and quality is mitigated.
The thesis closes with a proposal for further investigations. The concepts here should be reconsidered by using better performing optimisation algorithms from the literature, like CMA-ES. Beyond that, an adaptive scheme for the collocation points could be tested. Finally, the fitness function should be further examined.
Many test drives are carried out in the automotive environment. During these test drives many signals are recorded. The task of the test engineers is to find certain patterns (e.g. an emergency stop) in these long time series. Finding these interesting patterns is currently done with rule based processing. This procedure is very time consuming and requires a test engineer with expertise. In this thesis it is examined if the emerging field of machine learning can be used to support the engineers in this task. Active Learning, a subarea of machine learning, is used to train a classifier during the labeling process. Thereby it proposes similar windows to the already labeled ones. This saves the annotator time for searching or formulating rules for the problem. A data generator is worked out to replace the missing labeled data for tests. The custom performance measure “proportion of seen samples” is developed to make the success measurable. A modular software architecture is designed. With that, several combinations of Time Series Classification algorithms and query strategies are compared on artificial data. The results are verified on real datasets, which are open source available. The best performing, but computational intensive solution is an adapted RandOm Convolutional KErnel Transform (ROCKET). The custom query strategy “certainty sampling” shows the best results for highly imbalanced datasets.
Die cloud-basierte Verarbeitung von Datenströmen von IoT-Geräten ist aufgrund hoher Latenzzeiten für zeitkritische Anwendungen nur beschränkt möglich. Fog Computing soll durch Nutzung der Rechen- und Speicherkapazitäten von lokal vorhandenen Geräten eine zeitnahe Datenverarbeitung und somit eine Verringerung der Latenzzeit ermöglichen. In dieser Arbeit werden Anforderungen an ein Fog Computing-Framework erhoben, das die dynamische Zuweisung und Ausführung von Services auf ressourcen-beschränkten Geräten in einem lokalen Netzwerk zur dezentralen Datenverarbeitung ermöglicht. Zudem wird dieses Framework prototypisch für mehrere Transportkanäle, unterschiedliche Betriebssysteme und Plattformen realisiert. Dazu werden die Möglichkeiten der Skriptsprache Lua und des Kommunikationsmechanismus Remote Procedure Call genutzt. Das Resultat ist ein positiver Machbarkeitsnachweis für Fog Computing-Funktionalitäten auf ressourcenbeschränkten Systemen. Zudem werden eine geringere Latenz und eine Reduktion der Netzwerklast ermöglicht.
Pump-Probe-Elastographie
(2020)
Krebs ist die zweit häufigste Todesursache in Deutschland. Seine frühzeitige Detektion ist wichtig für eine erfolgreiche Behandlung. Die Detektion und Charakterisierung der Tumore kann unter anderem anhand ihrer mechanischen Eigenschaften erfolgen.
Die Pump-Probe-Elastographie (PPE) ist eine neue und vielversprechende Methode um die mechanischen Eigenschaften von Gewebe durch optische Anregung und Detektion der dadurch entstehenden akustischen Wellen zu charakterisieren.
In dieser Arbeit wurde ein PPE-Mikroskop aufgebaut und anschließend untersucht, wie die Wellenfronten möglichst deutlich sichtbar werden und welchen Einfluss die Pulsenergie auf die erzeugten Wellen hat. Dies geschah anhand zweier Proben – Wasser und Glas. Beide sind transparent und bieten daher auch die Möglichkeit der Messung im Volumen. Wasser unterscheidet sich zudem von Glas, da es kompressibel ist und somit die Entstehung von Stoßwellen erlaubt. In Glas wiederum können nicht nur Longitudinalwellen, sondern auch Transversalwellen und Rayleigh-Wellen entstehen.
Als Ergebnis dieser Arbeit wird gezeigt, dass die Messergebnisse in Wasser abhängig von der Pulsenergie sind, denn die entstehenden Stoßwellen breiten sich mit Überschallgeschwindigkeit aus, was bei zukünftigen Messungen mit Zellgewebe zu beachten ist. In Glas hat die Pulsenergie zwar keinen Einfluss auf die Wellengeschwindigkeit, jedoch werden die Messungen mit zunehmender Energie deutlicher. Auch ist bei Messungen im Volumen nur die Longitudinalwelle zu sehen. Misst man jedoch auf der Glasoberfläche, so entsteht auch eine Druckwelle in der Luft, welche wiederum abhängig von der Pulsenergie ist. Zusätzlich ist in einigen Messungen eine weitere Welle zu erkennen, welche unabhängig von der Pulsenergie ist und sich im Glas, oder auf dessen Oberfläche ausbreitet.
The classification of waste with neural networks is already a topic in some scientific papers. An application in the embedded systems area with current AI processors to accelerate the inference has not yet been discussed. In this master work a prototype is created which classifies waste objects and automatically opens the appropriate container for the object. The area of application is in the public space.
For the classification a dataset with 25,681 images and 11 classes is created to re-train the Convolution Neuronal Networks EfficientNet-B0, MobileNet-v2 and NASNet-mobile. These Convolution Neuronal Networks run on the current Edge \acrshort{ai} processors from Google, Intel and Nvidia and are compared for performance, consumption and accuracy.
The master thesis evaluates the result of these comparisons and shows the advantages and disadvantages of the respective processors and the CNNs. For the prototype, the most suitable combination of hardware and AI architecture is used and exhibited at the university fair KasetFair2020. An opinion survey on the application of the machine is conducted.
