Refine
Year of publication
- 2022 (3) (remove)
Document Type
- Bachelor Thesis (2)
- Conference Proceeding (1)
Has Fulltext
- yes (3)
Is part of the Bibliography
- no (3)
Keywords
- Neuronales Netz (3) (remove)
Recent developments in the field of deep learning have shown promising advances for a wide range of historically difficult computer vision problems. Using advanced deep learning techniques, researchers manage to perform high-quality single-image super-resolution, i.e., increasing the resolution of a given image without major losses in image quality, usually encountered when using traditional approaches such as standard interpolation. This thesis examines the process of deep learning super-resolution using convolutional neural networks and investigates whether the same deep learning models can be used to increase OCR results for low-quality text images.
We present a novel long short-term memory (LSTM) approach for time-series prediction of the sand demand which arises from preparing the sand moulds for the iron casting process of a foundry. With our approach, we contribute to qualify LSTM and its combination with feedback-corrected optimal scheduling for industrial processes.
The sand is produced in an energy intensive mixing process which is controlled by optimal scheduling. The optimal scheduling is solved for a fixed prediction horizon. One major influencing factor is the sand demand, which is highly disturbed, for example due to production interruptions. The causes of production interruptions are in general physically unknown. We assume that information about the future behavior of the sand demand is included in current and past process data. Therefore, we choose LSTM networks for predicting the time-series of the sand demand.
The sand demand prediction is performed by our multi model approach. This approach outperforms the currently used naive estimation, even when predicting far into the future. Our LSTM based prediction approach can forecast the sand demand with a conformity up to 38 % and a mean value accuracy of approximately 99%. Simulating the optimal scheduling with sand demand prediction leads to an improvement in energy savings of approximately 1.1% compared to the naive estimation. The application of our novel approach at the real production plant of a foundry proves the simulation results and verifies the capability of our approach.
Bis heute ist völlig unbekannt, ob wir allein im Universum sind. Um auf dieses Thema eine Antwort zu finden, überprüft diese Bachelorarbeit, ob Convolutional (CNN) und Recurrent Neural Networks (RNN) für die Erkennung außerirdischer Signale geeignet sind.
Das Ziel war dabei, in einem Datensatz bestehend aus Spektrogrammen mehr als 50% aller außerirdischer Signale zu erkennen, da nur so ein Neuronales Netzwerk ein besseres Resultat als eine zufällige Klassifikation liefert, bei der im Mittel 50% aller Signale erkannt werden.
Dabei zeigte sich, dass sich mit beiden Varianten der Neuronalen Netzwerke bis zu 90% aller Signale erkennen lassen, die Vorhersagen von CNNs allerdings verlässlicher sind. RNNs bieten hingegen aufgrund ihrer geringeren Größe einen deutlich leichtgewichtigeren Ansatz und führen zu einer signifikanten Speicherersparnis.
Daraus folgt, dass Neuronale Netzwerke bei der Suche nach außerirdischem Leben im Universum helfen können, um die Frage „Sind wir allein im Universum?“ endgültig zu beantworten.