A5045125750- Time Series Analysis and Forecasting
- Stock Market Forecasting Methods
- Fault Detection and Control Systems
- Anomaly Detection Techniques and Applications
- Neural Networks and Applications
- Forecasting Techniques and Applications
- Data Stream Mining Techniques
- Image and Signal Denoising Methods
- Industrial Automation and Control Systems
- Advanced Neural Network Applications
- Mineral Processing and Grinding
- Complex Systems and Time Series Analysis
- Industrial Vision Systems and Defect Detection
- Image and Object Detection Techniques
South Westphalia University of Applied Sciences
2022-2025
IRD Fuel Cells (Denmark)
2024
Many applications in the field of bulk material handling require determination Particle Size Distribution (PSD) handled materials. Typically, this can only be achieved by manual sampling and a subsequent laboratory analysis. However, availability novel image segmentation algorithms based on Convolutional Neural Networks (CNNs) leads to possibility for an implementation visual PSD estimation techniques. One prominent example such approaches is U-Net architecture. one fundamental problem...
One of the major goals in bulk good systems is to provide high-quality material. Sometimes corrupted by foreign objects, due insufficient process observations, human mistakes, or unforeseen events. These particles are unwanted process. Furthermore, they can lead unnecessary plant shutdowns even lock sub-parts plant. Current solutions work with filter techniques eject objects and ensure high quality. However, many difficult implement constraints give no precise information about corruption...
Increasing the linear modulation range
An active area of research in the field Machine Learning is optimization network structures, including activation functions. However, selecting a suitable function not that simple and usually requires an elaborate empirical process. Unfortunately, this process very resource intensive does always lead to best solution. This paper presents novel approach select, compose generate adaptive functions without any supervision. The best-suited learned from collection candidates while training neural...
In Machine Learning, large models need to have access a huge amount of training data. This requirement applies many applications in an industrial environment. Furthermore, specific processes it is not easy obtain such data for different reasons ranging from privacy, security, and even process affectations. Therefore, this work proposes the creation synthetic time series datasets simulate out given subset functional relationships. Moreover, using transfer learning improve performance four...
Denoising sequences of time series is one the elementary preprocessing steps in data mining. Current statistical methods work on univariate input stream and do not obtain long dependencies over whole series. Due to advanced architectures learning techniques, machine approaches are getting more effective. However, none existing propose a learnable denoising process select interpretable components out given subset dependencies. Therefore, we novel method field learning. In particular,...