Dilated Convolutions have been shown to be highly useful for the task of image segmentation. By introducing gaps into convolutional filters, they enable the use of larger receptive fields without increasing the original kernel size. Even though this allows for the inexpensive capturing of features at different scales, the structure of the dilated convolutional filter leads to a loss of information.We hypothesise that inexpensive modifications to Dilated Convolutional Neural Networks, such as additional averaging layers, could overcome this limitation. In this project we test this hypothesis by evaluating the effect of these modifications for a state-of-the art image segmentation system and compare them to existing approaches with the same objective.Our experiments show that our proposed methods improve the performance of dilated convolutions for image segmentation. Crucially, our modifications achieve these results at a much lower computational cost than previous smoothing approaches.

No abstract

In thermal recycling plants high performance fans are needed to transport gas through the process. For this application, the fans have to fulfill very special requirements. The pyrolysis and gasification processes transform different kinds of waste, e. g. biomass and municipal scrap waste, electronic or industrial waste, into highly clean efficient synthetic gas, i.e. usable energy. Therefore, the thermal operating range of these fans lies between ambient temperatures of about 20°C and 530°C. Due to the requirements and design space of the recycling facility, new radial fans were developed according to these particular specifications. Two different impellers were designed with an extended analytical method, as described in Epple et al. [1] and [2]. The first one was designed with a parallel flat shroud and the second one with a conical shroud. In order to fulfill the manufacturer requirements of ease of manufacturing, the impellers were designed with circular arc blades. With these two impellers and at a temperature of 530 °C, as it is needed in the recycling plants, a first Bommes spiral casing was optimized adjusting the spiral casing opening angle and the Bommes parameter, i.e. the ratio of the spiral casing width to the impeller exit width, BSC/b2 (Epple et al. [3]). The results show that the predictions of Bommes work well with impellers with conical shrouds, but for impellers with parallel shrouds the Bommes prediction does not work properly.

The flow under sluice gates is nowadays frequently still determined by empirical approaches, based on the Bernoulli equation and a specific discharge coefficient which depends on the geometry of the sluice gate. This discharge coefficient is determined either from potential theory approaches or from a variety of experimental series. Based on Malcherek’s new momentum balance theory approach, the flow under inclined sluice gates can be described now on a physical approach. While the pressure distribution of the vertical sluice gate, which was investigated by the authors in previous works, results in a horizontal acting force only, the pressure distribution of the inclined sluice gate brings an additional force component acting in the vertical direction. The magnitude and the sign of the force in the vertical direction depends on the sign and the value of the angle of inclination. With the commercial CFD solver Star-CCM+ from Siemens PLM, the flow under inclined sluice gates was investigated in detail, to be compared with Malcherek’s momentum balance theory. For this basic investigation, boundary layer effects and 3D effects were first neglected. For various angles of inclination and water levels the pressure distribution on the ground under the sluice gate and on the surface of the sluice gate itself, were evaluated and mathematical functions were introduced to describe the pressure distributions. These results were compared with the results from the vertical sluice gate. Also the discharge velocity and the volumetric flow rate were analyzed and compared with the Malcherek’s new momentum balance theory for inclined sluice gates.

Fans in industrial plants can be exposed to a strong erosion load due to particle flows. In the present work, the erosion behavior for large radial fans with spiral casings is investigated using the Finnie erosion model, see [1] [2]. Theoretical approaches concerning particle velocity and particle impact angle are validated by numerical methods. For this purpose, a baseline impeller and a parameterized baseline spiral casing have been designed and simulated using computational fluid dynamics. Than different geometrical variations of the spiral casing shape and the blade shapes of the impellers have been designed and simulated in order to determine their respective influence on the erosion behavior as well as on the performance characteristics. Finally, recommendations for an optimal design are presented and explained in detail.