06 Fakultät Luft- und Raumfahrttechnik und Geodäsie
Permanent URI for this collectionhttps://elib.uni-stuttgart.de/handle/11682/7
Browse
Search Results
Item Open Access Application of neural networks and transfer learning to turbomachinery heat transfer(2022) Baumann, Markus; Koch, Christian; Staudacher, StephanModel-based predictive maintenance using high-frequency in-flight data requires digital twins that can model the dynamics of their physical twin with high precision. The models of the twins need to be fast and dynamically updatable. Machine learning offers the possibility to address these challenges in modeling the transient performance of aero engines. During transient operation, heat transferred between the engine’s structure and the annulus flow plays an important role. Diabatic performance modeling is demonstrated using non-dimensional transient heat transfer maps and transfer learning to extend turbomachinery transient modeling. The general form of such a map for a simple system similar to a pipe is reproduced by a Multilayer Perceptron neural network. It is trained using data from a finite element simulation. In a next step, the network is transferred using measurements to model the thermal transients of an aero engine. Only a limited number of parameters measured during selected transient maneuvers is needed to generate suitable non-dimensional transient heat transfer maps. With these additional steps, the extended performance model matches the engine thermal transients well.Item Open Access Prediction of compressor blade erosion experiments in a cascade based on flat plate specimen(2022) Lorenz, Max; Klein, Markus; Hartmann, Jan; Koch, Christian; Staudacher, StephanErosion is an essential deterioration mechanism in compressors of jet engines. Erosion damage predictions require the determination of erosion rates through flat plate experiments. The applicability of the erosion rates is limited to conditions that are comparable to the prevailing boundary conditions of the flat plate experiment. A performed dimensional analysis enables the correct transfer of the flat plate erosion rates to the presented physical calculation model through limits in spatial and time resolution. This efficient approach avoids computationally intensive single-impact computations. The approach features a re-meshing procedure that adheres to the limits derived by the dimensional analysis. The computation approach is capable of describing local geometry changes on cascade compressor blades which are exposed to erosive particles. A linear erosion cascade experiment performed on NASA Rotor 37 provides validation data for the calculated erosion-induced shape change. Arizona Road Dust particles are used to deteriorate Ti-Al6-4V compressor blades. The experiment is performed at an incidence of i = 7°and Ma = 0.76 representing ground idle conditions. The presented parametric study for element size and time step revealed preferable values for the presented computation. Calculations performed with the determined values showed that the erosion prediction is within the measurement tolerance of the experiment and, therefore, high accordance between the computation and the experiment is achieved. To extend the current state of the art, it is demonstrated that the derived discretization is decisive for the correct reproduction of the eroded geometries and fitting parameters are no longer needed. The good agreement between the experimental measurements and the calculated results confirms the correct application of the physical model to the phenomenology of erosion. Thus, the presented physical model offers a novel approach to adapting deterioration mechanisms caused by erosion to any compressor blade geometry.Item Open Access Experimental and numerical investigation into the effect of surface roughness on particle rebound(2022) Altmeppen, Johannes; Sommerfeld, Heike; Koch, Christian; Staudacher, StephanErosion damage and particle deposition are crucial wear phenomena in gas turbine engines. As a result, compressor efficiency decreases, stability margin reduces, and maintenance cost increases. Hence, predicting these phenomena in an accurate manner is of paramount importance for a cost-efficient, safe, and sustainable operation. Erosion and particle deposition in the annulus are affected by particle transportation in the fluid and particle-wall interaction. The latter involves the particle impact, the potential damage of the surface and/or the particle, and the particle rebound. Particle rebounds are statistical in nature due to the target surface roughness, the variability in particle sizes, and superimposed effects caused by particle shapes as well as particle rotation and particle break-up during contact. Multiple studies investigated the statistics of particle rebound, providing empirical-based models for median and spread. However, modeling the particle-wall interaction and its data spread on a transparent physical basis allows separating the effect of target roughness from superimposed effects. The presented article pursues this objective by assessing the statistical spread of particle rebound data through multiple techniques and utilizing their interdependencies. It combines experimental, numerical, and analytical considerations. For the first time, coherent boundary conditions for the experimental, numerical and analytical setup allow the distinction of the effect of roughness from the integral effect of the superimposed phenomena. A sandblast test rig equipped with laser measurement equipment was used to measure particle rebound from flat titanium and stainless steel plates at different angles. The numerical setup is developed under OpenFOAM 6 using a RANS solver for transient simulations with compressible media in combination with one-way coupled particle flows. The numerical model includes the rebound spread model proposed by Altmeppen et al. combined with the quasi-analytical rebound model proposed by Bons et al. The statistical spread of particle rebound is investigated for roughness levels that are similar to the ones of deteriorated high-pressure compressor blades as discussed by Gilge et al. The measured surface roughness of the experimentally investigated targets is used as input parameters to the numerical framework. The rebound statistics obtained in the numerical simulation are compared to the rebound data measured in the experiment. Based on this study, conclusions are drawn about which part of the rebound spread is attributable to surface roughness and which is caused by superimposed effects. It was found that the effect of surface roughness as characterized by Altmeppen et al. contributes in the order of 46 % to the rebound spread for small impact angles. However, the share in spread due to roughness gradually decreases with increasing global impact angles to a level of 13 % for angles close to 90°. The remaining percentage of rebound spread is attributed to superimposing phenomena. In addition to the rolling and sliding of aspherical particles, further phenomena such as plastic deformation and erosion of the roughness peaks during contact and the associated dissipation of energy gain in importance for steeper impact angles. Therefore, the effect of surface roughness should not be neglected in numerical simulations of particle-laden flows. Modeling the superimposed phenomena which are observed to be dominating at high impact angles opens up a further field of research.