Refine
Year of publication
Document Type
Keywords
- Aerodynamik (1)
- Bearings (1)
- Blade Design (1)
- Density (1)
- Mehrkriterien-Optimierung (1)
- Optimierung (1)
- Optimization (1)
- Prozessintegration (1)
- Verdichter (1)
- Verdichterauslegung (1)
Institute
A Method for Using Multi-objective Optimization within the Rolls-Royce meanline Prediction Process
(2005)
Due to the demand of more and more detailed FEM simulation results of whole aero-engine models, the number of DOF of such models (Fig. 1) increases dramatically. Although the available computational power for the solution of such highly nonlinear, dynamic boundary value problems also increased during the last years, the overall computational time of these high-fidelity models for the simulation of a few milliseconds of the running engine is in the order of several weeks even on high performance clusters with thousands of CPU cores. Typically, for such highly dynamic FEM computations explicit time integration schemes are used in order to solve the equations of motion. To keep such a numerical integration stable, the time step size may not exceed a certain critical
time step size. Unfortunately, the critical time step size of an explicit time integration algorithm for the engine models under consideration is in the order of 10-8. This means that the simulation of a running engine over a time span of a few seconds results in a computational time of several years with clusters, which are available nowadays. Therefore, implicit time integration schemes, which are unconditionally stable and allow much bigger time steps, have to be used. Since in an implicit integration, in contrast to an explicit one, an equilibrium iteration is always necessary, the computational costs for an implicit time step are much higher than for an explicit time step.
Almost all commercial FE codes use the classical Newmark implicit time integration scheme. It will be shown that this algorithm fails especially for the simulation of fast rotating structures if bigger time steps are used. This is demonstrated at the example of academic examples as well as for more realistic rotor models. To overcome the problems of the Newmark integration, a 3-Point-Newmark-
Euler-Backward integration scheme [1,2] has been selected, which consists of alternating Newmark and 3-Point-Euler-Backward steps. The additional interpolation information for the Euler-step is provided by the previous Newmark step. This integration procedure has been implemented and successfully applied to the FEM simulation of fast rotating structures, which will be also demonstrated during the presentation.
Considerations About the Necessary Mesh Density of Bearings in Detailed Finite Element Models
(2014)
In order to predict properly the behavior of very complex mechanical models, much focus needs to be done on the level of model details as well as the techniques used to predict component movements based on mechanical and thermal loads. In particular bearings and joints play a significant role in big model assemblies such as aero-engines where most of the structures are modeled by solid elements rather than using traditional shell or beam elements. This approach is very complex and expensive in terms of computational effort however can lead to increased simulation accuracy when done properly.
This publication will answer questions related to the modeling techniques required for detailed bearings (not only rotor bearings) and joints of an aero-engine application where these components are not idealized or simplified, i.e. the ball in a ball joint is meshed by solid elements. The investigation will discuss the question about the number of elements needed to mesh a ball joint in such a way that a correct rotation is possible. If the mesh is too coarse, a rotation would be impossible due to the faceting of the rotating parts. It will be shown that a proper rotation depends on a correct relation between the radius of the inner ball, the gap size of the bearing and the number of nodes in the circumferential direction. For this relation an analytical formula has been derived and successfully tested at some examples. Furthermore, some finite element codes offer the possibility of smoothed contact surfaces. These options improve the situation but they might cause issues too. For this reason these smoothed contact options are also considered under the aspects mentioned above.
Copyright © 2014 by Rolls-Royce Deutschland Ltd & Co KG