Institute of Engineering and Computational Mechanics
Parametric Model Order Reduction in Elastic Multibody Systems
Project description
For many applications in elastic multibody dynamics, the mass-, stiffness-, and input matrix of the elastic bodies have to be modelled parametrically in order to obtain meaningful models. Classical model order reduction methods cannot be applied for such kind of problems since they do not preserve the parameter dependency in the reduced models. Therefore, the aim of this research project is the development and the investigation of parameric model order reduction methods which preserve the parameter dependency in the reduced order model explicitly. In the following some applications of parametric model order reduction will be presented.
Structural Optimization with Reduced Order Models
The increasing demand for energy and resources efficient technical products makes the use of lightweight structures necessary. Usually, the Finite-Element-Method is used for modelling complex components. Afterwards these models can be used for optimization to achieve the desired weight reduction. However, the numerical effort for solving these optimization problems may become tremendous for fine discretized models making the optimization problems unsolvable in acceptable computation times. The goal of this research project is therefore the development of methods for parametric model order reduction for structural and shape optimization problems to enable efficient solutions to these kinds of problems.
Figure 1: Shape optimzation of a cantilever beam.
Simulation of Moving Loads
Moving Interactions, as they occur, for instance, in gear-wheels, sliding components, turning and milling processes (see Figure 2), lead to a parameter dependent input matrix. One parametric reduction method which provides very satisfying results for moving point forces is based on the interpolation of reduced system matrices. Thereby, support systems for given parameter values are calculated offline and transformed to allow the interpolation. While the system is integrated in time with a moving load or interaction, the system matrices for any parameter value are generated by interpolation. The implementation of this method in MOREMBS and the implementation of the integration in the multibody simulation program Neweul-M² enables the efficient simulation of elastic multibody systems with moving interaction.
Figure 2: Linear axle with moving load
Figure 3: Workflow for parametric model reduction with interpolation of local reduced system matrices
Promotion
The research project is granted by the German Research Foundation (German: Deutsche Forschungsgemeinschaft):
Sonderforschungsbereich 1244 at the University of Stuttgart: "Adaptive Hüllen und Strukturen für die gebaute Umwelt von morgen"