This thesis is about a fast and accurate reanalysis procedure. It focuses on elastic structures used in parametric design, optimization and reliability analysis. The mathematical description bases on the manifold structure of parametric problems described by a non iterative system of equations. Therefore the method will be called approximate response manifold (ARM). The finite element method is used as the underlying method of simulating these structures. An efficient set of solution vectors is derived to embed the solution manifold of the parametric design space within an Euclidean subspace of the original problem. The solution is calculated using the Hamilton principle and the Rayleigh-Ritz procedure. Different strategies to obtain the solution vectors are discussed, depending on the parameter range of the underlying reanalysis problem. After introducing the procedure, the numerical effort and complexity is outlined in comparison to a traditional analysis. The procedure is designed for large and complex optimization problems, characteristic for real world problems in industrial structural design, such as the automotive design process. Focused on a real world problem the efficiency of the formulation is discussed and compared, in terms of accuracy and numerical effort, to a traditional surrogate model such as the response surface method (RSM). It is shown that the method can solve large scaled linear static structural problems with up to hundreds of design variables very fast on low end computing systems. It is therefore suitable for multi-criteria optimization, trade off studies and reliability analyses.
«
This thesis is about a fast and accurate reanalysis procedure. It focuses on elastic structures used in parametric design, optimization and reliability analysis. The mathematical description bases on the manifold structure of parametric problems described by a non iterative system of equations. Therefore the method will be called approximate response manifold (ARM). The finite element method is used as the underlying method of simulating these structures. An efficient set of solution vectors is...
»