A Hybrid Design Optimization Method using Enriched Craig-Bampton Approach

A hybrid design optimization method is presented which combines a number of techniques such as Component Mode Synthesis (CMS), Design of Computer Experiments and Neural Networks for surrogate modeling with Genetic Algorithms and Sequential Quadratic Programming for optimization. In the method, the FE analysis is decomposed and reduced by a well-known CMS technique called the Craig-Bampton method. Since the optimization method requires CMS calculations of the updated model at each of its iterations due to the changes in the design variables, one can either reuse the reduction basis of the initial components or compute new reduction basis for the condensation of the system matrices. The first option usually leads to inaccurate results and the last one increases the omputation time. In the method, instead of using one of these options, the Enriched Craig-Bampton method, proposed by Masson et al., is employed for efficient optimization. New basis for the modified components are generated by extending the corresponding initial reduction basis with a set of static residual vectors which are calculated using prior knowledge of the initial component designs. Thus, time consuming complete component analyzes are prevented. A theoretical test problem is used for the demonstration of the method.