Engineering design of a radar structure using design of experiments and structural optimization

Abstract

Models of complex systems contain a large amount of different parameters and several constraints, which make structural optimization very time consuming. Therefore, Saab AB wanted to develop a method based on Design of Experiments to use for optimization within structural analysis and concept generation. Consequently, the purpose of the project was to develop a method that efficiently evaluates concepts and new products with regard to structural analysis in product development. The method was used to ease the evaluation of different concepts, as well as the understanding of the impact that the parameters had on the system. The method was developed by using three different trial cases in order to learn to use the different software, as well as developing and fine tuning the method. Thereafter, a simplified version of an authentic, complex problem from Saab was evaluated using the developed method. Different Design of Experiments methods were investigated while executing the first sample cases. Thereafter, only one of the Design of Experiments method, called Hammersley, was used in order to make the simulations equivalent. The second and third cases were created in order to learn how to, for example, model and parameterize the concepts. The advanced case consisted of a simplified version of an existing analysis model from Saab's product portfolio. The original existing analysis and simplified models were both evaluated and the results obtained from the two cases were compared to each other. This was done in order to be certain that the behaviour of the simplified model resembled the behaviour of the original model. When the simplified model was assumed to be sufficiently accurate, a new concept based on the simplified model was created and analyzed using the developed method. The method used structural optimization and informed about which parameters that in uenced the specified objective function and which parameters that could be neglected. The advanced case had six different design variables and the number of values each parameter had ranged from 2 to 11. In the software HyperStudy, the Design of Experiments method Hammersley calculated that 105 concepts had to be created in order to get a sufficient optimization. Thus, 105 evaluations were executed, since one concept needs one evaluation. By using this method, the number of parameters was, in the advanced case, reduced from six to two, which saves both computational and optimization time.