Concurrent Subspace Architecture For Layout Optimization Design Of Composite Thin-walled Stiffened Structure

The weight of aircraft structure is one of the most fundamental performance index,and the reduction of the structure weight is the permanent theme of the aircraft structure engineers.The stiffened structures are widely used in the field of aerospace for its advantages of saving structure material and reducing structure weight.As the composite material is getting a higher percentage of use,the optimization problem of the composite stiffened panel structures are attracting wider attention recently.The optimization design methods of the structure and its development history are discussed in this paper.The common patterns of the stiffened structures,the basic concept of the buckling instability and the layout optimization design of the stiffened structure are introduced.To solve the layout optimization problem of the stiffened thin-walled plate structure,the parallel subspace method is adopted,which divides the whole optimization problem into three parallel sub-problems: layout optimization problem,the size optimization problem and the lamination optimization problem.The design variables are classified into three categories corresponding to each sub-problem,all of which with the same optimization goal of the lightest structure weight.Each sub-problem takes the variables from the other two sub-problems as the state variables during the process of optimization.When the process of optimization comes to an end,the value of the design variable is adjusted according to the optimization results of each sub-problem and their difference value to the mean value of the three sub-problem results.The adjusted variable values are then taken into account for the next iteration until the convergence condition is satisfied.The optimum lamination sequence result is obtained by using the equivalent bending stiffness method after the whole optimization process.To deal with the two sub-problems of the layout optimization and the size optimization,the parametric modeling method,the experiment design and the surrogate modeling techniques are adopted.The layout variables and the size variables,from which the sample points are chosen,are taken as the design variables.The Kriging surrogate model is constructed and its accuracy is improved by updating techniques.The optimum result is obtained by applying the multi-island genetic algorithm to the surrogate model.In other words,a set of test points plays a role in the optimization of three subspaces at the same time,which increases the utilization of the test set and improves the efficiency of computation.Two computational examples verifying the effectiveness and the practicability of the above-mentioned parallel subspace method are presented.