Optimization of the grid-stiffened structure via PSO-LSSVR surrogate modeling and sensitivity analysis

The grid-stiffened structure is the key load-carrying sections of launch vehicles such as the storage tank and the interstage section. Thus, its load-carrying performance is directly related to the load-carrying efficiency of a rocket. In this paper, a structural optimization method based on the load-carrying efficiency was proposed for grid-stiffened structures. Firstly, a general 3D parametric geometrical modeling method for grid-stiffened structure was proposed, which could optimize the position of stiffeners and incorporate the key geometric features such as welds. Based on this method, the buckling load of the structure considering the internal pressure was figured out by finite element analysis, and the solution accuracy was verified. Then, a surrogate model was constructed based on the Particle Swarm Optimization-Least Squares Support Vector Regression algorithm for accurately and efficiently predicting the buckling load of grid-stiffened structures. Subsequently, the sensitivity of each structural parameter to the buckling load was evaluated using the Sobol method to identify the key ones. Finally, targeting these key parameters, a load-carrying efficiency optimization model was proposed, considering constraints such as mass and geometric constraints. The results showed that the load-carrying efficiency of the optimized structure increased by 11.20%. The methods proposed in this paper could efficiently establish a practical 3D geometrical model of the grid-stiffened structure, effectively improve its load-carrying efficiency, and maintain the given parameters within the specified ranges. This approach could lay a foundation for reducing the launch costs and improving the transportation efficiency.