Abstract: Based on the form vulnerability theory of skeletal structures, a well-formation optimal model of single-layer spherical shells is established by taking the minimization of the standard deviation of the well-formation as the optimization objective and the cross-sectional dimensions of members as the optimization variables. This optimal model takes into account the constraints of a slender ratio, a deflection, bar strength and stability. Taking genetic algorithm and simulated annealing as sub algorithms, a genetic-simulated annealing algorithm (GASA) is put forward using a mixed strategy. Meanwhile, adaptive strategies are adopted to reduce the dependency of the algorithm upon the optimization parameters. Taking an example of a 70m span single-layer spherical shell, the uneven well-formation regions are identified through a clustering process. The cross-sectional dimensions optimization of members located in those regions is carried out using GASA. By the clustering process analysis and the seismic time history analysis, it is indicated that the optimized model and algorithm can effectively solve the collapse scenario optimization of large-scale single-layer spherical shells with several optimization variables under earthquake excitations.