Abstract: Due to the difficulties of the traditional reliability design in meeting modern design requirements, it was necessary to have a robustness optimization design to improve comprehensive reliability of the cabin. The reliability of composite monolayer was solved on the basis of two failure modes of matrix damage and fiber breakage by adopting the design point method. The idea of taking the structure as parallel systems that were constituted by series subsystems was put forward based on the hypothesis of the last layer failure. The main failure sequences can be figured out through combining the material stiffness ratio degradation criterion with monolayer reliability theory, and adopting the probability gradually failure analysis method, then the failure probability of structure can be worked out. Due to the complexity of design variable space of composite cabin, the idea of two-level optimization method was put forward. The first level was system-level layout optimization, which optimized some parameters, including profile shape of stiffener rib, the parameters of position, etc. The second level was subsystem-level size optimization, which optimized some parameters, including sectional dimensions of stiffener rib, each layer's thickness of composite, etc. Regarding the minimum mass of composite cabin as the objective function and the requirements of reliability as the constraint condition, the composite cabin with initial imperfection can be optimized by means of employing the genetic algorithm of self-adaption random search and coordinative optimization method of robustness. The method of optimization can provide references for optimization and design of composite structure.