Volume decomposition is a technique for decomposing a computer-aided design (CAD) model into subvolumes to improve the types of meshes that can be generated and enhance the accuracy of finite element analysis. Protrusions frequently occur on thin-shell CAD models for functional and structural purposes. Automatic decomposition of such features is difficult due to the complexity and variation of shapes. In this study, a method was proposed for decomposing protrusions on thin-shell CAD models. A feature recognition algorithm was first employed to recognize four types of protrusions on a boundary representation (B-rep) model: tubes, columns, ribs, and symmetric extrusions. A specific volume decomposition algorithm was then developed for each type of protrusion. A protrusion is divided into sweepable subvolumes, with each subvolume represented by a pair of main contours and several side contours that connect to both main contours simultaneously. The contours of all subvolumes are tightly adjacent to each other to preserve the entire volume of the feature. Realistic CAD models and analysis results are presented to demonstrate the feasibility of the proposed protrusion decomposition method. The integration of the proposed algorithm with the decomposition of thin shells is also discussed.