Underwater welding is commonly used to repair corroded offshore steel structures. Corrosion-damaged portions are covered by welded patch plates. According to the current design manual, a thickness of patch plate and a weld length can be determined. However, different weld patterns can be designed to achieve the same required weld length. In order to examine the effectiveness of these different weld patterns, this paper first proposes a method to model underwater welds in the finite element analysis based on mechanical properties of fillet welds obtained from weld strength tests. The weld model was firstly validated against a theoretical shear stress distribution in a longitudinal fillet weld and then further validated against experimental results of thickness-reduced steel pipes repaired with welded patch plates under compression. The proposed model was then applied to thickness-reduced steel pipes repaired by welded patch plates with different weld patterns that have the minimum required weld length. Behaviors of these repaired pipes under a compressive load were examined with respect to stiffness, load-carrying capacity, load share of patch plates, and failure modes. It was found that stiffness and load-carrying capacity of the thickness-reduced steel pipes under compression cannot be fully recovered by the welded patch plate repair when a patch plate thickness is the same as the thickness reduction of the damaged pipe. Among different weld patters, the one with four slits was found to show better performance.