This paper describes a structural optimization method for subsystems which realizes the desired value of coupling loss factors (CLFs) in statistical energy analysis (SEA). We have developed the structural design process on the basis of experimental SEA for the purpose of reducing structure-borne sound in real-world machinery. The process identifies the CLFs which should be changed in order to reduce the noise radiated from the machinery. The optimization method is implemented using the finite element method and optimization algorithms. The finite element model represents a part of a whole system which includes a junction together with their neighboring SEA subsystems, associated with the CLFs which need to be changed. In this paper, the proposed method for the structural optimization is demonstrated. Consequently, taking one CLF as the objective function, an optimization of the thickness of the shell elements is performed showing the efficiency of the structural optimization method.