Structural Health Monitoring (SHM) system is a promising technology to determine the health condition of a structure and localize its damage. SHM system is widely used, especially in gigantic structures for its strong requirements of safety. Many researches about SHM system have been conducting. However, conventional SHM system did not take account of an accidental collapse caused by earthquake. As a result, there is a possibility that sensor nodes, network links, and a center server cause failures in processing or storing of data gathered in the building. These failures may lose important data that contains useful information for post-analyzing the collapse. The data includes when, where and why the structure is damaged and collapsed. Theses information have a great deal of potential for preventing similar damage or collapse and it will make great contribution to accelerate future structural researches. To solve the problem of data losses in a damage, data sharing system for SHM system is proposed. This system shares data within sensor nodes. The proposed system achieves the following three processes; node search, backup node selection and backup data transfer process. It is important to conduct these processes under the condition of limited resources due to small and not powerful sensor nodes. Backup node selection is an important process of this system. This paper mainly focused on this selection. Round-trip time, a size of free memory space of backup nodes and value of displacement caused by vibration are used for the selection. In this paper, the selection method based on displacement was checked. An experiment was conducted by using practical sensor nodes and acceleration data. From the result of this experiment, proposed system could share data by using selection method based on displacement.