Large-span spatial structures have closely spaced frequencies, complex vibration modes and a huge number of components. The structural monitoring methods based on local responses have poor applicability in large-span spatial structures. Addressing the above problem, the AI-based global monitoring method and the corresponding optimal sensor placement (OSP) strategy are proposed in this work. The model updating technique based on the estimation of distribution algorithm (EDA) is adopted to realize OSP and structural damage identification (SDI) of large-span spatial structures. The objective functions in the OSP framework based on different criteria are proposed separately. The influence of the algorithm parameters and the target number of sensors are evaluated. In addition, two physical tests focusing on different SDI problems are conducted to demonstrate the effectiveness of the proposed OSP strategy. One test aims to identify the local damage of members in a single-layer cylindrical latticed shell based on frequency domain data, the other aims to identify the loosening bolts in a six-star model based on time domain data. The results indicate that the proposed OSP framework performs well in different SDI problems. The EDA-based model updating technique shows good potential in solving the OSP problem of large-span spatial structures.