After a severe earthquake, a practical disaster relief strategy requires accurate information related to the damage level of buildings [1][2]. To properly carry out such a strategy while maintaining urban resilience, it is necessary to determine damage levels of affected buildings as soon as possible [3].According to Japan Building Disaster Prevention Association [4], the most common method for damage evaluation is conducted through visual inspection. This method provides an accurate evaluation of the damage level of a building. However, an accurate visual-based assessment requires highly skilled engineers and is time-consuming. [5][6]. Additionally, in cases of severe damage, some buildings might become inaccessible, reducing the usefulness of the visual method. To overcome this, automatic processes for damage evaluation have been researched [7].A structural health monitoring (SHM) system is a viable alternative to visual inspections because the former can automatically and immediately record the structural response using output measurements such as floor acceleration [8][9]. Although it is not difficult to estimate structural response, the way to evaluate damage using said response remains unclear [10].Hence, the evaluation of the performance curve and the classification of the current damage level requires a detailed investigation. and proposed the implementation of the scaling and characteristic point method for updating the performance curve based on analytical models and observed data (introduced in Sec. 2), and a damage classification method based on the identification of characteristic points corresponding to the following limit states: cracking, yielding, flexural mechanism, and ultimate conditions. Nevertheless, these methods were applied for one case only, and therefore, a range of applications and its effect on the accuracy of the method must be verified.The limitations of the described procedure include the following: 1) For typical buildings, the response to low-intensity earthquakes is available but it is less common to obtain observations after yielding, and 2) differences between analytical model and observation tend to decrease the accuracy of the updating procedure and damage evaluation method.