Immediately after a seismic crisis, engineers are demanded to provide rapid post-earthquake damage evaluations to globally assess economic losses, residual overall serviceability, and the general safety conditions of the hit existing heritage. Since 1997 and subsequent amendments, in Italy, the inspectors must compile the AeDES forms. Composed of nine sections, the form-filling provides about sixty categorical features to aid the inspector to elaborate a categorical judgment of the rapid qualitative safeness of the structure. This judgment is categorized into six damage classes denoted with capital letters from A to F designating, respectively, fully usable buildings until condemned constructions. Despite the AeDES forms’ features should help the inspector to elaborate a quite fair judgment, there is still a certain level of subjectiveness, since there is not any strictly underlining model to objectively convey these sixty categorial features unequivocally toward a specific judgment. Therefore, in the current study, the authors analyzed the AeDES forms data coming from 878 public school buildings, thus exploring the possibility to aid the inspector with a multi-layer perceptron (MLP) neural network model. Precisely, the AeDES forms under investigation are referred to the L’Aquila city seismic event which hit the Abruzzi region of central Italy in 2009. Considering the sensitivity of the current multinomial classification performances in the presence of limited data regarding the MLP architecture topology, the authors formalized a multi-objective optimization problem to find an optimal architecture with the minimum number of hidden neurons and simultaneously maximize the classification accuracy.