Preliminary Study of a Neural Network Procedure for the Timely Detection of the Collapse of Historical Cultural Heritage Structures

The protection and preservation of buildings of historical and cultural heritage is often only guaranteed by a type of ordinary or extraordinary maintenance. The constructions are subject to restoration work when damage is advanced or partial collapses have occurred. In recent years, modern technology and research have ensured the development of continuous monitoring through structural health monitoring (SHM) systems. SHMs allow the real time acquisition of different physical quantities that are stored in a cloud. Often, these data are numerous and difficult to interpret through traditional analysis conducted by non-linear numerical simulations involving high computational burden and times. To this purpose, in this paper is presented the preliminary study of the implementation of a Neural Network that, trained on data obtained through a Virtual Optimal Sensor Placement method, is able to predict the collapse. This method presents encouraging results in classifying the acquired data and producing timely detection.