Validation of nondestructive methods for assessing stone masonry using artificial neural networks

Abstract

The aim of this study is to deepen the technical and scientific knowledge related to the characterization of granite masonry based on geophysical tests, mechanical techniques, and neural networks. We used a nondestructive test method to characterize traditional stone masonry and further obtained data on the mechanical parameters of the elements. Historic buildings are typically constructed with stone masonry and make up the urban heritage. The maintenance and rehabilitation of historic buildings are crucial for maintaining interest in history, owing to the historical and cultural values of such buildings. The preservation of buildings classified as historical and cultural heritage is of collective interest, as they mark the history of society. Because the research object is considered a traditional structure, the use of destructive test techniques is discouraged. Thus, a mechanical characterization simulation tool using artificial neural networks (ANNs) was developed and applied to traditional granite walls. This database was developed through ground-penetrating radar (GPR) and sonic tests to characterize wall samples built. The walls were analyzed in a controlled environment, and the elastic modulus was used in response to the ANNs. Two case studies representing traditional granite masonry buildings in Portugal were evaluated through nondestructive characterization tests. For the Mancelos church and Miguel Bombarda Street building, ANNs were applied based on the sonic and GPR test results. The feasibility of using ANN simulation tools for characterizing traditional and historic buildings constructed with granite stone masonry was demonstrated.