Título: Previsão de Camadas do Subsolo do Sítio da Usina Nuclear de Angra-2 através de Redes Neurais
Autores: Dyminski, Andréa Sell; Ribeiro, Eduardo Parente; Romanel, Celso; Pedreira, Carlos Eduardo
Resumo: In civil engineering, it is of great importance to know the subsoil characteristics to design the foundation for large constructions. This is usually achieved by drilling holes for tests on the desired construction area. Those investigations enable one to know the type of soils as a function of depth at that position. Ideally, an increased number of holes gives better knowledge of the terrain, but to reduce the costs only a few holes are drilled and the information is empirically interpolated to the neighbor regions. We propose the use of feedforward networks to classify the type of soil over the whole construction area. Data from SPTs performed at Angra II nuclear power plant site were used to train and test the networks. This technique shows promising results for three dimensional spatial interpolation problems in geotechnical engineering.
Código DOI: 10.21528/CBRN2001-139
Artigo em pdf: 5cbrn_139.pdf
Arquivo BibTex: 5cbrn_139.bib