Soil-structure interaction: A neural-fuzzy model to deal with uncertainties in footing settlement prediction

摘要

Engineers have always been aware of uncertainties in soil - structure interaction problems, related to soil inherent variability, site conditions, construction tolerance, failure mechanisms, in particular when the structure is founded in granular soils. Geotechnical and structure design deal with these uncertainty by heuristic and expertise knowledge using input data that fall in the category of non statistical uncertainties. This paper is devoted to the description of a Neuro-Fuzzy method to predict foundation settlements in granular soils and their effects on the structure response. The learning process analyses over 200 record of settlement of building foundations, tanks and embankments on sand and gravel. Uncertainties and vagueness into parameters of in situ tests, boundary conditions and external loads have been introduced as fuzzy numbers to be later reflected in the soil-foundation-superstructure system response.