Name: AMANDIO GONÇALVES DE OLIVEIRA FILHO
Type: MSc dissertation
Publication date: 01/11/2019
Advisor:
Name |
Role |
|---|---|
| KÁTIA VANESSA BICALHO | Advisor * |
Examining board:
| Name |
Role |
|---|---|
| KÁTIA VANESSA BICALHO | Advisor * |
| WILIAN HIROSHI HISATUGU | Co advisor * |
Summary: The compression index (CC) and the compression ratio (CR) are used to calculate consolidation settlement of foundations on soft soils when requested by external loads. Several empirical correlations have been published in the literature to predict CC and CR values for different soft soils as a function of their index properties. The multiplicity of published correlations to estimate CC and CR indicates the need for selection criteria in their use. This research investigates a database of 2,022 soft soil samples from different geological sites in Brazil and other countries. The objective of this study is to evaluate the prediction capacity of the compression index and the compression rate through empirical correlations previously published in the literature and by empirical correlations of adjustments proposed in this research, compared to the use of Artificial Neural Networks (ANNs). Artificial neural networks was trained with the Levenberg-Marquardt algorithm, with one or two hidden layers operating sigmoid activation functions and an output layer activated by a linear function. The ANNs are trained in different groups, first with all soil samples from the dataset, and then only with soil samples from the Brazilian coast, in order to evaluate the generalization capacity of ANNs. Forecasting performance have been assessed using statistical techniques that include: (i) the root mean square error (RMSE), (ii) the estimated and measured compression index ratio (K), (iii) the ranking index (RI) and (iv) the ranking distance (RD). The presented results reveal that the adapted ANNs created for estimation of soft soils Cc and CR from Brazilian coast and other countries have potential application as an alternative to the empirical correlations, especially during preliminary investigation of suitability of a foundation site during planning stages.
