Name: MORGANA MORESCHI
Type: MSc dissertation
Publication date: 17/02/2023
Advisor:
Name |
Role |
|---|---|
| KÁTIA VANESSA BICALHO | Advisor * |
Examining board:
| Name |
Role |
|---|---|
| KÁTIA VANESSA BICALHO | Advisor * |
| WILIAN HIROSHI HISATUGU | Co advisor * |
Summary: This research analyzes and compares the performance of artificial neural networks
(ANN), statistical methods and empirical and semi-empirical correlations in predicting
values of the permeability coefficient of saturated soils (ksat), based on index
properties that characterize the granulometric distribution and the fine fraction. A
number of 8258 experimental data of ksat from soils composed of coarse and fine
grains (2.50x10-13 ksat (m/s)  4.50x10-2), published in 08 databases in the literature,
were compiled and compared, to understand the hydraulic properties of saturated
soils and to descript ksat prediction problems. Subsequently, a set of samples was
selected and analyzed with a combination of different input variables to predict the
log(ksat), using linear and multiple polynomial regression and ANN. The input variables
considered were percentage of fines (silt and clay) (%Finos), liquidity limit (LL),
effective diameter (d10), uniformity coefficient (Cu) and voids index (e). The results
were evaluated from the values of the coefficient of determination, the root mean
squared error and the mean absolute error. The performance of the ANNs surpassed
the regressions and correlations in the literature. Of all the results of the analyzes
carried out, RNA302, which considered as independent variables the %Finos, LL, Cu
and d10, numerically presented the best results. The addition of a third hidden layer
reduced the accuracy of the networks. The regressions and the ANN were better than
the empirical correlations for ksat prediction, for the investigated database, and
showed that the choice of variables that characterize the granulometri c distribution
and the fine fraction was satisfactory for the experimental database. Considering that
ksat is a highly variable property and a function of several interdependent properties,
the ANN technique proved to be viable, mainly because it does not require prior
knowledge of the mathematical relationship between the variables and because of its
ability to describe more complex problems. The importance of including information
on the granulometry and nature of the fines in ksat databases is highlighted, mainly for
characterizing the permeability of samples with hydraulic properties dominated by the
fine fraction.
