Abstract

The swell index which is the slope of the rebound curve of void ratio versus the logarithm of the effective pressure curve is used to estimate the consolidation settlement of overconsolidated fine grained soils. Because determination of swell index from oedometer tests takes a relatively long time, empirical equations involving index soil properties, are needed to estimate it for preliminary calculations and to control the validity of consolidation tests. Geotechnical engineering literature involves empirical equations for the estimation of compression and swell indexes. In this study the performance of widely used empirical equations were assessed using a database consisting of 42 test data. In addition to this, new empirical relationships with single and multiple dependent variables were developed with better estimation capability. An artificial neural network (ANN) which has two input variables, one hidden layer and eight hidden layer nodes was also developed to estimate swell index. It was concluded that the performance of the ANN is better than empirical equations.

Key words: Swell index, clay soils, regression analyses, artificial neural networks.