The characteristic of expansiveness represented by swelling pressure is considered as major parameter in design of engineering structures. In order to determine the potential expansion, the swelling pressure experiments are usually conducted. In the present study, neuronets have been used to predict swelling pressure of undisturbed as well as remolded soils from given geotechnical parameters such as grain-size distribution, consistency limits, activity, and deferential free swell. Neuronet models relating the potential expansiveness to some geotechnical properties are derived in order to overcome the need to perform lengthy swelling pressure determination experiments. While the remolded soil neuronets (Testing sets) developed in this study might be considered as specific, the remolded soil neuronets (Training sets) can be used to predict swelling characteristics for most soils in India or elsewhere in the world. Similarly the combined datasets used in training & testing may significantly assist earthwork engineers in designing the sub grades and control the detrimental effect caused by volume changes associated with swelling of soil for roads, canals, buildings etc. Moreover, a neural network used in the prediction of expansion may assist the designers in selection of the appropriate water content to compact clays.

Swelling Pressure Neuronets Input layer Input variables Hidden layer Output layer