Soft computing based prediction of support pressure in tunnels

Abstract

Prediction of Tunnel support pressure up to an accurate and reliable degree is difficult, but of utmost importance. Empirical models are available with different set of parameters, mostly are based on the rock classification parameters. A feed forward neural network based predictive models from the data collected from literature for the Himalayan tunnels have been developed. The input variables in the developed neural network models were depth of over burden, radius of tunnel, normalised closure. The fourth input variable was rock mass quality or rock mass number or rock mass rating. The output was a support pressure. Sensitivity analysis relating the variables affecting the support pressure has been performed. The developed neural network models were compared with models developed based on the multiple linear regression analysis as well as with empirical models already available in literature. Finally, model equations have been presented based on the connection weight.