The scour around the bridge piers has been estimated using conventional empirical formulae; however, these formulas are unable to predict the scour depth precisely. The present study is conducted in two parts (a) experimental investigation for evaluating the behaviour of local scour around twin piers positioned in the transverse direction of flow and (b) an empirical equation to estimating scour depth is proposed utilising new evolutionary Artificial Intelligence technique Gene Expression Programming (GEP). Experimental results of the present study demonstrate the influence of the rate of flow and clear spacing between the piers on the scour depth. Additionally the results of the soft computing technique GEP during testing and training of proposed modelling, fitness function Root Mean Square Error is observed as 0.00133 and 0.00113, with the coefficient of determination as 0.950 and 0.955, respectively. Furthermore, in order to find out the role of each variable for scour depth sensitivity analysis has been conducted. The findings of the sensitivity analysis show that the pier spacing and rate of flow play the most significant role in scour depth estimation. Results of this study demonstrate good agreement with the proposed GEP model and conclude that it is a better approach for forecasting scour depth.