www.tjprc.org editor@tjprc.org International Journal of Civil, Structural, Environmental and Infrastructure Engineering Research and Development (IJCSEIERD) ISSN(P): 2249-6866; ISSN(E): 2249-7978 Vol. 4, Issue 3, Jun 2014, 1-10 © TJPRC Pvt. Ltd. ESTIMATION OF SHEAR STRENGTH PARAMETERS OF SOILS USING ANN TECHNIQUE ABEER ABDULJABBAR ABDULABBAS & Y. K. BIND Department of Civil Engineering (Geotechnical Engineering), Sam Higginbottom Institute of Agriculture Technology & Sciences, Deemed University, Allahabad, Uttar Pradesh, India Republic of Iraq, Babylon University, Hillah, Iraq ABSTRACT The present study emphasizes on development of artificial neural network (ANN) models for prediction shear strength parameters like cohesion and angle of shearing resistance of the soil. Pertinent input parameters like depth, SPT- N value, liquid limit and plastic limit have been chosen for estimating two pertinent output parameters of shear strength. A total data sets of soil geotechnical investigation area were collected, out of which 72 datasets were from site A, Cohesion-less soils were not included for prediction of shear strength parameters. ANN models for each site were developed. First two models consist cohesion and angle of internal friction as output separately whereas third model consist both cohesion and angle of internal friction as output parameters. In the ANN analyses, the data set is generally normalized to obtain better convergence, prior to the training stage, a certain range in which the inputs and targets values fall is determined. The normalized dataset was then used to train neural networks. At the end of analysis, the network outputs were post processed to convert the data back into non- normalized units. Some precautionary measures adopted during the development of the models were the separation of cohesion-less soil, selection of at least two or more than two pertinent input parameters and training each network several times for different iteration to avoid anomaly in result. The results of ANN models were examined on the basis of mean square error and regression. Some of the pertinent input parameters were introduced in all networks to compare the result of actual and predicted output by the networks. Network error, percentage error, root mean square error and mean absolute error between calculated and predicted value for all networks were calculated. This indicate that ANN is a strong tool for predicting shear strength parameters of soil. The result of models showed the feasibility of ANN in geotechnical engineering. KEYWORDS: Artificial Neural Network (ANN), Mean Square Error (MSE) and Regression, Root Mean Square Error (RMSE) INTRODUCTION One of the most important engineering properties of soil is its shear strength or ability to resist sliding along internal surfaces within a mass. The stability of cut, the slope of an earth dam, the foundation of structures, the natural slopes of hillsides and other structures built on soil depend upon the shearing resistance offered by the soil along the probable surface of slippage. There is hardly a problem in engineering which does not involve the shear properties of soil in some manner or other. The basic concept of friction is applicable to the soils which are purely granular in character. Soils, which are not purely granular, exhibit an additional strength which is due to cohesion between the particles.