www.ijecs.in International Journal Of Engineering And Computer Science ISSN:2319-7242 Volume 6 Issue 6 June 2017, Page No. 21845-21848 Index Copernicus value (2015): 58.10 DOI: 10.18535/ijecs/v6i6.49 Arunav Chakraborty, IJECS Volume 6 Issue 6 June, 2017 Page No. 21845-21848 Page 21845 Slope Stability Prediction using Artificial Neural Network (ANN) Arunav Chakraborty 1 and Dr. Diganta Goswami 2 1 Civil Engineering Department, Tezpur University, Tezpur, Assam, India aru243@gmail.com 2 Civil Engineering Department, Assam Engineering College, Guwahati, Assam, India Abstract: Artificial neural networks (ANN) usually called neural networks are very sophisticated modeling techniques which are capable of modeling extremely complex functions. They are used for predicting the outcome of two or more independent variables. Predicting the stability of slopes is a very challenging task for the geotechnical engineers. They have to pay particular attention to geology, ground water and shear strength of the soils in accessing slope stability. In this paper, a prediction formula has been developed for predicting the factor of safety (FOS) of the slopes using ANN. A total of 110 cases with different geometric and soil conditions were analyzed using Bishop’s Simplified Method. Out of these, 100 cases were used to train up the prediction model. The computational method for the training process was a back propagation learning algorithm. The prediction model is validated by comparing the results with the remaining 10 cases. Keywords: Artificial Neural Network, Back-propagation, Factor of Safety, Shear Strength, Slope Stability. 1. Introduction The introduction of artificial neural network (ANN) continues to captivate scientists and engineers from a variety of disciplines. This growing interest among the researchers is stemming from the fact that these learning machines show excellent performance in understanding the patterns and developing the non-linear relationships of multivariate dynamic systems. In this paper, an investigation has done to validate the utilization of ANN in the physical problem of slope stability prediction. The accurate estimation of the stability of rock or soil slope is a very challenging task for the geotechnical engineers. This is mainly due to the complexity of the physical system itself and the difficulty in determining the geotechnical input data parameters. The analysis must be carried out by considering the site subsurface conditions, ground behavior, and applied loads. It is due to its practical importance that slope stability analysis has drawn the attention of many investigators. The judgments regarding the risk factor and the safety factor must be made to evaluate the results of analyses. Therefore, slope investigation and classification are important for the community [1], [2], [3], [4]. Although the slope stability prediction is a very challenging task yet it has developed its existence to a great extent in the last two decades. Many researchers from the geotechnical background are constantly working to find new prediction models for determining the slope stability. Sakellariou and Ferentinou used ANN to predict the stability of slopes for circular failure and wedge failure mechanism and came up with the conclusion that the input parameters are having close relationships with the output parameters [5]. Kayesa predicted the slope failure of Letlhakane mine using Geomos slope monitoring system which contributed a lot in avoiding potentially fatal injury and damage to mining equipments [6]. Davis and Keller studied on uncertainty behaviour of soil and developed a slope stability prediction model based on fuzzy sets and Monte Carlo simulation [7]. The use of evolutionary polynomial regression (EPR) technique for predicting the stability of soil and rock by Ahangar-Asr et al. is found to be very effective and robust in slope behaviour modeling [8]. Mohamed et al. used the concept of fuzzy logic system for prediction and found that the results are having higher degree of accuracy [9]. Erzin and Cetin developed another prediction model uasing ANN and multiple regression (MR) for estimating the FOS of an artificial slope subjected to earthquake forces [10]. The obtained indices make it clear that ANN model has higher prediction performance than the MR model. Sternik made a comparison study of two slope stability prediction models developed by shear strength reduction (SSR) method and gravity increase (GI) method with Bishop’s Simplified Method [11]. The comparison results culminated that SSR method gives more close results compared to GI method. 2. Methodology In this research, 110 slope cases having different geometrical and slope parameters were selected along Guwahati-Shillong Highway (NH-40), India. Soil samples were collected and laboratory tests were performed to find out the various soil parameters. These slope parameters were used to analyze the various slopes using Bishop’s Simplified Method to find the FOS. Out of these, 100 cases were used to develop the prediction model using ANN. In the proposed model, several important parameters including, height of the slope (H), cohesion (C), angle of internal friction (φ), angle of the slope (β) and unit weight of soil (γ) were used as input parameters whereas the FOS was used as the target value. The ANN model was prepared in Matlab 2011a (Figure. 1).