A Comparative Study on the Performance of ANN, MLR and MNR in the Assessment of Slope Stability for Kalla - Coonoor Hill Road Stretch of Nilgiris

Abstract

Artificial neural network (ANN), Multiple linear regression (MLR) and Multiple Non linear regression (MNR) analysis are known as Machine learning techniques which solves complex problems for which doesn’t have any standard algorithms. These methods forms their own rules of learning and solve the complex problems. Among the various problems affecting the hill system, slope instability’s are very important. These slope failures have been observed more frequently in recent days because of the tremendous developmental activities in hilly regions. Road networks play major role in the developmental activities of hill regions. Human interactions with the nature during the construction of roads, increases the landslide susceptibility. In present study, a road network between 76.806–76.874 E longitude and 11.334–11.351 N latitude passing through Nilgiri hills was considered for stability assessment. Assessment of slope-stability/landslide-susceptibility of the hills is a common geotechnical problem and it requires detailed information about several parameters such as slope inclination, height, shear strength, density, etc. Even though limit equilibrium methods (LEM) are most accurate methods for the estimation of slope stability, their application to analyze huge number of slopes especially for hilly regions are very tedious and time consuming. In order to minimize these problems associated with LEM, ANN, MLR and MNR are soft computing method which can solve these problems. There is often confusion in the selection of suitable method from these methods, for this all these methods are applied to Kalla - Coonoor Hill road stretch and then results are analyzed and find suitable method out these. A part from these their advantages and disadvantages of these alternative approaches are also discussed in this study.