Application of Support Vector Machine Technique for Damage Level Prediction of Tandem Breakwater

For decades, breakwaters played a vital role in the development of the port and which in turn assist in improving the economy of the country. Due to the development of industries and urbanization the coastal region is facing threat towards its normal and safe functioning. It challenges for the coastal engineers protect the environment and also support the urbanization in a protective way. Therefore, a need of protective structure along the coast arises. The tandem breakwater is one of the innovative types of structure, which consists of a conventional breakwater along with the submerged reef. It has been proved that submerged reef acts as a protective structure, when placed in-front of the conventional breakwater. The final geometry and the layout of tandem breakwater have been achieved through physical model studies. Support Vector Machine (SVM) which accounts for structural risk minimization compared to neural networks is used to model various problems of real time scenarios where mathematical modeling is difficult. By assigning appropriate weights, biases and e-insensitive loss function of the problem, SVM proves to be robust in addressing non-linear and non-stationary problems. In the present study SVM is applied to predict the damage level of the conventional breakwater of tandem breakwater. The experimental data available in the department of Applied Mechanics & Hydraulics are used for the analysis. Finally, comparison of the predicted damage level is made with the observed data of the experimental work using the statistical measures such as Root Mean Square Error (RMSE), Correlation Coefficient (CC), Scatter Index (SI) and Nash Sutcliffe Efficiency (NSE). It is observed that SVM technique using Radial basis kernel function (RBF) performed better with 0.9357 CC, 0.0765 RMSE, 0.5293 SI and 0.8327 NSE.