Artificial Intelligent Systems and Machine Learning

Gumageri Nagaraj ¹, G. S. Dwarakish ², Sreenivasulu Dandagala ², Usha Natesan ³

Affiliations
1 Department of Applied Mechanics and Hydraulics, National Institute of Technology, Karnataka, Surathkal, 575025, IN
2 Department of Applied Mechanics and Hydraulics, National Institute of Technology Karnataka, Surathkal, 575025, IN
3 Center for Water Resources, Anna University, Chennai-600025, IN

Abstract

Forecasting of wave parameters is necessary for many
marine and coastal operational related activities. In this paper, artificial neural network (ANN) as a robust data learning method is used to forecast the waveheight for the next 3hr, 6hr, 9hr, 12hr, 24hr, 48hr, 72hr, 96hr and 120hr in the Mangalore region, southwest coast of India. For this purpose two different models namely, Feed Forward Back Propagation (FFBP) and Nonlinear Auto Regressive Model with eXogenous input (NARX) of the ANN were used. The performances of developed models were evaluated using performance indices such as RMSE and CE. The CE values in FFBP model ranged from 0.997 to 0.785 while in NRAX model CE values are between 0.995 and 0.806 for the prediction time from 3hr to 120hr. A better agreement was observed between the observed and predicted waves for NRAX than that of FFBP for smaller (3-12hr) and larger lead period (24-120hr). Thus the NARX model performs better than the FFBP in terms of prediction capability and accuracy. 

Keywords

Waveheight, Prediction, ANN, FFBP, NRAX, RMSE.

Full Text

     

Sreenivasulu Dandagala ¹, M. V. Subba Reddy ¹, D. Srinivasa Murthy ¹, Gumageri Nagaraj ²

Affiliations
1 Department of Civil Engineering, Sree Vidyanikethan Engineering College, Sri Sainath Nagar, Rangampet-517102, IN
2 Department of Civil Engineering, Bearys Institute of Technology Mangalore, Innoli, Boliyar Village, Near Mangalore University, Mangalore, IN

Abstract

Reliable groundwater level forecasting is crucial and has become challenging task for the groundwater hydrologists. An accurate forecast plays a vital role for proper planning and utilization of groundwater resources in a sustainable manner. In the present study an attempt has been made to study one of the soft computing techniques such as Artificial Neural Networks (ANN) and its various applications in groundwater hydrology. Accurate prediction of Groundwater Level (GWL), assessment of water quality, concentrations of contamination, estimation of various aquifer properties and dynamic groundwater levels are the complex problems in the domain of groundwater hydrology. The ANN applications are studied in detailed manner and discussed its merits and demerits and given a brief discussion on its scope for future research work in the groundwater hydrology. Moreover, type of time series data, quality of data and other aspects would limit the applications of the modeling and requires further any model development. In the time series analysis, ANN has wide applications in the domain of civil engineering due to its capability of non-linear modelling in real world complex phenomena. ANN is non parametric method and prior knowledge is not mandatory. Above all these features makes ANN more attractive for time series modelling and forecasting. However, in the past a lot of successful applications have shown that ANN provide powerful tool for time series modeling. Therefore, based on the literature cited on ANN applications in the domain of groundwater hydrology, ANN can be suggested to be one of the effective tool for better GWL forecasting even with limited data.

Keywords

Groundwater Level, Artificial Neural Networks, Time Series.

Full Text