One of the useful measures of wave directional spreading at a given location is the directional spreading parameter. This paper presents a new approach to arrive at its characteristic value using the computational technique of Artificial Neural Networks (ANN). Different networks were developed in order to obtain the characteristic spreading parameter from the unidirectional parameters of significant wave height and average zero cross wave period, Ursell number, spectral width, spectral peakedness and maximum spectral density. An ANN was also developed to get the spreading parameter from the directional width. Training of the networks was trained using alternative algorithms. The data used belonged to a site off Goa along the west coast of India. The network output compared well with the target spreading parameter values. Use of alternative training was some times found necessary to get accurate output. The spreading parameter was more strongly correlated with characteristic wave height and period ― in either individual or grouped form ― than the unidirectional spectral characteristics.
展开▼
