In the last few years, artificial neural networks (ANNs) have advanced rapidly. Fundamentally, these networks are developed by using data to recognize patterns through the use of complex mathematics; the resulting tools can then be used for mathematical modeling. Numerous publications have appeared in the past 20 years, dealing with ANNs that correlate information such as both temporal and spatial variations in wave heights and water levels, as well as data pertaining to wave overtopping, pipeline and pier scour, coastal erosion and other coastal phenomena. These developments are reviewed here after first providing an introduction to ANNs and a brief primer for developing them. The literature reviewed suggests that ANNs may be best used as a supplementary tool to enhance the efficiency and reliability of physics-based forecasting tools, as a spatial correlation method in some cases to deploy more efficiently data-gathering devices, and as a means to create more sophisticated curve fits for processes such as scour and overtopping. While other applications may also be found, the penetration of ANNs into engineering practice still requires much effort. It is noted that this may be accomplished through judicious selection of future research problems as well as more detailed investigation of ANN characteristics, keeping in mind the capabilities of alternative technologies (e.g. physics-based models).
ASJC Scopus subject areas
- Earth and Planetary Sciences(all)