Apuntes Redes Neuronales

Coastal Engineering 54 (2007) 586 – 593 www.elsevier.com/locate/coastaleng

Neural network modelling of wave overtopping at coastal structures
Marcel R.A. van Gent a,⁎, Henk F.P. van den Boogaard a , Beatriz Pozueta a , Josep R. Medina b
b a Delft Hydraulics, The Netherlands Universidad Politécnica de Valencia, Spain

Received 10 April 2006; received in revised form 15 November 2006;accepted 5 December 2006 Available online 1 February 2007

Abstract A method has been developed to estimate wave overtopping discharges for a wide range of coastal structures. The prediction method is based on Neural Network modelling. For this purpose use is made of a data set obtained from a large number of physical model tests (collected within the framework of the European project CLASH, seee.g. [Steendam, G.J., Van der Meer, J.W., Verhaeghe, H., Besley, P., Franco, L. and Van Gent, M.R.A. (2004). The international database on wave overtopping. World Scientific, Proc. 29th ICCE, vol. 4, pp. 4301–4313, Lisbon, Portugal.]). Moreover, a method was developed to obtain confidence intervals for the overtopping predictions of the neural network. © 2006 Elsevier B.V. All rights reserved.Keywords: Wave overtopping; Breakwaters; Dikes; Coastal structures; Rubble mound structures; Neural network modelling; Physical model tests

1. Introduction For the design, safety assessment and rehabilitation of coastal structures reliable predictions of wave overtopping are required. Several design formulae exist for simplified types of dikes, rubble-mound breakwaters and verticalbreakwaters. Nevertheless, often no suitable prediction methods are available for structures with non-standard shapes. This paper describes a method that leads to a conceptualdesign tool to estimate wave overtopping discharges for a wide range of coastal structures. Only one schematisation is used for all types of coastal structures, where not only dikes, rubblemound breakwaters or vertical breakwaters aredefined, but also other non-standard structures are included. Additionally, not only is the effect of the most common parameters (i.e. wave height, wave period and crest freeboard) analysed herein, but also the effects of many other wave and structural characteristics are considered. The prediction method described is based on Neural Network modelling. Neural network modelling is discussed inSection 3 of this paper. For the preparation of the

neural network a data set is used that was obtained from about 10,000 physical model tests. The present investigation focuses on the development of a neural network for estimating mean overtopping discharges. Moreover, a method is developed to obtain confidence intervals for the overtopping predictions of the neural network. The latter is anessential extension since the neural network model results in a tool that acts for users as a kind of black box. Therefore, it is important that predictions are extended with information regarding their reliability or uncertainty. 2. Description of database and parameters involved 2.1. Description of the database The data set used for the set up of the present neural network (hereafter: NN), isthe database created within the framework of the European project CLASH. This database includes tests collected from several laboratories. This database is described in detail in Verhaeghe et al. (2003) and Steendam et al. (2004). Results from about 10,000 overtopping tests are included in the database. Each of these tests is described by a number of parameters that represent hydraulicinformation (i.e. incident wave characteristics and measured overtopping discharges) as well as structural information (i.e. parameters characterising the

⁎ Corresponding author. Delft Hydraulics, P.O. Box 177, 2600 MH Delft, The Netherlands. Fax: +31 15 2858712. E-mail address: marcel.vangent@wldelft.nl (M.R.A. van Gent). 0378-3839/$ - see front matter © 2006 Elsevier B.V. All rights reserved....