Neural Network Modelling of Forces on Vertical Structures (original) (raw)
Related papers
Applications of a neural network to predict wave overtopping at Coastal Structures
2006
Introduction For design, safety assessment and rehabilitation of coastal structures reliable predictions of wave overtopping are required. Several design formulae exist for dikes, rubble mound breakwaters and vertical breakwaters. Nevertheless, often no suitable prediction methods are available for structures that do not resemble rather standard shapes. In the European research project CLASH a method is developed to provide a generic design tool to estimate wave overtopping discharges for a very wide range of coastal structures. The paper gives results from the CLASH project on this subject. It is focused on the extensive database gathered (see Verhaeghe et al., 2003), the neural network (see Pozueta et al., 2004) that has been developed on the basis of this database, and on applications of both.
Prediction of wave impact occurrence on vertical and composite breakwaters
Research studies and case-histories of the collapse of vertical breakwaters have revealed the destructive potential of impact loads due to breaking waves and confirmed that during severe storms a large number of wave impacts can be generated of such severity as to cause the failure of the breakwater-foundation system. The problem of addressing the probability of occurrence of impact loading from breaking waves is very complicated as it involves different aspects, closely interconnected, of hydrodynamic and morphological nature. Moreover, the breaking of waves in presence of vertical breakwaters is likely to be different in character from that most usually studied which relate to the case where waves are breaking due to shoaling water depths or to a driving wind, and till now there has been very little study of wave breaking in the presence of reflection. This gave rise to studies, most of them experimental, aimed at improving the reliability of the design procedures of vertical breakwaters. The analyses of hydraulic model tests have stressed the importance of particular combinations of wave conditions, bottom slope, berm and vertical wall profile on the generation of impact loads. In this paper recent research results on the prediction of the occurrence of wave impacts on vertical structures are described and general design formulae and graphs, based on extensive bidimensional and tridimensional hydraulic model tests with random waves, are proposed. standardisation, environmental implications, construction and maintenance costs, when compared with the traditional rubble -mound type. Although these structures have been used for many decades, some of them have suffered damage from storms that can be considered catastrophic as the major failures of vertical breakwaters cost 2 -3 times more to rebuild than the original construction costs.
In the present study, Artificial Neural Networks (ANNs) with different topologies have been evaluated to be used to predict hydrodynamic coefficients of permeable paneled breakwater. Two neural network models are constructed, one to predict wave transmission coefficient (K t) and another for the prediction of wave reflection coefficient (K r). Back propagation algorithm was used to train a multi-layer feed-forward network (Levenberg Marquardt algorithm). The capability of ANN topologies to estimate these coefficients is evaluated using the Mean Squared Error (MSE). Based on training patterns of different ANNs, a 5-7-1 topology has been selected to predict both coefficients. The results of the developed ANN models proved that this technique is reliable in such field. A good match between the measured and predicted values was observed with correlation values varying in the range (0.9508-0.9805) for the training set and (0.9159-0.9877) for the testing set.
Prediction of geometrical properties of perfect breaking waves on composite breakwaters
Applied Ocean Research, 2011
Breaking wave loads on coastal structures depend primarily on the type of wave breaking at the instant of impact. When a wave breaks on a vertical wall with an almost vertical front face called the ''perfect breaking'', the greatest impact forces are produced. The correct prediction of impact forces from perfect breaking of waves on seawalls and breakwaters is closely dependent on the accurate determination of their configurations at breaking. The present study is concerned with the determination of the geometrical properties of perfect breaking waves on composite-type breakwaters by employing artificial neural networks. Using a set of laboratory data, the breaker crest height, h b , breaker height, H b , and water depth in front of the wall, d w , from perfect breaking of waves on composite breakwaters are predicted using the artificial neural network technique and the results are compared with those obtained from linear and multi-linear regression models. The comparisons of the predicted results from the present models with measured data show that the h b , H b and d w values, which represent the geometry of waves breaking directly on composite breakwaters, can be predicted more accurately by artificial neural networks compared to linear and multi-linear regressions.
Neural network modelling of wave overtopping at coastal structures
Coastal Engineering, 2007
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, see e.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, Lisbon, Portugal.]). Moreover, a method was developed to obtain confidence intervals for the overtopping predictions of the neural network.
Comparison of Neural Network Error Measures for Simulation of Slender Marine Structures
Journal of Applied Mathematics, 2014
Training of an artificial neural network (ANN) adjusts the internal weights of the network in order to minimize a predefined error measure. This error measure is given by an error function. Several different error functions are suggested in the literature. However, the far most common measure for regression is the mean square error. This paper looks into the possibility of improving the performance of neural networks by selecting or defining error functions that are tailor-made for a specific objective. A neural network trained to simulate tension forces in an anchor chain on a floating offshore platform is designed and tested. The purpose of setting up the network is to reduce calculation time in a fatigue life analysis. Therefore, the networks trained on different error functions are compared with respect to accuracy of rain flow counts of stress cycles over a number of time series simulations. It is shown that adjusting the error function to perform significantly better on a spec...
Water
Cutoff walls are widely used to limit seepage, piping, and the uplift under hydraulic structures. Therefore, this study focused on a numerical investigation of the hydraulic performance of cutoff walls beneath hydraulic structures during both static and dynamic conditions, considering location and inclination angle influences. The results confirmed that placing the cutoff wall at the upstream heel was more effective in reducing uplift pressure compared to other placements during static conditions. The inclination angles for the different placements of the cutoff wall had a significant impact on the total uplift pressure, exit hydraulic gradient, and seepage discharge during both static and dynamic states. The earthquakes had a noticeable effect on uplift pressure, seepage discharge, and exit hydraulic gradient. During static conditions, the inclination angle of 90° was the most effective angle for decreasing seepage discharge, irrespective of the cutoff wall position. During an eart...