برآورد نيروهاي ناشي از امواج بر روي سازه هاي دريايي با استفاده از تلفيق تئوري پراش و شبكه هاي عصبي مصنوعي

چكيده انگليسي :

80 Estimation of wave load on offshore structures using a combined diffraction theory and neural networks Mohammad Javad Khayyami mj khayyami@me iut ac ir 16th June 2010 Department of Mechanical Engineering Isfahan University of Technology Isfahan 84156 83111 Iran Degree Master of Science Language Farsi Supervisor Ahmad Reza Azimian azimian@cc iut ac ir Abstract It is well known that floating structures such as ocean platforms breakwaters and wave energy devices are often used in ocean engineering The most important hydrodynamic properties of which are the hydrodynamic coefficients wave excitation forces and transmission and reflection coefficients are of major interest of designers and many research have been carried out These activities have been done based on experimental methods theoretical methods and numerical methods To analyze the hydrodynamic properties of floating structures various methods such as the Boundary Element Method BEM the Finite Element Method FEM and some analytical methods can be used In this work we use Boundary Element Method named diffraction theory to analyze simple offshore structures In this theory assuming potential flow and distribution of individual points on the body surface to calculate force and moments coefficients Using a special form of green s function that can issue boundary conditions such as the free surface sea bed and the radiation will include an important part of work When the structural dimensions compared with wavelength are large this theory can be used To estimate hydrodynamic coefficients we have to calculate Green function and its derivative s Green function is a very complex function and its calculation is very time consuming On the other hand one of the ways that the ability to calculate complex functions in a very short time is use artificial neural networks ANNs Artificial neural networks with input and output processing are able to establish a relationship between them In cases where the possibility of establishing a relationship between some physical parameter is not present can be very useful We utilize this ability to estimate Green function and its derivations The ANNs are trained to learn the relationships by experience so we must have a data set to learn ANN At the first step we calculate Green function and its derivative s by classical proposed Green function then we learn ANN with this data set In the last step we use the ANN which is learned to approximate green function In this thesis different neural networks were trained and then used these networks to calculation of force coefficients on the vertical cylinder floating disc and floating Buoy were We use two types of neural network namely backpropagation networks and radial basis networks This network has six inputs body dimension water depth etc and six output six parameters to calculate green function Force and moment coefficients on the geometry calculated by using neural network and without using it The results shows that if these networks properly trained can be calculate hydrodynamic coefficients with high accuracy CPU time comparison shows that elapsed time clearly decreases when we use ANN Comparison between radial basis and perceptron networks shows that radial basis networks is more accurately then perceptron networks but its elapsed time is also higher The advantage of this Technique is that green function and its derivative s is calculated just once but in classical method for any geometry it is calculated separately Keywords Diffraction Theory Artificial Neural Network Wave Load Green s Function Offshore Structure

خيامي ، محمد جواد

برآورد نيروهاي ناشي از امواج بر روي سازه هاي دريايي با استفاده از تلفيق تئوري پراش و شبكه هاي عصبي مصنوعي

نيروي امواج , شبكه عصبي مصنوعي , تابع گرين