| The Artificial Neural Network (ANN) technique has been successfully applied in many industries. In the shipbuilding industry, the ANN technique has already been applied in the modeling of ship type and in some cases being used in real production as well. Currently, using ANN technique to fit the hull curve and fairing is still in theory. In real world, BP algorithm is used to fit the hull curve and it is used mainly to ditty rate and simple line style ships. The dissertation adopts the neural network of the SWF (Spline Weight Function) to carry on the curve fitting of more complicated hull lines. It makes the fairing of multi-set molded lines possible. In the research of fitting and fairing hull lines, we have done the following works:(1) Study the basic theories of BP neural network and neural network of the SWF. Analyze the two kinds of algorithms and theirs merit and shortcoming.(2) Use the two kind of neural network algorithms to carry on the curve fitting to differentiate molded lines of Three Gorges Reservoir ship. Compare the application of two kinds of algorithms in the actual curve fitting. The conclusion is that using the SWF neural network to fit is closer to actual engineering requirement.(3) A method of fitting curve is introduced and it solves two endpoint condition problems existed in the fitting hull lines. One is to deal with the endpoint condition at the corner. Another is to deal with the endpoint condition at the central symmetry position.(4) A method is presented which determines cave and convex of curve and fairs hull lines. The curvature of fairing curve should change equably. It is convenient to calculate the curvature while the neural network of the SWF is used to carry on curve fitting. This makes it possible for fairing multi-set molded lines. When performing the mathematics fairing of molded lines, the SWF can express the slope rate of each point since the function of each segment curve is existed, that unreasonable point can be found and correction can be made.(5) A practical method is proposed which solves the problem of big turn corner and partly distortion when curvature is big. Because of ship lines complication, small turn corner condition which is required by SWF can not be completely satisfied when the neural network of the SWF is used to carry on the curve fitting. To solve this problem, three methods are used:â‘ to circumvolve the coordinate of molded lines with 90 degrees;â‘¡to split curve into different segments and to deal with them separately;â‘¢to add points at dividing position.The research result of the dissertation makes small scaled shipyard to be able to use computer lofting. In real world, SWF can fit curve better. By comparing slope rate, it can quickly find out the point which curvature changes greatly and perform fairing on the molded lines to achieve a better effect. The experiment on sailing of ship(a Three Gorges Reservoir ship)shows a good ship speed (25 km/h).
|
PDF Full Text Request