FRP Ship's Principal Optimization Based On General Regression Neural Network And Genetic Algorithm

As the premise condition of ship lines design, arrangement design and structure design, in the process of ship design, the selection and determination of the ship principal dimensions is an important part of the ship global design, directly produce an influence on the overall technical performance and economic benefits. Due to its remarkable performance advantage, the FRP (fiber reinforced plastic) fishing vessel has become the development trend of the world fishing vessels. By establishing the information database of the FRP vessels, the data base to research the relationship between the form elements of the FRP vessel, the selection of the form elements, and the optimization of the form elements is easily acquired.In this paper, firstly, the ship data in the FRP fishing vessel information database is regression analyzed, and then a regression equation is established, a series statistical formula of the FRP fishing vessel form elements are follow formed, which can be used in the preliminary design to estimate the form elements. And then using RBF neural network and GRNN neural network of MATLAB neural network toolbox to establish the mathematical model of the ship form elements, which is finally established by the error analysis to determine the GRNN neural network based prediction model, then using the neural network model predictions to look for the fine ship factor combination. Reference adaptive genetic algorithm and niche genetic algorithm ideas to improve the basic genetic algorithm, with improved genetic algorithm based on GRNN neural network mathematical model to optimize the ship form elements to get the best elements of several groups of coefficients, and finding it in the database coefficient of the same group elements of ship hull elements.Using the block coefficient (Cb), the ratio of the length and the displacement (L/â–³1/3), the ratio of the breadth and the draught (B/T) and the Freund's velocity coefficient as the input-variables, and using the Freund's resistance coefficient as the output-variable, to establish the ship form elements optimization model with the purpose of the best resistance performance, which not only meet the ship itself technical performance requirements, but also consider the economic performance, reducing the fuel consumption and the effect of energy saving and emission reduction. When using the genetic algorithm to make optimization of the mathematical model which established according to the neural network, it is no need to know the specific functional form, forwardly, its inherent heuristic random searching characteristics can be used to realize the global optimization of the discrete variables.