The Research Of Trawler's Seakeeping Performance Base On Neural Network

With the decline of coastal fishery resources, the international community increasingly stringent control of fisheries resources, and offshore fishing becomes the inevitable trend of development. Currently, the large ocean-going trawler in variable water level is a main kind of fishing vessels, designing and constructing short-fat-based trawlers which has excellent performance and economic benefits become a hot research spot. In the conceptual design stage of trawlers, we must first analysis and demonstrate main dimensions and hull form of the design vessel,and obtain the best ones to arrive at benefit of the overall performance. Taking into account the climate and operations characteristics of trawler sailing on the ocean, the importance of researching seakeeping performance have no doubt. With the rapid development of modern fishing technology, trawlers'hull form has huge changes, if we predict seakeeping performance of multiple-scale trawlers quickly and accurately only according to the original statistical data, these result of principal dimensions and seakeeping performance relationship is not credible.According to the analysis of the foreign lines plan and the domestic research hull form in this thesis, the author used a home-based trawler's lines as the initial plan and generates hull form database with the same characteristics in the scales based on a ship form conversion program; then through detailed 3D time-domain panel method calculation, the seakeeping database was established correspondingly. Faced with such large amounts of data, in order to more efficiently and accurately predict the any dimensions of the trawlers'seakeeping performance, based on neural network theory and algorithms, the author used the hull form database as input samples and the seakeeping database as output samples,then determined the type of network structure, the training functions, transfer functions and the other network elements.A BP neural network model which meets the accuracy requirements for trawlers seakeeping prediction was establishd and it can predict the frequency response function of the motion. With the neural network model of the calculation results, the disciplines of the parameters and the hull form coefficients on the impact of ship movement was concluded.In the view of the weights from the network, the importance of each parameter on seakeeping performance was analyzed.Finally in the paper, the trawler conceptual design methods were studied through a program which contains 5 initial designs the as an example. Using the designed neural network model, and the sea conditions of trawlers operating, the motion of trawlers were short-term forecasted through spectral analysis method, and the 5 kinds of programs were evaluated and optimized, and the result can be used for pre-trawlers conceptual design or preliminary design and it has some significance and practical value in engineering.