奥里油溢油扩散面积预测非线性方法研究.pdf

29 4 Vol. 29 No. 4 201111 THE OCEAN ENGINEERING Nov. 2011 : 1005-9865( 2011) 04-0087-05 龙新峰, 邱 平 ( , 510640) : , , , Matlab , , , BP , , Sobel BP 0. 88% , : ; ; ; BP : X55 : A Area fore ast resear h on Orimulsion spill diffusion at sea by means of nonlinear method LONG Xin-feng, QIU Ping ( Key Laboratory of Enhan ed Heat Transfer and Energy Conservation of the Ministry of Edu ation, S hool of Chemi al Engineering and Ener- gy, South China University of Te hnology, Guangzhou 510640, China) Abstract: Orimulsion is originally developed in Venezuela. It is a potential fuel whi h an repla e oal and petroleum. In re ent years, Orimulsion has been applied to many fuel pla es su h as boilers in power plants and industrial furna es in severam ountries. With the in rease of its quantity and fields of appli ation, the risk of Orimulsion spill aused by marine transport is also in reasing. Fo using on this phenomenon of Orimulsion spill, a nonlinear pro essing is put forward to fore ast the area of Orimulsion spill based on image pro essing and edge dete tion. It means that with the help of MATLAB, the image edges of Orimulsion spill are dete ted, the area of Orimulsion spill is al ulated, and the next time area of Orimulsion spill is fore asted using BP neural network. At the same time, the data from these methods is ompared with the data obtained from the experiments. The results show that the edge extra ted by Sobel operator is thinner, more pre ise than any other operators and it is better to deal with the noise and gray hanges using Sobel operator. The results also show that fore asted area data by BP neural net- wor