www.bewet.fora.pl

loczytzn

Based on neural network prediction of flap rudder lift coefficient

. rlb = 15t. } A I, lL} j, l05101520253035 I (.) Lift coefficient curves of Figure 6 (Cy A, = 15.) Fig. 6Theliftingcoefficientcurve (Cy A, = 15.) Lift coefficient curves of Figure 7, C-(a = 15.) Fig. 7Theliftingcoefficientcurve (Cy A, a = 15.) Table 1 flap rudder lift coefficient table Table1Theliftingcoefficientofflaprudder set the initial learning rate of 0.4 for O. 8,[link widoczny dla zalogowanych], the tinkling of 1. O5,[link widoczny dla zalogowanych], expected error is O. O2. After learning to predict the effects of network testing, in order to compare, the paper also gives the approximate cost plus formula Ostrovsky also the results obtained, comparing the results shown in Figure 6 and 7, the figure \curve, \Plus also Ostrovsky approximate formula described below: Formula scope: 1) aspect ratio of 0.52.5 rectangular rudder; 2) thickness ratio of 28% of the symmetrical wing profile; 3) does not exceed the critical angle of the rudder rudder angle. At this point there is 1:4.75 × × (2.752 +2 a) 1 (2.752 +5.6). (11) Supplement flap rudder lift coefficient of Liu Sheng, et al: Neural Network Based Prediction of flap rudder lift coefficient C = kxCn1XCOSt2 '. (12) when the rear wing angle relative to the main rudder 20. , The coefficient k is calculated by Kara Ferry: k: 1 + Dan. (13) when> 20. Time, k is calculated according to Bao Liang Ke Siji: vii + ~ fT (1-0.17F); = (a (14) 7coc | .5 UU data showed that the BP neural network prediction of hydrodynamic coefficients of the flap rudder accurate than the value calculated by the approximate formula, indicating that the neural network has good self-learning function, to meet the needs of engineering applications. spectrum curve due to restrictions on the number, the sample can not be fully comprehensive data, leading to neural network nonlinear part of the lift coefficient curve approximation there is a big error, but with the approximate formula compared to the results obtained, the error is still acceptable .4 Conclusion flap rudder lift coefficient of the neural network prediction method for the flap Calculation of hydrodynamic properties of the rudder to provide a new way to manipulate the controls on the real ship, the hydrodynamic theory of computation is too complicated, and approximate formula for the calculation accuracy is difficult to satisfactorily study the rudder flap based on neural network hydrodynamic performance prediction is of great significance. From the prediction results,[link widoczny dla zalogowanych], simply select the appropriate network structure and learning algorithm, and obtain a full and comprehensive training data, BP neural network prediction accuracy can be achieved fully the requirements of engineering applications. Of course, to solve specific problems, the network structure and learning algorithm selection is not ready to follow the law, access to sample data has also been the original test conditions, therefore the use of neural networks is also a need for further research. References: [1】 Lingling Ding, Liu Sheng. warship main rudder / flap rudder control law of GPC joint [J]. Harbin Engineering University,[link widoczny dla zalogowanych], 2000 (6) :1-6. [2】 Yang Jianmin. Flap Rudder Hydrodynamics properties [J]. Shanghai Jiaotong University, 1997 (11) :133-136. [3】 LIUSheng, DUYanchun, LIWanlong, ZHENGXiuli.Sonararrayservosystembasedondiagonalrecurrentneuralnetwork [A]. IEEEInternationalConferenceonMechatronicsandAutomation,[link widoczny dla zalogowanych], ICMA2005 [C1.Canada ,2005:1912-1917. [4】 HAGANMT, DEMUTHHB, BEALEMH.NeuralNetworkDesign [M]. Beijing: China Machine Press ,2002:201-234. [5】 CYBENKOG.Approximationbysuperpositionsofasigmoidalfunction [J]. MathContrSignalSyst, 1989,2 (4) :303-314.

相关的主题文章：

[link widoczny dla zalogowanych]

[link widoczny dla zalogowanych]

[link widoczny dla zalogowanych]