RBF神经网络在深V型滑行艇阻力预报中的应用

基于SV、JYK系列滑行艇的阻力、浸湿面积、航行纵倾角试验数据,采用RBF神经网络建立了深V型滑行艇阻力预报数值图谱;针对艇艉底部横向斜升角变化的有限试验数据,提出了一种基于小样本试验数据的阻力修正方法。试验表明,该方法对深V型滑行艇(折角线长度与最大折角线宽度比在4~5.5,面积负荷系数在5.5~7,重心纵向相对位置在3%~9%,艉部艇底斜升角在5°~25°之间变化)阻力预报是可行的。在相同精度下,针对该文研究的问题,RBF神经网络所需时间少于BP神经网络。

Application of RBF neural networks to resistance prediction of deep-V planning craft

The RBF neural network was applied to predicting the resistances of deep -V planning craft based on the measured data of resistance, wetted surface area, and trim of series SV and JYK. According to the limited tested data, a new resistance modified method was presented, which can be used to predict the resistance of planning craft with a range of dead rise angles. The experiment verifies the method of predicting the resistance of deep -V planning crafts, with the ratio of projected chine length to the maximum breadth over the chine 4～5.5, the area coefficient 5.5～7, the longitudinal location of the center of gravity 3%～9%, the stern dead rise angle 5°～25°. In the same precision, the time used by RBF network is less than that used by BP network in solving the problem.