基于BP神经网络的VVP水动力系数预报

Hydrodynamic Coefficient Prediction of VVP Based on BP Neural Network

将BP神经网络应用于全方位推进器周期螺距状态推力系数和转矩系数预报。针对BP神经网络在训练中存在的学习速度慢、易于陷入局部最小等缺点,采用变学习率的BP算法加以改进。对全方位推进器周期螺距状态推力系数和转矩系数进行预报,结果表明,预报值的精度明显高于由近似公式计算所得的值,采用BP神经网络对全方位推进器周期螺距状态的推力系数和转矩系数能进行预报是可行的,能够满足工程应用的要求。

BP neural network is applied to the thrust coefficient and model coefficient prediction of variable vector propeller in cyclic pitch. The variable learning rate of BP algorithm is adopted to mend the drawbacks that are common for BP neural network in training, such as, low learning rate, easy to get into the local minimum points. The thrust coefficient and model coefficient of variable vector propeller in cyclic pitch is predicted. Prediction the result shows that, compared with the value calculated by approximate formula, the value predicted by neural network has higher precision. So it is feasible to adopt BP neural network to predict the hydrodynamic performance of variable vector propeller in cyclic pitch, and it can also satisfy the need of engineering application.