Ship resistance prediction based on neural network
Conventional resistance prediction method of proxy models takes main scale ratios,ship form coef-ficients,and other similar parameters as inputs.Compared to CFD calculations,in which the complete hull form is used as input,prediction method with lower information density of proxy models results in lower pre-diction accuracy.In this paper,a high-dimensional,high-precision resistance prediction method was pro-posed for ship hulls using 4108 sets of complete hull geometry feature tensors as input and employing neural networks as proxy models.The total resistance coefficient of the ship was taken as the output.Dimensionless treatment of the hull forms was conducted at first and feature tensors were extracted as inputs.Next,a neural network model was constructed,comprising input layers,hidden layers,and an output layer.Finally,the fea-ture tensors of the hull forms and the corresponding total resistance coefficients were fed into the neural net-work,and the model was trained using error back propagation until the loss function converges.The research findings in this paper can provide theoretical and technical support for high-dimensional proxy model-based resistance performance prediction.
万方数据
维普
