Research On Fast Prediction Method Of Ship Resistance Performance Based On Convolutional Neural Network

Hull form design optimization has always been the focus of people’s research.Engineers usually use the Computational Fluid Dynamics(CFD)method to predict ship performance based on model scales,and based on this,carry out hull form optimization,then select the best solution from it and apply it to real ships.However,considering the effect of scale effect,the optimization results based on model scale cannot necessarily achieve the desired optimization effect on the real ship.Moreover,the accuracy of CFD method is affected by a variety of factors,such as mesh generation,turbulence model,etc.,which requires users to have a lot of experience.In recent years,with the rapid development of computer hardware and software technology,classification and regression of data with complex relationships based on machine learning algorithms has become a research hotspot in various fields.It is becoming a trend to use machine learning algorithms to solve complicated problems in the field of ship engineering,which has high engineering application value and scientific research significance.In this paper,a fast prediction method for ship resistance performance is established based on convolutional neural network,and the effectiveness of the method and the correlation between ship type and ship resistance are systematically analyzed,the specific research work is as follows:First of all,this paper used Python language to build a convolutional neural network model in the TensorFlow framework.It studied the functions used in actual programming and the settings of their parameters,clarified the meaning of parameters and their impact on the data processing and performance of the model,and provided a code basis for building a convolutional neural network-based ship resistance prediction model.Secondly,in this paper,the nominal wake velocity field of different ships obtained after the modification of the 90000 DWT bulk carrier was taken as an example,a convolution neural network was built to predict the propeller torque coefficient KQ of the wake current field,which proves the convolution,proved the neural network suitable for the regression forecasting of matrix-type data similar to the ship’s accompanying flow field,and provided a practical basis for the establishment of ship resistance prediction model based on the convolutional neural network.Finally,this paper used SHIPFLOW software to calculate the total resistance of large oil tankers.Using the ship type data and total resistance value as input data and sample labels,a fast prediction model of ship resistance performance based on convolutional neural network was established.The resistance prediction accuracy of the model was verified by the hull model obtained by the fusion of the original ship type of a large oil tanker.The performance of the convolutional neural network model under different structures and different parameters was compared,and the influence of the model structure on the model performance was studied.Taking the large oil tanker as the research object,the sensitivity analysis is carried out based on the established ship resistance performance prediction model,and the correlation between the half-width of the hull profile and the ship resistance is studied,and some of them have a greater impact on ship resistance.This research provided a certain reference for hull form optimization design and lays a foundation for the intelligent hull form optimization design in the future.