Application Of Surrogate Model In The Prediction And Optimization Of Ship Resistance

With the development of computer and data processing technology,many scholars have carried out the work of ship hydrodynamic performance prediction and optimization by surrogate models.However,at present,most of the research are based on the data set about ship form and performance,and a large number of ship form sample points extracted from the design space to construct the surrogate model of the single form,there are relatively few studies on the influence of model form selection and number of sample points on the prediction accuracy.Therefore,focusing on the ship resistance performance,this paper takes open test data sets of ship model and common ship forms as the research objects,carries out the research on the prediction and optimization of ship resistance performance by the surrogate model.Firstly,the test data of 60 series ship model were used to analyze and compare the prediction effects of many common single form of surrogate models.Aimed at the problems existing in the performance prediction of single models,a combined surrogate model based on genetic optimization algorithm was proposed.By testing the combined surrogate model,the results show that this method can improve the prediction accuracy on the premise of ensuring the construction efficiency and stability.In order to further verify the applicability of the method,Taylor series data sets were simulated with the same test process,and good prediction results were also obtained.Then the Wigley ship form was taken as the research object,free form deformation method was applied to change the ship form within the range of parameter variation.The commonly used Sobol,orthogonal test and Latin hypercube sampling methods were used to extract a different number of samples in the design space,and the SHIPFLOW software was used to calculate the wave-making resistance under the design speed of the ship form samples to generate the data sets about ship form and performance.By comparing the density of the samples and the prediction accuracy of the model,a new method of selecting small ship form samples of Latin hypercube based on sensitivity analysis was proposed.Compared with the above three sampling methods,this method not only improves the stability,but also can be used to construct the surrogate model that meets the requirement of accuracy under the condition of selecting a small number of sample points of ship form.Finally,the method of sampling and combined surrogate model construction proposed in this paper were used to optimize the total resistance of KCS's bulbous bow at the design speed under the constraint conditions combined with CFD and genetic algorithm.The optimized scheme was checked by CFD,and the total resistance was reduced by 1.1%.Compared with simply using CFD technology for resistance optimization,the optimization process can greatly improve the optimization efficiency.The above results show that the sampling and combined surrogate model construction method established in this paper can realize the rapid prediction and optimization of the ship's resistance performance,and provide a feasible scheme for the surrogate model to be applied to ship form optimization.