Research On Rapid Assessment Method Of Overall Damage Of Hull Subjected To Non-contact Underwater Explosion Based On Machine Learning

With the diversification and replacement of anti-ship/anti-submarine weapons,ships are faced with an increasing threat of damage.How to quickly and effectively evaluate the damage effects from anti-ship/anti-submarine weapons during the ship design phase is critical.This paper mainly studies the overall damage assessment of hull under the action of underwater non-contact explosion.Due to the high cost of real-ship trials,in order to be able to complete the preliminary assessment of the overall damage of the ship during the ship design phase,historical data and finite element simulation techniques must be used to form the original data and a data processing method should be used to obtain the assessment model.As for this,two research objectives are proposed:(1)It is possible to quickly assess the overall damage of the hull under any impact environment preliminary;(2)Ability to apply data processing capabilities of artificial intelligence to complete existing vessels based on the real ship or model test data or the calculated simulation data,establish a set of prediction and evaluation models.In order to achieve the research purpose of this topic,this paper takes the hull girder as the research object.Based on the previous work of the research group,it puts forward the requirements for rapid assessment.This article carried out the following research work:(1)The underwater non-contact explosion theory and numerical calculation methods are introduced.The theory of underwater explosion physics,calculation of underwater explosion load and numerical calculation of underwater non-contact explosion based on Abaqus are introduced respectively.The numerical simulation process and keyword setting of Abaqus underwater non-contact explosion are described in detail.(2)The comparison between the submarine non-contact explosion test and the numerical simulation results of the hull girder model is verified.First,the hull girder model test was introduced,and four groups of numerical simulation models corresponding to the test conditions were established using Abaqus software,and corresponding simulation calculations were performed.Then,the test and simulation results under each operating condition were compared to verify the model and method reliability of the numerical simulation.The verification results show that the simulation model and method are reliable.Finally,the results of the validated numerical simulation model are further analyzed in detail.(3)Matlab/Python/Abaqus data combination technology was used to perform batch numerical simulation of hull girder model.First of all,we introduced in detail the use of Matlab/Python to jointly modify Abaqus inp calculation files,submit inp files through Matlab to Abaqus background to calculate,Matlab use Python script to read odb data and do analysis and processing of these three Abaqus secondary development sub-topics,wrote a related program script.Then,based on the numerical simulation model in the work(2),the amount of explosives and the explosive distance were changed within a certain range,and 50 sets of random data were generated.Finally,using Matlab/Python/Abaqus data combination technology,the inp file corresponding to the numerical simulation calculation in work(2)was batch-modified,batch task submission and batch post-processing of odb file data,and the odb file of each group was obtained.The data of the maximum deflection assessment index of the hull under the non-contact explosion.(4)The assessment index of the remaining damage factor of the hull under the action of underwater non-contact explosion is defined,and the calculation method of the index is studied.Two methods,empirical formula and finite element simulation,were used to calculate the index.When the finite element simulation was used to calculate the residual intensity factor,the remaining intensity factor data corresponding to 50 random conditions were calculated using the quasi-static method for the hull girder intact and damaged cabin segments,respectively.(5)Using the support vector machine neural network method,the rapid damage assessment of hull girder model under underwater non-contact explosion is realized.Based on the total damage evaluation index data(maximum deflection and residual intensity factor)corresponding to the 50 sets of random conditions obtained in the previous section,using the GA and PSO parameters of the Matlab toolbox to optimize the support vector machine neural network model,the overall hull damage index Training,testing.The prediction error of the two-parameter optimization method is compared with that of the support vector machine neural network,and the high-precision parameter-optimized support vector machine neural network(GA-PSO)model is selected as the rapid assessment and prediction of the hull girder model under underwater non-contact explosion model.This paper combines Matlab/Python/Abaqus data combination technology and machine learning methods to achieve rapid assessment of the overall hull damage under non-contact underwater explosion.This is the first domestic exploration in this field.This method provides support for the calculation verification and preliminary evaluation of the technical design stage,and it has a certain guiding role in the design and research and development of ship structure impact resistance.