Research On Data Mining Method Of Ship Impact Environment Based On Dimensional Analysis

In the field of ship anti-explosion and impact resistance research,the calculation of ship impact environment is an important means of assessing the impact resistance of ships and equipment.It is of great engineering significance to achieve rapid and accurate prediction of impact environment.Based on the large amount of calculation data obtained by simulation,data mining methods can be used to learn and mine the potential digital laws in these data,and finally realize low-cost and high-efficiency ship impact environment forecasting.In this context,this thesis firstly screens and analyzes the characteristic parameters in the structural impact problem of ships,etc.,and then performs a dimension reduction analysis on the input characteristic parameters of the mining model based on the dimensional analysis method and similarity theory;in addition,in view of the dimensional analysis Based on the different influences of different basic quantities in the selection,the paper constructs supervised and unsupervised models to analyze the selection of basic quantities.Finally,data mining models such as RBF,PNN,and SVM are used as the basic forecast framework to realize the numerical forecast of the impact environment of the stiffened plate frame,partial cabins,and the whole ship respectively.The specific research contents are as follows:(1)At present,there are many examples of data mining in the analysis of ship impact environment.This thesis summarizes these examples,and believes that the main problems and difficulties in data mining research on ship impact environment are the insufficient selection of characteristic parameters in the sample database and the sparse sampling density of each feature,which makes it difficult to effectively control the stability and accuracy of the data mining model when performing numerical forecasting.In response to the above problems,this thesis first classifies the factors that may affect the ship’s impact environment,and then filters out more complete characteristic parameters for subsequent use in parameter dimension reduction.(2)For improve the interpretability of parameter dimension reduction and make it comply with a certain physical essence,this thesis mainly analyzes and explains the necessity of the application of dimensional analysis method in feature dimension reduction,and puts forward a new dimension reduction method,namely the dimension reduction method "DA-ST",which combines the similarity theory and the dimensional analysis method.This method is based on the similar analysis of dynamic problems,and achieves the purpose of dimension reduction through the splitting of original features and the reorganization of dimensionless quantities.The feasibility of this method is discussed,deduced and proved in detail in the article.In addition,at the application level of this method,a calculation formula for the approximate shielding rate coefficient suitable for surface ships to absorb the explosion shock energy is established.(3)In view of the fact that this article is mainly based on the dimensional analysis method to study the characteristics of impact environment dimension reduction,when applying the DAST,it also analyzes the influence of different basic quantities on the mining model forecasting.The linear regression model and Gaussian mixture clustering model are used to evaluate the different choices of basic quantities.Finally,based on the determination of the optimal basic quantity and the use of new dimension reduction methods,the accuracy and stability of the data mining and analysis of the stiffened plate frame,local cabins and the entire ship’s impact environment have been improved to varying degrees.To sum up,the dimension reduction of the characteristic parameters of the ship shock environment is the focus of the research.This thesis provides a new idea for the numerical prediction of the ship shock environment by using the original features to reduce the dimensionality reasonably and effectively.