Study On The Data Feature Processing Method Of Ship Shock Environment Forecast

With the further study of ship-borne equipment and personnel impact resistance in recent years.A large amount of ship impact environment data has been accumulated,and intelligent forecasting methods represented by neural networks have emerged.However,the use of neural networks for impact environment forecasting still has the problem of insufficient forecast accuracy,unstable forecast results and so on.One of its core problems is that the input data processing of the forecast model is inadequate.In this context,in order to improve the accuracy and stability of intelligent forecasting methods.This paper studies in depth the processing method of neural network input data characteristics when the ship impact environment is intelligently forecasted,and then improves the forecast accuracy.This paper provides a method for the acquisition of the impact environment for the demonstration stage of the ship’s plan and the demonstration of the impact resistance of the new ship.The details of the study are as follows:(1)Based on the real ship,this paper obtains a large amount of impact environment data through simulation calculation.Using My SQL software to classify and store data,the impact environment data is structured and standardized storage.Form an impact environment database and build an operating interface.Easy to increase,delete,modify,query and other operations of data.The PSO optimization algorithm is selected to optimize the PNN neural network.(2)The adaptability of PNN network model is analyzed and studied,and the relevant parameters of optimization model and PNN neural network are set respectively.Based on the previous research and analysis experience,the parameters with higher environmental impact weight on ship impact are screened.The input parameters are eventually defined as 11 × 9)-dimensional parameters that have been harmonization by magnitude.This parameter is used as input to forecast the impact environment of the ship’s inner bottom and deck 1.(3)From the point of view of volume analysis,a new method of data feature processing and dimensional reduction is proposed.Refer to the idea of the analysis of the outline,and classify the physical quantity involved in the problem by attribute.Then find out the connection between different physical quantities,and convert the original data features into non-measured physical quantities.In this paper,the method of volume analysis is applied to the processing of data features.Its basic ideas and methods can also be used in the study of data processing and dimensional reduction of other problems.(4)The influence of the processed data on the forecast results is analyzed comprehensively by using the feature extraction method.Neural network input parameters are processed.The characteristics of neural network input data are processed by PCA main component analysis and PLS partial least-multiplication,respectively.Reduces the dimensions of input parameters and reduces the correlation between the original data.The duplicate information between the data is eliminated.Get more optimized input data.The spectral values of the ship’s impact environment output are forecasted by using the data obtained.Finally,the stability and accuracy of the forecast results of different methods are compared and analyzed.