A Fast Prediction Of Cabin's Vibration And Noise Based On Artificial Intelligence Algorithm

In this paper,the problem of how to apply the artificial intelligent algorithm to the predic-tion of vibration and noise in cabin is studied.In the process of ship and offshore platform construction,the prediction of cabin vibration and noise is an essential step.Because all kinds of domestic and foreign standards have a strict limit on the vibration and noise of various com-partments,in order to avoid the remedial work due to vibration and noise exceeding the stand-ards,a prediction of the vibration and noise in cabin in the preliminary design stage and taking necessary measures are needed.On the other hand,in the stage of ship trials,because of the test time,working conditions and other reasons,it's usually difficult to get the results of cabin vi-bration and noise in all required conditions.So how to predicting vibration and noise in the rest of cabin according to the part which is measured is a practical problem.Intelligent algorithm has been widely used in engineering,it can deal with nonlinear map-ping problems,which have more parameters and complicated relationships.As for the complex ocean structures such as ships and offshore platforms,the vibration and noise of cabin are in-fluenced by many factors.The existing prediction methods are limited by accuracy or time-consuming,to solve this problem given article proposed the method of using BP neural network and support vector machine algorithm to predict the vibration and noise of cabin.First of all,the two intelligent prediction models used in this paper are introduced in detail,the important algorithms are deduced,and the corresponding MATLAB programs are written and the correctness is verified.Then based on the characteristics of ships and offshore platforms,the corresponding structural parameters are selected as input variables of intelligent prediction models.And a database for intelligent prediction models is set up,the source of the data is mainly come from two parts,one part is from sea trials of offshore platforms,another is from simulation by VA one software in statistical energy analysis method.Before putting the data into intelligent forecast model,the eliminate the polynomial trend item and normalization meth-ods are used for data preprocessing to solve the zero drift and the problem of parameters can not be used directly due to different types of variables which influence the vibration and noise of cabin.For both models,there are some parameters should be set up before training,such as active function,cost function and Kernel.The effects of predicting accuracy of vibration and noise are studied when choosing various parameters and functions of the intelligent algorithms.After the training was done,the order of significance of the input data related to ship features,such as thickness,numbers and areas of bulkhead,are studied in correlation analysis.The priority of the vibration and noise reducing measures has been determined.Vibration and noise of cabins which belongs to the same ship with training dataset or other ship and offshore platform are predicted respectively by well-trained intelligent algorithm,and got a relatively high accuracy.