The Effect Analysis Of Semi-parametric Regression Model In The Influence On Fuel Consumption By Navigable Environment In Yangtze River

Navigation environment is the external environment of vessel navigating,and this article focuses on the influence of external factors on the ship's fuel consumption which is also the basis of modeling and improvement.Due to the relatively simple navigation environment of inland waterway traveling,the article studies on the relationship between fuel consumption and rate of main engine revolution on the condition of sailing in the Yangtze River from three aspects which are wind,water speed and water depth respectively.According to difference in the navigation environment,the whole channel has been divided into three segments including high,medium and low level areas by adopting the method of combination of pedigree cluster analysis and K means clustering validation,then build models to fit in an area of them.Extract the variables those are used to build models by adopting calculating correlation coefficient and principal component analysis.Meanwhile,select fuel consumption as the dependent variable and rate of main engine revolution as main independent variable which has significant effects on fuel consumption.In order to make comparison with the situation which factor are introduced in,reflect and analyze the effects of factors on the fuel consumption directly,at the beginning of modeling,nonparametric fitting is used to discover the tendency between fuel consumption and rate of main engine revolution.Meanwhile,compare the regression models without parametric form of different spline bases and adjust the fitting optimization in order to improve generalization ability by introducing regularized penalty factor.After taking the effect factors of navigation environment into consideration,multiple regression models with parametric form would be taken into consideration.Based on the significance of parametric estimation of multivariate regression model,the effect factors including interaction of factors are selected by stepwise regression.Make a compromise and integration combining the advantage and disadvantage of parametric regression model with those of non-parametric regression model in so that a semi-parametric regression model would be built.Introducing the effect factors of navigation environment as parameter regulatory factors into nonparametric model between fuel consumption and rate of main engine revolution can not only meet the needs of research object in this article,but also maintain the effect of nonparametric fitting.Considering the random data introduced into model in process of solution,mixed effects models are used to make the model more reasonable and convictive.The results come from adopting the test of significance and likelihood ratio test to make a simulation trial analysis of environmental factors and evaluate the model prediction.As a result,the semi-parametric mixed model is selected as the most appropriate regression model to channel fitting.Predict the consequence in channels of different grade velocity according to the clustering result that has been oriented from difference of navigation environment,the semi-parametric mixed effects model is built to compare difference of fuel consumption tendency on condition of a whole navigation environment.Use the final fitting-figure of trend to analyze the discrepancy in different channels so that we can figure out how to make fuel more efficient at a certain speed.At the meanwhile,provide practical suggestions for channel design of Yangtze River and optimization of ship speed.