Research On Influencing Factors Of Ship Collision Damage Based On Logistic Regression

In the recent years, ship collision accident rate has been effectively controlled but the number of ship collisions, especially accidents with high severity level, remains high. As a result, an in-depth analysis of the influencing factors of ship collisions and the method of carrying the research results into accident prevention, loss reduction and post-accident grade prediction and finalization thus respond with effective search and rescue forces have been an important research subject for the maritime researchers.Based on the review of previous study results and actual cases, the article selected possible factors that may affect ship collision results. Through the collection of more than 200 domestic and international ship collision accident reports, data have been extracted and analyzed. The quantitative relationships and criteria between accident results and affected factors have been proposed and a SPSS database containing detailed data on 172 accidents has been established. Firstly, the independence among possible factors and the correlation between these factors and ship damage have been tested with SPSS contingency table. Results show that the sizes, types, speeds of both ships involved in collision, collision location, collision angle, ship ago of the ship being hit and encountering trend are related to accident result. Secondly, Logistic Regression Modeling has been used to build ship collision accident model and the differences between the independence testing with contingency table and the Logistic regression model have been analyzed. With the application of the Logistic model, the influences of the four factors, gross-tonnage ratio, collision position, speed of the ship being hit and collision angle, on the seriousness of the collision results have been measured. Furthermore, the model provides a way of prediction on the probability of severer accident result from each of the above factors under different combination.