2021 Volume 12 Pages 11-18
The near miss reports are utilized for preventing accidents and incidents in various industrial fields. The authors analyze both actual near misses and hypothetilcal ones which crew expected to likely occur.
First, the data of 21,118 reports from Japanese domestic tankers are categorized into 21 risk factors based on 4M(Man, Machine, Media and Management) by text mining technique.
Secondly, the data is analysed by Principal Component Analysis ; it is shown that the degree of nature risk on tanker in each risk factor is corresponding to the first principal component and sensitivity against risk is corresponding to the second one. It is clarified that these two conponents closely relate to accidents and incidents ; they can be used as an index of factors behind accidents and incidents.
Furthermore, it is shown that there is a blind spot in near miss reports, not all actual unsafe event is reported, because poor sensitivity against risk cannot detect one.