| Project/Area Number |
16510109
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Social systems engineering/Safety system
|
|---|
| Research Institution | Yokohama National University |
|---|
Principal Investigator |
SEKINE Kazuyoshi Yokohama National University, Graduate School of Engineering, Professor, 大学院・工学研究院, 教授 (40017934)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
YAMADA Minoru National Research Institute of Fire and Disaster, Project Program Division, Director, プロジェクト研究部, 部長 (60358778)
|
|---|
| Project Period (FY) |
2004 – 2005
|
| Project Status |
Completed (Fiscal Year 2005)
|
| Budget Amount *help |
¥3,700,000 (Direct Cost: ¥3,700,000)
Fiscal Year 2005: ¥1,200,000 (Direct Cost: ¥1,200,000)
Fiscal Year 2004: ¥2,500,000 (Direct Cost: ¥2,500,000)
|
|---|
| Keywords | Risk estimation / Statistical analysis / Safety policy / Large-scale accident / Risk curve / Dangerous material facilities / Decision making |
|---|
| Research Abstract |
In the risk assessment using statistical accident analysis by regulatory agencies, the main concerns are often on the risk of extreme events in large-scale accidents. This study demonstrates a statistical methodology for the risk estimation of extreme events on various industrial accidents and corrosion damages in bottom floors of oil storage tanks, based on the risk curve previously proposed by the authors, in which the exceedance cumulative probability in event occurrence are plotted on log-log scale as a function of their magnitude of damages. The analytic expression for asymptotic distribution for maxima in damages has been derived by means of the statistics of extremes and this concrete form has been shown to be the second type asymptotic distribution for maxima (Frechet type distribution). The relationship between characteristic parameters in the risk curves and their asymptotic form has been investigated. A link between statistical parameters for risk curve and asymptotic distributions has been shown and this has been verified by using the statistical real data on aircraft crash, labor accident, fire accident and leakage accident of hazardous materials.
From the practical point of view, to evaluate the predictability of the expected damage magnitude is considered to be significant. Through the analysis with actual accidents data and series of percolate simulations, possibility to predict the expected maximum magnitude of events by applying the extreme value theory has been presented. As the result, prediction by using return period might be applicable to the rough estimation. According to theses considerations, it is demonstrated that the proposed methodology is applicable to assess the safety management level.
|
