This paper provides an application of fuzzy logic to the risk assessment in transport of hazardous materials in order to evaluate the uncertainties affecting risk estimates. In fact, in spite of the significant steps forward taken in the last years in increasing knowledge of input variables (for example accident frequencies) and involved physical phenomena, no one could be able to calculate an 'exact' risk value. Therefore uncertainty evaluation is necessary for establishing the width of a risk interval, delimited by a minimum and a maximum risk value; this evaluation is even more important if risk measures are used to define the minimum risk route: the minimum risk route should be determined by comparing risk intervals instead of single risk values or risk bands instead of risk curves. In this paper both road and pipeline transportation of ammonia has been considered and both local and societal risk (represented by means of F/N curves) have been evaluated with efficient mathematical algorithms. Owing to the complexity of these mathematical procedures, a significant deal of computer time is required to achieve the final risk measures. Thus a traditional technique for quantifying uncertainty like the Monte Carlo simulation, which requires a huge number of iterations, can not be applied. When complex computer codes are required a modern approach to uncertainty evaluation by means of fuzzy logic is more appropriate, since the computational effort is substantially reduced. Furthermore the limited values for input data available are just sufficient to describe the uncertain parameters as fuzzy sets, but do not allow to represent them by means of probability density functions which would be necessary for the Monte Carlo simulation. Using fuzzy logic, the uncertain input parameters are described by fuzzy numbers and calculations are performed using fuzzy arithmetic. The outputs will be fuzzy numbers too. This paper describes how the uncertain input data (like the accident frequencies or the hole sizes through which ammonia is released) can be represented by fuzzy numbers, i.e. how the central value, the lower and upper limit and the shape of the fuzzy sets should be chosen. Combined uncertainty and sensitivity analysis are performed to show which effect each uncertain input has on the output uncertainty. Finally test results are presented and discussed.
Go to:
Table of Contents of the 1997
SRA-Europe Conference
Program of the 1997 SRA-Europe Conference
RiskWorld Welcome Page
Tec-Com, Inc.