방사선 시설에 대한 확률론적 위험도 평가에 관한연구

Abstract

A procedure for the radiological risk assessment was established for radiation facility with vague information on risk contribution factors. In contrast to the nuclear power plant systems for which probabilistic risk assessments(PRAs) have been performed over three decades, risk models and associated variables used in PRAs are often vague in radiation facilities because of the scarcity of data. In that, experts provides valuable insights through his or her judgment on the uncertain but needed elements for risk analysis. Especially, the Delphi method was confirmed as a useful research tool for elicitation of expert opinions for a system with vague information. In addition, the streamlined procedure employs advanced techniques particularly useful for an object system with vague information, which include the Bayesian update and two-dimensional Monte Carlo analysis(2D MCA). The methodology developed in this study was applied, as an illustration, to risk assessments for two selected types of radiation facilities; 131I therapy facility and field radiography. Firstly, the utility of Bayesian updates was testified for risk assessments of the selected systems. The model variables for analyzing risks were obtained through three-round Delphi surveys. The resulting risks with the Bayesian updating of the variables were compared with both those without updating(3rd round Delphi survey) and those estimated by employing data within the 95% confidence intervals after the third round Delphi survey. The cumulative distribution functions(CDFs) without Bayesian updating showed unrealistically extreme values of radiation dose in the lower and upper tails. On the other hand, the risks estimated with Bayesian updating agreed with the risks reflecting those data in 95% confidence interval. The later, however, suffers intentional rejection of certain data collected. With this observation, the Bayesian updates were incorporated in the radiological risk assessment procedure. With the established procedure, radiological risks associated with those selected systems were assessed for both the workers and any members of the public possibly being involved. The exposure situations were groped into two categories according to the expected frequencies; normal and accident situations. Scenarios leading to exposure events were constructed by analyzing the observations gathered during visits of the workplace and the case accidents in Korea. Tasks with exposure potential were analyzed and the corresponding safety factors were identified. So were collected the risk model variables. Expert panel sizing 18 and 20 persons for 131I therapy facility and field radiography, respectively, from the operation groups and the regulation groups participated in the Delphi studies. The CDFs for each task generated through 2D MCA were used to plot the 90% confidence interval for the percentiles, from which the power of 2D MCA was exemplified by comparing the results with those from 1D MCA. In addition, the variability of radiological risk due to the normal tasks and accidents from 2D MCA were statistically summarized for different levels of percentile. In case of using-task in field radiography, 95%ile of median of the expected annual dose from normal tasks are 9.85mSv for the worker and 1.66×10-4mSv for the public, respectively, while in accidents the doses were 25.4mSv and 0.091mSv. The corresponding doses for the administration task in a 131I therapy facility were 3.35 mSvfor the worker in normal situations, while in accidents 0.35 mSv and 5.64×10-4mSv for the worker and the public, respectively. Doses from normal tasks do not exceed the domestic annual dose limits; 20mSv for workers and 1mSv for the public. These representative values of dose are judged to be reasonable when they are compared with the dose records of workers for normal operations and the expected exposure potential under an emergency situation. In spite of the reasonable representative values, the resulting uncertainties of the risk estimates for the two illustrative systems appeared to be unsatisfactorily large. Such wide ranges of the estimates can be attributed to several factors including unmatured risk models, lack of field data, limited size of the Delphi panels, and conservatism in the expert judgments even with the three-round surveys. However, this paper is of worth in view points that it established a framework for an advanced procedure of risk analysis for the radiation facilities where information on risk contributing factors are vague, by encompassing the PRA approach and those new techniques in data production and improvement as well as in quantification of consequences and their uncertainties with 2D MCA simulations. It is expected that the procedures established in this study, with certain refinements, will open the way leading to quantification of risks and their uncertainties in similar systems with vague risk information, particularly other type of radiation facilities.