An Experimental Analysis on the Use of Microworld Simulator

Abstract

Lack of data is a major challenge in human reliability analysis (HRA) [1, 2]. To date, the most broadly used HRA methods—such as the Standardized Plant Analysis Risk HRA—depend on a dataset provided by the Technique for Human Error Rate Prediction [2], which was generated from the early 1970s until the late 1980s, mostly from non-nuclear experience. Furthermore, although new technologies like digital main control rooms (MCRs) are already implemented on new or upgraded nuclear power plants, HRA methods still have been applied as is without modification to accommodate differences due to digital technologies. Accordingly, to update these data, several institutes and researchers tried to collect HRA data from different data sources, such as actual historical measurements, expert judgements, and simulator studies [3]. The majority of recent HRA studies focus on collecting data using full-scope MCR simulators with operators. For most current studies, these are predominantly concentrating on collecting data from simulator studies with full-scope simulators. The largest current efforts are led by the U.S. Nuclear Regulatory Commission (U.S. NRC) and Korea Atomic Energy Research Institute (KAERI). These efforts are collecting data from full-scope simulators using the Scenario Authoring, Characterization, and Debriefing Application (SACADA) database [4] and Human Reliability data Extraction (HuREX) [3] framework, respectively. But in fact, full-scope studies pose several intrinsic challenges for securing adequate data. A full-scope study entails high expense in securing a full-scope facility and numerous operators. Many experts in operating nuclear power plants (NPPs) also participate in the research. Because this work is relatively resource-intensive and time-consuming, in addition to presupposing utilities’ cooperation in partially releasing collected data, it is strictly limited to those few organizations able to satisfy such conditions. In keeping with the need for human reliability data sources, Idaho National Laboratory (INL) began collecting HRA data via the Rancor Microworld simulator, a simplified simulator, with student participants. In this way, INL has identified to investigate whether students could be used as subjects for collecting HRA data instead of operators in nuclear power plants when using the Rancor Microworld. As a first step toward achieving this goal, this paper compares human performances between operators and students measured across benchmark experiments, so that we can understand how much differences in the performances there are between the two subject groups. A randomized factorial experiment design was developed with two independent variables: type of scenario and type of subject. Six human performance measurements—1) time, 2) error, 3) workload, 4) situational awareness, 5) attention, and 6) number of manipulations—were selected. A couple scenarios and related procedures to be simulated by Rancor were then developed. The data collected from the experiment is analyzed using several statistical analysis methods like an analysis of variance (ANOVA) test. |인간신뢰도분석 (Human Reliability Analysis, HRA)은 확률론적 안전성 평가에서 요구되는 운전원 행위를 평가하고 그에 대한 오류확률을 정량적으로 분석하는 방법이다. 이를 위해, HRA 연구자들은 실제 데이터, 시뮬레이터 연구 또는 전문가의 판단 등으로 인간 신뢰도 데이터를 수집하여 인적오류확률을 추정하기 위해 노력해왔다. 그러나, 인간 신뢰도 데이터 부족 문제는 아직까지 HRA의 주요 현안으로 남아있는 상태이다. 대표적으로, 현재까지 수많은 HRA 방법론들이 개발되었음에도 불구하고 대부분의 방법론들은 1970년도 초부터 1980년도 후반까지 수집된 Technique for Human Error Rate Prediction (THERP) 데이터에 의존하고 있다. 이에 따라, 미국 Idaho National Laboratory에서는 HRA 데이터를 수집하기 위한 목적으로 Rancor Microworld 시뮬레이터를 개발하였다. 기존에 수행되는 연구들의 경우, 주로 실제 발전소와 유사도가 높은 시뮬레이터와 운전원들을 대상으로 실험을 수행하여 HRA 데이터를 수집해온 반면에, Rancor Microworld 시뮬레이터의 경우 비교적 단순한 시뮬레이터로서 운전원뿐만 아니라 학생들을 대상으로 실험을 수행할 수 있어 저렴한 비용으로 비교적 많은 데이터를 수집할 수 있다는 장점을 갖는다. 본 연구는 Rancor Microworld 시뮬레이터를 HRA 데이터 수집 용도로 활용하기 위한 사전 연구로서, Rancor Microworld 시뮬레이터 운전에 대한 운전원과 학생 피실험자의 수행도를 실험적으로 분석하는 것을 목적으로 한다. 본 연구에서는 총 6개의 인적 수행도를 고려하였고 몇 가지 통계분석기법을 활용하여 운전원과 학생 피실험자의 인적 수행도를 비교하였다.