Welcome to the PSAM 18 Abstract Status page.
Abstract TA247Full Paper + Presentation
Quantitative Assessment of Human Error Probabilities and PSF Multipliers for Railway: An Expert Elicitation Study
Railway accidents occurring during operation, facility management, or infrastructure maintenance can lead to catastrophic casualties, as well as significant social and economic losses. According to statistics from the Ministry of Land, Infrastructure and Transport, South Korea experienced 86 major train accidents (e.g., collisions, derailments, and fires) between 2020 and 2024, resulting in 178 casualties. Notably, various research and reports indicate that approximately 50% of these incidents are attributed to human errors by personnel, including drivers and controllers.
Consequently, Human Reliability Analysis (HRA) methods are needed within the railway industry to quantitatively evaluate Human Error Probability (HEP), reduce error occurrence, and enhance task performance. However, performing reliable HEP assessments remains a challenge due to the scarcity of empirical data required to quantify nominal HEP and Performance Shaping Factors (PSFs).
The present study aims to generate highly reliable HRA data by applying a systematic approach to use experts’ knowledge. This study employs Cooke's Classical Model to quantitatively estimate the nominal HEPs and PSF multipliers required for HRA by systematically aggregating expert judgments. Expert opinions were elicited via a structured questionnaire administered to a panel of 16 experts with extensive experience and specialized knowledge in the South Korean railway industry. Regarding the target questions, nominal HEPs for diagnosis and execution errors are elicited, along with multiplier values for each level of eight predefined PSFs. For the seed variables, seed questions are formulated based on nominal HEPs and PSF multipliers drawn from existing HRA research in railway and other industrial domains. Based on the experts' judgments regarding the seed variables, informativeness and statistical accuracy are evaluated to calculate information and calibration scores. Utilizing these metrics, specific weights are assigned to each expert, and an aggregated assessment of the target questions is generated using the "expert" package in R, as proposed by Vincent et al. (2009).
← Check another abstractConsequently, Human Reliability Analysis (HRA) methods are needed within the railway industry to quantitatively evaluate Human Error Probability (HEP), reduce error occurrence, and enhance task performance. However, performing reliable HEP assessments remains a challenge due to the scarcity of empirical data required to quantify nominal HEP and Performance Shaping Factors (PSFs).
The present study aims to generate highly reliable HRA data by applying a systematic approach to use experts’ knowledge. This study employs Cooke's Classical Model to quantitatively estimate the nominal HEPs and PSF multipliers required for HRA by systematically aggregating expert judgments. Expert opinions were elicited via a structured questionnaire administered to a panel of 16 experts with extensive experience and specialized knowledge in the South Korean railway industry. Regarding the target questions, nominal HEPs for diagnosis and execution errors are elicited, along with multiplier values for each level of eight predefined PSFs. For the seed variables, seed questions are formulated based on nominal HEPs and PSF multipliers drawn from existing HRA research in railway and other industrial domains. Based on the experts' judgments regarding the seed variables, informativeness and statistical accuracy are evaluated to calculate information and calibration scores. Utilizing these metrics, specific weights are assigned to each expert, and an aggregated assessment of the target questions is generated using the "expert" package in R, as proposed by Vincent et al. (2009).