Human Error Prediction with Reliability Life Data Analysis Compared to Classical Human Reliability Assessment: a Case Study in Freight Train Railways

The adoption of automation technologies in freight train operations has significantly reduced operational risks; however, maintenance tasks remain critical to railway safety. Consequently, identifying, addressing, and predicting human error is essential for improving railway reliability. This study proposes the use of Reliability Life Data Analysis (LDA) as an alternative approach for predicting human failure and compares its performance with the classical Human Reliability Assessment (HRA) method, SPAR-H, in the context of a large-scale railway logistics operation. A quantitative research approach was adopted using a case study methodology. Human failure events were stratified according to task type and Performance Shaping Factors (PSFs) and distributed along a timeline. The categorized events were then transformed into error-interval data for reliability analysis. In parallel, selected human errors were assessed using the SPAR-H method. The results obtained from both approaches were qualitatively compared, and human failures were analysed in relation to the incidence of selected PSFs. The analysis was based on a historical database of human failures recorded over one year, classified according to the PSFs defined in API Guidance 770. Results were presented using reliability and probability plot, highlighting PSFs and task types with the best goodness-of-fit. The findings indicate that LDA enables predictive insights for specific task categories, while SPAR-H provides static probability estimates adaptable to different contexts. As the methods differ in nature, they are complementary and can jointly support the identification of failure patterns and the development of more effective error reduction strategies.

Keywords: Human Reliability, Human Error, Freight Train Railways, Life Data Analysis, SPAR-H, API 770.