Human Error Recovery Modeling Using Fault Tree Analysis (FTA): Case Studies in the Steel Industry

Human reliability is determinant for process safety and operational consistency in steelmaking. Error recovery contributes directly to the final failure probability, yet applied studies addressing how system characteristics influence recoverability and how to represent error and recovery in an integrated model remain limited. Human Reliability Analysis (HRA) techniques provide a basis for recovery modeling, but their guidance is often insufficient for complex industrial applications. This study presents a methodology that integrates Fault Tree Analysis (FTA) with the HRA techniques HEART, SPAR-H, Petro-HRA, and APOA to model human error recovery in operation and maintenance tasks in steelmaking processes. The approach is structured around a cognitive recovery sequence composed of detection, diagnosis, planning, and execution, represented in FTA by an AND gate combining the failure of each stage. Human Error Probabilities (HEP) were estimated using HEART and SPAR-H to assess recovery under operating conditions. Two FTAs were developed as case studies: hot metal sampling in the blast furnace, including selection of the sampling point and positioning of the lance in the metal stream, and replacement of the planetary gear shaft of an overhead crane. The results show differences between HEP estimates, indicating that HEART provides a more detailed representation of recoverability due to its contextual criteria. The proposed methodology supports the identification of critical tasks and deviations affecting recovery and can be generalized to other human-centered activities.

Keywords: Human Reliability Analysis; Steelmaking; HEART; SPAR-H; Petro-HRA; APOA; Fault Tree Analysis; Error Recovery; Human-Machine Interface.