Human reliability study on the door operation from the view of Deep Machine Learning

Abstract

Abstract Despite the advances made in developing human reliability analysis (HRA) methods, a gap remains between the methods developed by HRA researchers and the ways HRA practitioners actually use. Automated process control and control room operating supervision are important in major process industries. This paper introduces operator’s operational and cognitive errors in the process reliability analysis based on a review of current research on HRA techniques, a procedural framework of qualitative screening phase for the control rooms in the metro industry context, and panel discussions with HRA analysts, systems analysts, and control operators. HRA method Standardized Plant Analysis Risk-HRA (SPARH) based on Machine Learning is utilized to evaluate human error probabilities (HEPs) of action and diagnosis in the Door Operation of Guangzhou Metro Corporation (GMC) Line 1. The results show that improving staff’s experience/training level will greatly enhance operating reliability of door operation. Considering diagnosis complexity and situational awareness, different decision makers have different effects on the operating reliability. We suggest that the quality of ergonomics/Human Machine Interface (HMI) and the availability of existing ergonomics/HMI assessment tools should be evaluated based on industry standards.