A Hazard Analysis Approach for the SOTIF in Intelligent Railway Driving Assistance Systems Using STPA and Complex Network

Abstract

The Intelligent Railway Driving Assistance System (IRDAS) is a novel kind of onboard system that relies on its own situational awareness function to ensure the safety and efficiency of train driving. In such systems, the use of situational awareness brings about a new fault-free safety problem, i.e., the safety of the intended functionality (SOTIF). It is essential to analyze the SOTIF-related hazardous factors for ensuring a safe train operation. In this paper, a hazard analysis approach is proposed to capture and evaluate SOTIF-related hazardous factors of IRDAS. This approach consists of an extended STPA-based hazardous factor identification part and a complex network-based hazardous factor evaluation part. In the first part, an extended control structure of STPA is designed for the modeling of the situational awareness process, followed by a new classification of SOTIF-related causal scenarios to assist the identification of causal scenarios. In the second part, a modeling method for heterogeneous complex networks and some customized topological indexes are proposed to evaluate the hazardous factors identified in the STPA causal analysis. The outcomes of the approach can help develop targeted hazard control strategies. The proposed approach has been applied to a new IRDAS operating in Tsuen Wan Line of Hong Kong MTR. The result shows that the approach is effective for the analysis of hazardous factors and is helpful for the formulation of hazard control strategies.

A Review on Risk Assessment Methodologies of Decision-making for Virtually Coupled Trains

“…Within the railroad sector, attempts have been made to use the system's analysis approach for risk analysis. Zhang et al (2021) used STPA for risk analysis of IRDAS to identify risk factors. Hao et al (2020) used STPA for risk analysis of virtually coupled trains to identify some potential hazards that are difficult to detect by traditional safety analysis methods, which helps to assist in system safety decision-making.…”

Safety Assurance of Autonomous Systems using Machine Learning: An Industrial Case Study and Lessons Learnt

“…STPA allows to identify a larger set of causes compared to traditional methods. Since it focuses more on the overall control loop and the safety impacts of interactions in the system or interaction of the system with the environment, STPA is suitable to analyze SOTIF aspects (see (Zhang, et al, 2021) or (Becker, et al, 2020)). However, the STPA method is complex itself and requires a lot of manual effort to perform the safety analysis.…”

Dynamic complex network based on status value for infectious disease

“…The effectiveness of complex network has been verified in lots of application [5][6][7] . In this part, the basic theory of complex network is introduced.…”