Contribution of artificial intelligence and machine learning to the assessment of the safety of critical software used in railway transport

Abstract

As part of the process of certification and commissioning of a new guided or automated rail transport system, the domain experts and in particular the National Safety Authority are responsible for reviewing the safety of the system to ensure that the safety level of the new transport system is at least equivalent to the railway systems already in service and deemed safe. This critical task of evaluating safety essentially concerns all the safety files prepared by the manufacturer and in particular safety studies such as the Preliminary Hazard Analysis (PHA), the functional safety analysis (FSA), the analysis of failure modes, their effects and of their criticality (AFMEC) or Software Error Effect Analysis (SEEA). The study presented in this paper is part of the SEEA analysis. To respect the completeness and consistency of this safety analysis (SEEA), the experts carry out complementary analyses of safety. They are brought to imagine new scenarios of potential accidents to perfect the exhaustiveness of the safety studies. In this process, one of the difficulties then consists in finding the abnormal scenarios being able to lead to a particular potential accident. This is the fundamental point that motivated this work. To help experts in this complex process of evaluating safety studies, we agreed to use artificial intelligence techniques and in particular machine learning to systematize, streamline and strengthen conventional approaches to safety analysis and critical software certification. The approach which was adopted in order to design and implement an assistance tool for safety analysis involved the following two main activities: – Extracting, formalizing and storing hazardous situations to produce a library of standard cases which covers the entire problem. This process entailed the use of knowledge acquisition techniques; – Exploiting the stored historical knowledge in order to develop safety analysis know-how which can assist experts to judge the thoroughness of the manufacturer’s suggested safety analysis. This second activity involves the use of machine learning techniques in particular the use of case-based reasoning (CBR).

Keywords

• artificial intelligence
• machine learning
• case-based reasoning
• Software Error Effect Analysis (SEEA)
• safety of rail