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

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