CLASCA: Learning System for Classification and Capitalization of Accident Scenarios of Railway
Résumé
In the process of analysis and assessment of the safety of a rail transportation system, one of the difficulties is to ensure the completeness of the accident scenarios taken into account by all the actors involved in the development of the system. The present work is to formalize, classify and archive the historical scenarios experienced on transportation systems in French already certified and/or approved such that the VAL, MAGGALY, TVM 430 of the TGV Nord. The goal is to develop a database of historical scenarios from the know-how of the manufacturers, masters of book and experts and researchers from the French Institute IFSTTAR to help examine the completeness of safety analyzes. The development and the operation of this basis of scenarios have need resort to the techniques of knowledge acquisition and automatic learning. The application of methods for the acquisition of knowledge has resulted essentially on the constitution of a database of historical knowledge which comprises 70 scenarios relative to the risk of "collision". The exploitation by machine learning of this basis of scenarios in order to extract the relevant knowledge in a purpose explanatory or made decision-making the object from the system "CLASCA" presented in this paper.