Advances of Machine Learning in Clean Energy and the Transportation Industry

Abstract

The seriousness and negative consequences of accidents and incidents in rail transport have forced all players in the rail system to set up an feedback system (REX) in order to improve operational safety. In this context, the knowledge of accidents and incidents results essentially from the contribution of lessons learned and experiences acquired. This process of analyzing and using experience data can be facilitated by the use of artificial intelligence techniques, especially machine learning, in order to understand the origins and circumstances of accidents and therefore propose solutions to avoid the reproduction of similar insecurity events. The approach proposed for the prevention of railway accidents is a Hybrid method built around three learning algorithms and one knowledge-based system (KBS): • The acquisition of knowledge to gather knowledge of railway safety and in particular the scenarios of potential accidents, • Learning by classification of concepts to group accident scenarios into homogeneous classes such as the class relating to train collision or derailment problems. • Rule-based machine learning (RBML) to automatically identify, from a base of historical scenarios (experience feedback), the relevant safety rules that are often difficult to extract manually from safety experts, • Knowledge-based system (KBS). Production rules, previously induced by machine learning, are transferred to KBS to form the knowledge base of the safety assessment support tool. • Case-based reasoning system (CBR). At the previous level, the KBS is used to evaluate safety at the highest level of the safety analysis hierarchy and can deduce a possible risk of accident not taken into account and likely to jeopardize the safety of the system and by therefore the safety of hardware and software equipment. This risk of accident requires the implementation of new prevention or protection measures during the various safety analyzes of hardware and software equipment (low level of the hierarchy). In this context, the CBR makes it possible to look for the most similar cases to this new risk of accident and proposes the appropriate measures.