The development of autonomous transportation has emerged as a great opportunity and big challenge for both researchers and manufacturers. When it comes to the autonomous train, the foremost objective is to maintain and assure the same level of safety as conventional trains. In case of the conventional trains, the driver has to stay aware of the train's surrounding environment and entities. The state of these surroundings change dynamically and the situation awareness of the drivers permit them to dynamically assess the risk associated with operation and take the right and safe driving decisions. However in the case of autonomous train, this dynamic risk assessment task must be performed by the autonomous driving system. Furthermore, in order to achieve this, the autonomous driving system shall remain self-aware of its surrounding environment at all times. In this paper, we present a novel framework of Situation Awareness-based Dynamic Risk Assessment for autonomous trains. This framework will ensure an acceptable level of safety by allowing the system to anticipate and adapt itself to unknown situations and take safe decisions. Finally, we present a use case of the proposed framework through application on obstacle detection function.