Abstract: During train operation, there are numerous and complex factors that can cause derailment, and there are certain limitations in detecting derailment using derailment coefficient and wheel load reduction rate, making it difficult to collect related physical quantities. Firstly, through analyzing the mechanisms and factors of vehicle derailment, the key characteristic physical quantities that are easy to collect for derailment detection, namely, lateral acceleration and roll angle, are verified and determined via virtual simulation. Secondly, addressing the technical challenges of online derailment detection for rail vehicles, a derailment detection system scheme is designed. Then, the matching test equipment is selected according to the test scheme, and the software and hardware of the derailment detection system are established. Finally, based on the completed bogie derailment detection system, the tests are conducted on the railway freight car dynamics simulation model and the rail test line in Shanghai Institute of Technology for verification. Research result indicates that the derailment detection system designed based on lateral acceleration and roll angle can accurately detect the derailment of rail vehicles, which has important engineering application value for ensuring the safety of train operation.