Train-movement situation recognition for safety justification using moving-horizon TBM-based multisensor data fusion

Train-movement situation recognition is used in safety analyses to discern whether trains are running according to preset mechanisms. When a train deviates from the preset operation mechanisms, potential operation risks exist among the trains. Based on a transferable belief model (TBM), a mechanism of train-movement situation recognition is proposed in a rolling horizon through multisensor data fusion. The recognition procedure involves frame-of-discernment (FOD) definition, plausibility and belief calculation, pignistic probability calculation, and decision, which is applied to a dynamic process inference. The multisensor data fusion is formulated using the rolling-horizon TBM. The recognition mechanism permits the FOD cognition incompleteness and its discovery, which might involve the identification of aberrant dangerous situations. Using multiple positioning facilities, i.e., track circuits, balises, and global positioning systems, the risk prevention performance is verified through a train accident. The results demonstrate that the proposed recognition mechanism can correctly perceive train-movement situations and provide an early warning to avoid accident occurrences incurred by unfavorable train-movement situations.