Fire Risk Assessment and Simulation of Electric Logistics Vehicle based on Data-driven Bayesian Networks

International dual carbon strategy and net zero action has significantly facilitated electric logistics vehicles. However, the fire accident is a threat for the logistic safety. A data-driven Bayesian network model is proposed to assess the fire consequence in electric logistics vehicles (ELV) based on 469 fire accidents, focusing on safety concerns, fire probability, and the evolution process. Firstly, accident data is collected to identify risk factors. Secondly, the EM algorithm is used to obtain the conditional probability table of the Bayesian network and establish an assessment model for fire consequence in ELV. Thirdly, FDS software is used to simulate specific fire scenarios in ELV and collect fire parameters. The results show that there is a 61.5% probability that the thermal runaway of the electric vehicle (EV) battery will lead to a complete vehicle fire, with a 67.1% probability of freight being lost totally. After the combustion becomes stable, the area 2.45m away from the ELV is relatively safe, and the thermal radiation intensity of the rear door of the vehicle compartment is lower. Therefore, targeted design and management recommendations should be proposed to reduce the risk of fire.