Abstract: The Bayesian Network model is used to explore the relationship between times of port state control inspection and ship accidents. The data mainly include ship data from Lloyd’s Register of Shipping（LR）, accident data from International Maritime Organization（IMO） and PSC inspection data of three MoUs including Tokyo MoU, Paris MoU and India MoU. The research indicates that for Tokyo, accident rate goes up first then down as the times of inspection increase. However, for other two countries, increase of inspect times has little effect on the accident rate, suggesting inefficient of the inspections there.