ABSTRACT：The road with a Curve radius less than 400 meters has the highest rate of traffic accidents, and we have to use the straight-road barrier on curve-road for the lack of research on curve -road barrier. To change this unreasonable situation, we must obtain the optimal parameters of concrete barrier with curve radius of 400 meters. Therefore, this paper takes barrier with safety grade of the SS into account base on the research of parameters of the NJ-concrete-barrier. The total mass of research bus model is 18 t, the collision speed is 80 km / h, and the collision angle is 20 °, based on the ETA / VPG and LS-DYNA software platform and " vehicle – barrier – curve-road" integration vehicle - road system dynamics model, the research use combination of orthogonal experimental design methods and dynamic explicit finite element method. The research sets the rear wheel elevation as criteria for the evaluation and creates six cross-section shape design variables. The application of L25(56) Orthogonal table for simulation and the using of SPSS statistical package helps the variance analysis of simulation results, we find out the results of various factors on the degree of influence, then determine optimal parameters of concrete barrier. 25 Groups of the simulation results show that overturn accident occurred to all when large bus collide concrete barrier with a radius of 400 meters, and then bus run out of concrete barrier and fall down to cliff outside of the barrier. The optimal design results showed that the increasing in height of the concrete barrier is a important way to protecting bus from running out of barrier, and a concrete barrier with reasonable parameters can control the overturn-accident occurs inside, then avoid the occurrence of major badly-damaged accidents. The six cross-section variables which determined the sharp of the barrier have impact on the accident, the existence of priorities as following: height of lower slope> height of flange at the bottom > height on the upper slopes> top width> width of the lower slope > width of the upper slope, and the influence of height factors is greater than that of width, the height of lower slopes impact more significantly than that of the upper.