Abstract: In view of the problem of frequent accidents of trackless rubber-tyred vehicles on curves in underground coal mine, a risk warning index system of underground trackless rubber-tyred vehicles curve was put forward considering multiple factors. Based on UWB positioning track data and objective factor data, and taking the five-grade scoring method as the quantitative indicator of early warning factors, the grading early warning mechanism for the accident risk of underground trackless rubber-tyred vehicle on curves in underground coal mine was established. The SSA-BP neural network early warning model was constructed, simulation learning and training were carried out, and compared with BP and SVM models. The results show that the established grading early warning index mechanism is helpful to improve the safety of underground trackless rubber-tyred vehicle on curves, and the constructed SSA-BP neural network has high accuracy and stability for the early warning of accident risk of vehicle on curves in underground coal mine.