With the intersections of buried pipelines and highways increasing, the pipelines are under the threat of heavy loads. In this article, a full-section dynamic stress testing system is designed and constructed. A transient vehicle–road-soil-pipeline numerical model based on ADAMS and ABAQUS is established. On this basis, the influence of the vehicle rolling position on the stress distribution is analyzed. The influence of seven important factors, such as vehicle load and pipeline internal pressure, on the maximum stress of the pipeline is investigated. It is found that the maximum stress point always appears in the 12o'clock of the pipeline and the vehicle load most significantly affects the maximum stress of the pipeline. Furthermore, based on the simulation results and the GA-BP neural network, a maximum stress prediction model for the buried pipeline is established. The prediction model shows a high accuracy with a maximum relative error of 11.1%. Through this prediction model, an ultimate bearing capacity method for buried pipelines under vehicle loads is established and applied in three in-service buried pipelines.