To more accurately evaluate the vulnerability of urban road networks under disruptive events, aiming at the problems that the progressive evolution of links is difficult to be characterized by traditional cascading failure models and drivers'' actual route choice behaviors are difficult to be reflected by traffic reallocation methods, a study is conducted based on cascading failure theory. First, a bi-layer urban network model is constructed. On the basis of the traditional load–capacity model, an overload state is introduced to establish a three-state cascading failure model comprising normal, overloaded, and failed states, through which the response process of the road network under disruptions can be more realistically reflected. In addition, an origin-destination (OD) reconstruction strategy for overloaded links based on alternative path cost is proposed, by which drivers’ actual decision-making behaviors when facing congestion are simulated. Subsequently, a vulnerability evaluation index system is developed, including average system travel time, demand-weighted network efficiency, OD demand loss, and the number of overload occurrences caused by failed links. Finally, a simulation analysis is conducted taking the regional road network of Xuancheng City, Anhui Province, as a case study. It is indicated by the results that, after the overloaded state is considered, the model is capable of capturing the transition process of links from normal operation to failure, thereby avoiding the overestimation of failure range inherent in traditional binary models. Meanwhile, the OD reconstruction strategy based on alternative path cost is shown to significantly reduce demand loss, whereby the model''s capability to represent actual traffic behaviors is substantially enhanced. It is demonstrated that the proposed three-state cascading failure model can more realistically describe the dynamic evolution of road network states under disruptive events, and can provide a basis for critical link identification and emergency traffic management.