The proper location of weigh-in-motion (WIM) stations in road networks is critical to effectively reduce the impact of overweight trucks. Truckers may quickly learn the locations of WIM stations, and take detours to bypass these checkpoints, with the aid of advanced navigation systems. This reaction needs to be considered in the location models for the deployment of WIM systems. This paper proposes a bilevel model to deal with the WIM location problem in a road network. The model includes an upper level and multiple lower level models respectively, represent the decision-making behavior of the law enforcement officials and truckers for different origin-destination pairs. The upper level model determines the optimal number and locations of WIM stations so as to minimize the damage due to evasive overweight trucks, considering the truckers' route choices. The lower level models simulate the truckers' route choices in response to the WIM locations determined by the upper level model. A heuristic is developed to efficiently solve the bilevel model, as it is difficult to obtain the exact solution. The proposed model and heuristic are evaluated using a test instance generated based on the Nevada road network. The results show that the heuristic outperforms a classical approach based on k shortest paths.