Hybrid bayesian networks for traffic load models from weigh-in-motion data

The Weigh-in-Motion (WIM) systems are used, among other applications, in pavement and bridge engineering, in infrastructure monitoring and assessment and inspection and reinforcement strategies. In the Netherlands and some other countries, the video-WIM system was implemented for pre-selection, and for continuously monitoring overloads. The systems record pictures of vehicles analyzed in the system together with measurements for number of axles, inter axle separation, vehicle length, weight per axle vehicular speed and total vehicular weight. In one month the WIM system in the Netherlands stores approximately 150,000 registrations of trucks. Currently, the WIM system operates in eight locations in the Netherlands. Because of the nature of the traffic, the quantities measured are regarded as random variables. The dependence structure of the data of such complex systems as the traffic systems are also very complex. Thus, the aim would be to be able to represent the complex multidimensional-distribution with a model where the dependence may be explained in a clear way and different locations where the system operates may be treated simultaneously. Bayesian Networks (BNs) offer an alternative for a model with the characteristics listed above. In this paper we discuss the construction of a large scale BN and results concerning its ability to adequately represent the data. The paper places special attention in the construction of the model. Furthermore, ideas as to how the model may be extended to include locations where the WIM system does not operate are given. These ideas benefit from structured expert judgment techniques previously used to quantify large scale Hybrid Bayesian networks (HBNs) with success. Copyright © (2012) by IAPSAM & ESRA.