Use of regression trees to predict overweight trucks from historical weigh-in-motion data

The traffic of overloaded trucks is a critical problem in highways. It affects pavement performance life, reduces the service life of bridges, and has a negative impact on road safety, average speed and level of service. There are several practices to prevent the truck overloading issue, i.e., enforcement activities to verify the truck's compliance with the legal weight limits. This paper investigates the development of a method that uses available weigh-in-motion (WIM) data to identify overloaded truck weight and travel patterns. The proposed approach is based on regression trees method, a simple and easily understandable analytic tool used to build prediction models from a large set of data. An overall analysis of the overloaded truck regression tree model shows that the most important variable to classify and predict overloading is the truck type. Regarding the axle overloading, the most significant variable is the time of the day (most of the overloaded trucks travel at late night or early morning). The regression tree results can be used to optimize the efficiency of administration activities by planning truck enforcement operations based on the more critical scenarios. Also, the results improve the knowledge about the load characteristics of trucks, which can lead to more effective pavement management systems and more assertive pavement structure designs. •Traffic data collected from a WIM system installed in a Brazilian highway (BR-381).•Five regression trees are developed to analyze the truck and axle overloading.•The most important variable to classify and predict overloading is the vehicle classification.•The probability of a truck to overload at BR381 is 28.9%.•Results can be used to commercial vehicle inspection and monitoring.