Abstract Bridge Weigh-in-Motion (BWIM) systems aim to monitor the axle loads of vehicles crossing the bridge without the need to stop these vehicles. Overloaded trucks can be filtered out in this way. This paper aims to analyze how the a priori knowledge of some vehicle parameters and the loss of data due to sensor failure affect the axle load estimation errors of uniaxial strain gauge-based BWIM systems. First, an incremental examination of Moses’ matrix axle load estimation method is executed providing deeper insights on the utility of the a priori knowledge of each vehicle parameter under different environmental conditions. Then, the effect of data loss due to sensor failure is also examined. A multivariate linear time-series regression model based signal reconstruction algorithm is proposed to reduce the axle load estimation errors caused by sensor failure. Results on a real measurement-based annotated corpus show that the proposed BWIM system with the signal reconstruction algorithm achieves lower errors than without it. The COST 323 classes without the proposed algorithm are in the D-D+ classes, while using the signal reconstruction algorithm, it is in the A-C classes. The proposed solution can also be used in other stages of the BWIM pipeline like axle detection or speed estimation in the future.
Szinyéri et al. (Sun,) studied this question.