A gradient based optimization procedure for finding axle weights in probabilistic bridge weigh-in-motion method
Bridge weight in motion (BWIM) consists in the use of sensors on bridges to assess the loads of passing vehicles. Probabilistic Bridge Weight in Motion (pBWIM) is an approach for solving the inverse problem of finding vehicle axle weights based on deformation information. The pBWIM approach uses a probabilistic influence line and seeks the most probable axle weights, given in-situ measurements. To compute such weights, the original pBWIM employed a grid search, which may lead to computational complexity, specially when applied to vehicles with high number of axles. Hence, this note presents an improved version of pBWIM, modifying how the most probable weights are sough. Here, a gradient based optimization procedure is proposed for replacing the computationally expensive grid-search of the original algorithm. The required gradients are fully derived and validated in numerical examples. The proposed modification is shown to highly decrease the computational complexity of the problem.