A weighted multi-output neural network model for the prediction of rigid pavement deterioration

A novel weighted multi-output neural network (NN) model is proposed for predicting the deterioration of rigid pavements based on Iowa pavement management system data. This first-of-a-kind simultaneously predicts four condition metrics concerning pavements, including IRI, faulting, longitudinal crack and transverse crack. It provides an opportunity to efficiently evaluate conditions make treatment decisions multi-condition metrics, such as index (PCI) budget allocation models. Compared traditional single-output NN models, this capable incorporating correlations among different metrics. During training, each metric assigned a weight reflect its relative importance. When weights equal those in formula metric, prediction performance PCI optimal (13% lower MSE than optimal, models). The improves three individual compared Results show that consideration could improve single Finally, variable weighting critical achieving balance various dictated by needs decisionmaker.