Estimation of Vehicle Counts from the Structural Response of a Bridge

Abstract

In this paper an offline method for counting vehicles travelling across a bridge through its structural response under loading is developed. Readings from a single vertically oriented accelerometer fixed to the bridge are normalised and then summarised by instantaneous amplitude envelopes. The envelopes for a single vehicle have a profile that is log-normal in appearance. Least squares fitting is used in conjunction with the Akaike information criterion to fit the envelope time series as the superposition of 𝑛𝑛 log-normal functions (other functional forms are also considered). It is hypothesised that this fit describes 𝑛𝑛 vehicles travelling across the bridge. This method is applied to data from a previous study on rapid deployment of structural health monitoring systems undertaken on the Clifton Suspension Bridge (CSB) in Bristol. The data from a single strain gauge-based accelerometer installed as part of a wireless sensor network (WSN) on the bridge is used and demonstrates the value added by this method to pre-existing asset monitoring systems. A prediction accuracy of 74% is achieved on a labelled test set.