Published: Građevinar 78 (2026) 3
Paper type: Scientific research paper
Download article (Croatian): PDF
Download article (English): PDF

Preliminary investigation of vehicle weight estimation employing acoustic signals from bridge expansion joints

This study introduces a method for identifying vehicle load levels by analysing the characteristics of sound signals generated when vehicles pass over bridge expansion joints. The proposed approach is non-invasive and independent of lighting conditions, offering significant advantages. Experimental tests were initially conducted and effective impact sound signals were extracted by analysing the trend of the sound pressure amplitude following filtering, pre-emphasis, and signal detection processes. Typical characteristics of sound signals in the frequency, time, and time-frequency domains were extracted. The relationships between short-term energy, empirical mode decomposition (EMD) energy entropy, spectral centroid features, and vehicle weight were found to be significant for vehicle weight identification. Finally, a classifier based on the k-nearest neighbour (KNN) algorithm was employed for vehicle weight classification by analysing the identification results under varying vehicle speeds and feature parameters. The results indicated that the KNN classifier achieved high accuracy in vehicle weight identification: 90.8% at low speeds and 83.1% at high speeds. The confusion matrix revealed that misclassifications tended to predict vehicle weights in adjacent categories.

LICENCE:

This paper is licensed under a Creative Commons Attribution 4.0 International License.