A Crowdsensing-Based Framework for Indirect Bridge Monitoring Using Mel-Frequency Cepstral Analysis Considering Elimination of Operational Effects

Abstract

This paper puts forward an indirect bridge monitoring method using Mel-frequency cepstral analysis of inverse-filtered drive-by acceleration signals collected through smartphones. Crowdsensing-based approaches using data collected by smart cars and smartphones opened a new chapter in bridge monitoring by reducing the costs and increasing the efficiency of the bridge monitoring process. However, the major challenge of the dominancy of the operational effects in the recorded drive-by vibrations overshadows the bridge monitoring objective. This paper proposes an inverse filtering-based monitoring approach to suppress operational effects. The inverse-filtered spectrum is later employed in a Mel-frequency cepstral analysis, leading to the calculation of the abnormality index, which is then used to detect the change in the bridge state. The performance of the proposed method in suppressing operational effects is assessed through a series of laboratory and real-life experiments. Afterward, the damage detection capability of the method is investigated for two damage levels at different locations along the bridge, modeled in a laboratory environment. The results provide evidence for the capability of the proposed method in drive-by damage detection of bridges. Moreover, using the smartphone as the data acquisition device paves the path toward the implementation of the method for crowdsensing-based bridge monitoring in future smart cities, although more operational factors such as passenger interactions and resulting smartphone motions need to be considered in future studies.