Continuous bridge displacement estimation using millimeter-wave radar, strain gauge and accelerometer

Monitoring of bridge displacement is essential for providing critical information regarding the health of bridge structures. However, continuous and accurate monitoring of structural displacement remains challenging. This study proposes a bridge displacement estimation tech-nique that combines an accelerometer, a strain gauge, and a millimeter-wave radar considering intermittent radar target occlusion common in long-term displacement monitoring. The technique primarily estimates displacement using radar and accelerometer measurements but switches to using strain and acceleration measurements when radar targets are occluded. A radar target occlusion detection algorithm is developed to automatically achieve this switching. Automated initial calibration is performed to select suitable targets from all radar-detected targets and es-timate the conversion factor for converting radar-based displacement from the line-of-sight di-rection to the actual movement direction. An artificial neural network model is automatically trained for strain-displacement transformation without the need for any pre-knowledge of a target structure. The proposed technique was validated via a laboratory test on a 10-m-long beam -type structure and a field test on a pedestrian steel box-girder bridge. The proposed technique correctly detected the intermittent radar target occlusion, and continuously estimated displace-ments for up to 75 min. The intermittent radar target occlusion leads to errors of a few millimeters in the displacements estimated using radar and accelerometers. However, the proposed method accurately estimates displacements with errors of less than 0.1 mm.

Automated summary

Monitoring bridge displacement accurately and continuously is challenging, but essential for assessing the health of bridge structures. This study proposes a technique that combines accelerometer, strain gauge, and millimeter-wave radar to estimate bridge displacement, considering intermittent radar occlusion. The technique switches between radar and accelerometer measurements when radar targets are occluded, and includes an automated calibration process and an artificial neural network model for accurate strain-displacement transformation.