Displacement data under operational loads are an important aid for the estimation of structural performance, but accurate measurement of structural displacement remains as a challenging task, especially for long-span bridges. The global positioning system (GPS) is the common choice for long-span bridge displacement monitoring but the measurement accuracy is not satisfactory. The purpose of this study is to improve the GPS accuracy using a practical data fusion method. Although the main algorithms of data fusion method based on multi-rate Kalman filter are already reported, the detail about how to select the noise parameters required for Kalman filter is not provided. This paper demonstrates that maximum likelihood estimation (MLE) can be used to determine the necessary noise parameters. The proposed approach was validated on numerical and field data, the latter from a single-day displacement monitoring campaign on the Humber Bridge in the UK. The direct measurement by GPS was merged with the collocated acceleration data and the resulting displacement signal was evaluated by comparing it to the displacement signal from an independent vision-based system. Through the comparison, it is shown that MLE enhanced data fusion is practical to improve the GPS measurement accuracy and to widen the frequency bandwidth. The MLE provides an estimation about the GPS noise (assumed as zero-mean Gaussian process) with the standard deviation varying from 6 mm to 16 mm in the test day. The normalised root mean square deviation of GPS measurement compared with the vision-based measurement was decreased from 3.17% to 2.37% after applying the data fusion.
|Publication status||Early online date - Jul 2017|
- long-span bridge
- vision-based system
- KAlmen filter
- Maximum likelyhood estimation