Assessing and Mitigating impacts of structural inclination on acceleration measurements and estimated displacements using integrated GNSS and accelerometer structural health monitoring systems

• The characteristics of acceleration measurement errors due to bridge structural inclination are analyzed in both time and frequency domains for the first time. • A new method for fusing GNSS and accelerometer data while considering the acceleration measurement errors due to structural inclination is proposed where a dual-rate measurement update process is developed. • The limitations of the existing methods in addressing the problem are investigated for the first time. • The method was evaluated based on datasets collected from a shaking table experiment and a real-world long-span cable-stayed bridge in Hong Kong. Global Navigation Satellite Systems (GNSS) technique is usually integrated with accelerometers to better monitor structural health conditions. When an accelerometer is used in structural health monitoring (SHM), it is typically attached to the structure. In the case that the structure tilts due to structural deformation, the inclination angle of the structure introduces biases to the acceleration measurements. Such biases are often overlooked in current accelerometer and GNSS integration SHM systems. In this study, we present experimental results of the structural inclination impacts on accelerometer and GNSS SHM system using the dataset from a controlled shaking table test and a real-world cable-stayed bridge under heavy vehicle loadings and typhoon loadings. Results show that the significant impacts of bridge structural inclination on accelerometer measurements and estimated displacements derived from integrating the measurements. We also extend the existing multi-rate Kalman filter (MRKF) to include accelerations and their biases in the state vector and to estimate them simultaneously with displacements and velocities. The test results reveal that structural inclination of about 0.1° can induce over 10 mm/s 2 of acceleration measurement biases, even making the accuracy of GNSS and accelerometer fusion results using MRKF much lower than that of GNSS-only solutions. The frequency domain analysis shows that the acceleration biases mainly affect signals in low-frequency bands (i.e., less than 1 Hz) and do not affect the structural mode identification. In this case, the modified MRKF can accurately estimate and mitigate the acceleration biases, achieving about 13 % and 73 % of accuracy improvement compared to GNSS-only solutions and to current MRKF solutions, respectively. The acceleration measurement errors should always be considered if accurate SHM with accelerometers is required. Our findings could be applicable to other SHM systems that equipped GNSS and accelerometer.