Accurate Correlation Modeling between Wind Speed and Bridge Girder Displacement Based on a Multi-Rate Fusion Method

Yan Wang, Dong-Hui Yang, Ting-Hua Yi2021 · DOI

TL;DR

This paper introduces an improved multi-rate Kalman fusion method to accurately fuse displacement and acceleration data from bridge sensors. This enhanced fusion allows for more precise modeling of the relationship between wind speed and bridge girder displacement, crucial for structural health monitoring and performance warning systems.

What Problem Does This Paper Address?

Long-span bridges are susceptible to environmental factors, particularly wind, which can induce significant static and dynamic effects, leading to structural damage over time. Accurately understanding the relationship between wind speed and bridge girder displacement is vital for early detection of performance degradation and preventing catastrophic failures. However, existing monitoring systems often suffer from noise and differing sampling rates between sensors (e.g., GPS for displacement and accelerometers for acceleration). Traditional fusion methods struggle to simultaneously correct both displacement and acceleration data, limiting the accuracy of correlation modeling. This paper aims to bridge this gap by developing a more robust fusion technique.

Methods

The core of this paper is the development of an improved multi-rate Kalman fusion method. This method builds upon traditional multi-rate Kalman filtering by introducing two Kalman gain coefficients. This modification allows for the simultaneous correction of both displacement and acceleration monitoring data, even when they are sampled at different rates. The fusion process involves:

  • State Estimation: Using a Kalman filter framework to estimate the bridge's state (displacement, velocity, acceleration).
  • Time and Measurement Updates: Iteratively updating the state estimate based on the system model and sensor measurements.
  • Multi-Rate Handling: Specifically addressing the differing sampling rates of displacement (low frequency) and acceleration (high frequency) sensors. The improved method leverages the strengths of each sensor type – displacement data for low frequencies and acceleration data for high frequencies – to produce a more accurate fused signal.
  • Fixed-Interval Smoothing: Applying a post-processing step to further refine the fused data by combining forward and backward Kalman filtering results.

The authors then use this fused data to establish a correlation model between wind speed and bridge girder displacement using linear regression.

Datasets

The study utilizes real-world monitoring data from the Anqing Yangtze River Bridge, a long-span cable-stayed bridge. Specifically, they analyze data collected throughout 2014, focusing on periods of strong wind action. The data includes:

  • Wind Speed: Measured by an ultrasonic anemometer.
  • Lateral Displacement: Measured by GPS sensors.
  • Acceleration: Measured by accelerometers.

The paper also uses simulated data generated from a swept-sine signal with a linear trend to initially validate the performance of the fusion method before applying it to real-world bridge data.

Key Insights and Ideas

  • Synergistic Fusion: The improved multi-rate Kalman fusion effectively combines the low-frequency accuracy of GPS displacement sensors with the high-frequency accuracy of accelerometers, overcoming the limitations of individual sensors.
  • Simultaneous Correction: Unlike traditional methods that might focus on one data type, this approach simultaneously corrects both displacement and acceleration, providing a more comprehensive and accurate representation of the bridge's dynamic response.
  • Frequency Domain Improvement: The fusion method not only enhances time-domain accuracy but also improves the frequency domain representation of displacement, aligning well with both GPS and accelerometer data across different frequency bands.
  • Enhanced Correlation Modeling: By using the fused data, the correlation between wind speed and bridge girder displacement becomes significantly stronger, leading to more reliable linear regression models.
  • Improved Performance Warning: The fused data leads to higher warning rates for structural degradation, enabling earlier and more accurate detection of damage.

Key Results

  • The improved multi-rate fusion method demonstrated a lower Root Mean Square Error (RMSE) compared to traditional multi-rate fusion for displacement, velocity, and acceleration. For instance, the RMSE for displacement after fusion was 0.062, compared to 0.142 for smoothed traditional fusion.
  • Correlation coefficients between lateral wind speed and displacement increased after fusion. For example, in Case 4, the coefficient rose from 0.715 before fusion to 0.768 after fusion, an increase of 7.37%.
  • The confidence interval width of the linear regression models for wind speed and displacement decreased after fusion, indicating more precise modeling. For Case 1, the confidence interval width decreased by 9.01%.
  • The performance warning capability was significantly enhanced. The warning rate for fused data was higher than for unfused data, and the method could detect damage of 30 mm with 100% accuracy in Case 7.

Limitations and Future Work

  • Limitations: The paper acknowledges that the linear drift term in acceleration is undetectable by acceleration sensors alone, highlighting the necessity of fusion. While the method is shown to be effective, the specific impact of different types of noise beyond Gaussian white noise is not extensively detailed.
  • Future Work: The authors suggest that the improved fusion method can be better applied to monitoring data of bridges, implying potential for broader applications in structural health monitoring. Further research could explore the application of this method to other types of structures or environmental loads. The paper also hints at using the fused data for system identification and damage identification, which could be areas for future exploration.

Summary updated

BibTeX
@misc{wang2021accurate,
  author = {Yan Wang and Dong-Hui Yang and Ting-Hua Yi},
  title = {{Accurate Correlation Modeling between Wind Speed and Bridge Girder Displacement Based on a Multi-Rate Fusion Method}},
  year = {2021},
  doi = {10.3390/s21061967},
}
Figure