TL;DR
This paper introduces a novel two-step method to accurately measure bridge surface roughness using the response of a passing two-axle vehicle. By employing an Extended Kalman Filter with Unknown Inputs (EKF-UI) and a contact residual analysis, the technique effectively separates bridge deflection from vehicle response, overcoming a key challenge in existing methods. This advancement offers a more reliable and efficient way to assess bridge health and pavement conditions.
What Problem Does This Paper Address?
Measuring road and bridge surface roughness is crucial for infrastructure maintenance and understanding vehicle-bridge interaction (VBI) dynamics. Traditional methods, like manual profilographs or laser profilers, are often time-consuming, expensive, and not ideal for continuous monitoring. The "vehicle scanning method" (VSM) offers a more practical alternative by using a moving vehicle's response. However, a significant challenge arises when scanning bridges: the bridge's own dynamic response (deflection) caused by the vehicle contaminates the vehicle's measured response. This makes it difficult to isolate the true surface roughness. Existing VSM techniques struggle to fully eliminate this bridge deflection, leading to inaccuracies in roughness estimation. This paper aims to address this by developing a method that can reliably extract bridge surface roughness even in the presence of this coupled, non-stationary VBI problem.
Methods
The proposed solution is a two-step technique:
- Estimating Unknown Inputs with EKF-UI: The core of the first step is an adaptation of the Extended Kalman Filter with Unknown Inputs (EKF-UI) algorithm. This algorithm is designed to estimate unknown inputs to a system based on its measured outputs. In this context, the EKF-UI is used to process the noise-contaminated response of the two-axle test vehicle. It estimates two key components: the vehicle's improved (noise-reduced) states and the "unknown input" vector. This unknown input comprises both the bridge surface roughness and the bridge deflection caused by the vehicle.
- Retrieving Surface Roughness via Contact Residuals: The second step focuses on separating the roughness from the estimated unknown input. This is achieved by first calculating a "roughness-free" contact residual response. This residual is derived from the improved vehicle states obtained in Step 1. By analyzing the correlation between these residuals and the vehicle's contact displacements (which are also estimated), the paper derives coefficients that allow for the calculation of the bridge's contact displacements. Crucially, these contact displacements are independent of the bridge's flexural rigidity. Finally, the estimated bridge surface roughness is obtained by subtracting these calculated contact displacements from the total unknown input (roughness + deflection) estimated in Step 1.
Datasets
Not discussed in the paper. The study relies on numerical simulations and theoretical validation rather than real-world datasets.
Key Insights and Ideas
- Decoupling Vehicle and Bridge Dynamics: The most significant insight is the successful decoupling of the vehicle's response from the bridge's dynamic deflection. This is achieved by treating the combined roughness and deflection as an "unknown input" and then using a clever residual analysis to isolate the roughness.
- EKF-UI as a Powerful Tool: The paper demonstrates the effectiveness of the EKF-UI algorithm in handling complex, non-stationary VBI systems where multiple unknown inputs are present.
- Contact Residuals as a Key Indicator: The concept of using "roughness-free" contact residuals to estimate bridge displacements is a novel approach that bypasses the need for prior knowledge of bridge properties like flexural rigidity.
- Robustness to Vehicle Parameters: The study suggests that the method is relatively robust to variations in vehicle mass, which is a practical consideration for real-world applications.
Key Results
- High Accuracy in Simulation: Numerical validation shows that the retrieved bridge surface profiles are in excellent agreement with the original (assumed) profiles, both in the spatial and frequency domains, across different roughness classes.
- Noise Reduction Effectiveness: The EKF-UI algorithm effectively removes measurement noise from the vehicle's responses, as evidenced by the close match between retrieved and original (noise-free) vehicle states.
- Speed Sensitivity: Estimation accuracy decreases slightly with increasing vehicle speed, but remains acceptable (errors < 5%) for speeds up to 16 m/s.
- Noise Sensitivity: The method is sensitive to environmental noise; higher noise levels lead to reduced accuracy, with RMSE reaching 14% at a 10% noise level.
- Bridge Damping Impact: Increased bridge damping generally improves estimation accuracy, with RMSEs decreasing as damping increases.
Limitations and Future Work
- Theoretical Study: The paper presents a theoretical study and numerical validation. Field testing and validation on actual bridges are necessary to confirm its practical applicability.
- Noise Level: The method's sensitivity to environmental noise suggests that it is best applied in relatively quiet conditions. Future work could explore more advanced noise filtering techniques.
- Vehicle Speed: While acceptable, accuracy decreases at higher speeds. Further research could investigate methods to maintain accuracy at higher speeds.
- Multi-Axle Extension: The paper mentions that the approach can be extended to multi-axle vehicles, but the current study focuses on a two-axle vehicle.
- Bridge Types: The current validation is for a simple beam bridge. Applying the technique to bridges of different types (e.g., suspension bridges) is a logical next step.
Summary updated