A novel method is presented for estimating the bridge surface roughness scanned by a single-axle dual-wheeled 3D test vehicle and processed by an augmented Kalman filter (AKF). Two acceleration sensors are installed atop the axle near the two wheels of the vehicle to measure its vertical and rocking motions. Meanwhile, the Kalman filter algorithm is augmented specially for the vehicle-bridge interaction (VBI) system, allowing the bridge surface roughness to be treated as the only unknown in the state-space formulation. To meet the invertibility criterion for resolving the dynamic VBI problems using the AKF, the observation vector is restructured by consolidating the accelerations recorded for the two wheels and their derivative displacements. The effectiveness of the present method was validated by the finite element method and demonstrated in a parametric study encompassing various system properties. In addition, a self-made, single-axle, dual-wheeled test vehicle was adopted in the field test to verify the theory presented. The reliability of the present technique was confirmed by its application to a real three-span continuous concrete girder bridge. The results indicate that the present technique is suitable for detecting bridge surface roughness of all levels with low sensitivity to noise interference and vehicle damping. Moreover, the surface elevations identified along the traces of the left and right wheels of the moving vehicle are “spatial” in nature. For practical application, it is recommended that the vehicle operates at speeds not exceeding 12 m/s to keep errors below 2 %.

中文翻译：

---

双轮 3D 测试车扫描并由增强卡尔曼滤波器处理的桥梁粗糙度：理论与应用


提出了一种新方法，用于估计由单轴双轮 3D 测试车扫描并由增强卡尔曼滤波 （AKF） 处理的桥梁表面粗糙度。两个加速度传感器安装在车轴顶部，靠近车辆的两个车轮，用于测量其垂直和摇摆运动。同时，卡尔曼滤波算法专门针对车桥交互 （VBI） 系统进行了增强，允许将桥梁表面粗糙度视为状态空间公式中唯一的未知数。为了满足使用 AKF 求解动态 VBI 问题的可逆性准则，通过合并两个车轮记录的加速度及其导数位移来重组观测向量。该方法的有效性通过有限元方法进行了验证，并在包含各种系统特性的参数研究中得到了证明。此外，现场测试采用了自制的单轴双轮测试车，以验证所提出的理论。该技术的可靠性通过应用于真正的三跨连续混凝土梁桥得到证实。结果表明，该技术适用于检测各级桥梁表面粗糙度，对噪声干扰和车辆阻尼敏感度低。此外，沿着移动车辆的左右车轮的痕迹识别的表面高程本质上是“空间”的。对于实际应用，建议车辆以不超过 12 m/s 的速度运行，以将误差保持在 2% 以下。