基于自适应卡尔曼滤波加速度与位移融合的结构位移实时估计

实时高精度位移测量在工程结构的安全和寿命评估方面有着重要作用。为提高基于全球导航卫星系统技术的位移测量的精度及稳定性，本文提出了一种融合加速度和位移数据的自适应多速率卡尔曼滤波方法，来实时获取精度提升的位移信息。由于不合理的噪声参数设置会使位移估计的精度严重下降，本文利用加速度和位移数据测量噪声各自的特点，以分开估计相应噪声方差的思路来实现自适应估计。考虑传感器噪声的性质，自适应滤波中对噪声参数的估计可简化为仅对位移噪声方差进行估计。利用Sage-Husa估计器实现位移噪声方差的自适应估计，使滤波能在噪声参数未准确获知的情况下进行稳定的位移实时估计。首先，讨论了自适应滤波中初始噪声参数的影响，确定了初始系统噪声参数的选取原则。然后，分别在时不变与时变位移噪声环境下，观察该滤波应用于不同频率的谐波位移信息下的估计性能。最后，以某1.5MW风电塔在风-地震耦合作用下塔顶结构响应的数值模拟，来说明本文的自适应滤波在一般工程结构应用中的有效性。结果表明，即使初始噪声参数设置有误或位移噪声具有时变性，本文方法依然具有较好的估计效果及鲁棒性。上述研究成果可为结构实时高精度位移监测提供一定理论支撑与参考。

Real-time Structural Displacement Estimation by Fusing Acceleration and Displacement Data with Adaptive Kalman Filter

Real-time displacement measurement with high-precision is quite important for the safety and life-cycle assessment of engineering structures. To improve the accuracy and stability of displacement measurement based on Global Navigation Satellite System (GNSS) technology, an adaptive multi-rate Kalman filter is proposed to fuse the acceleration and displacement. Due to unreasonable settings of noise parameters, the accuracy of displacement estimation will be seriously degraded. By utilizing the characteristics of acceleration and displacement measurement noises, adaptive estimation is realized through estimating the variance of their corresponding noises separately. Considering the noise characteristics of accelerometer and GNSS device, the estimation of noise parameters in the adaptive filter is simplified to estimate only the variance of displacement noise. The Sage-Husa estimator is used to realize the adaptive estimation of displacement noise variance so that the filter can reach a stable real-time displacement estimation under inaccurate noise parameters. First, the settings of initial noise parameters in the proposed adaptive filter are discussed to determine its rule. Then the displacement estimation performance of the filter at different signal frequencies is discussed through the harmonic displacement under time-invariant noise and time-varying noise. Finally, the effectiveness of the proposed technique is demonstrated using a numerical simulation response from a 1.5MW wind turbine tower under wind-earthquake coupling. Results show that even if the initial noise parameters are inaccurate and the displacement measurement noise is time-varying, the proposed technique still has satisfactory performance and robustness in real-time estimation. This research can provide certain theoretical reference for real-time and high-precision displacement monitoring of structures.