一种提高时域参数识别精度的新方法

在结构损伤识别的时域法研究中,检测的加速度信息中包含结构特征信息和各种干扰信号,积分法计算速度和位移时也可能引入计算中带来的低频信号,使得输出信号严重失真,从而导致反演结构物理参数的失败.为了从测试信息中提取含有结构固有特征的信息,针对小波包分析能对信号频率多层次划分,具有多分辨、高精细的特点,设计一种适用性较强的信号滤波器.该滤波器可根据小波包分解层数,任意确定频率分辨率,选择不同小波包系数,确定滤波频带和频率范围选择.通过剪切型结构物理参数反演模型的算例分析,验证该滤波器具有较好地提取结构特征信息的性能.该滤波器可用于具有此类要求的信号处理.

WAVELET PACKET ANALYSIS BASED FILTERING TECHNIQUE FOR IMPROVING TIME DOMAIN IDENTIFICATION OF PHYSICAL PARAMETERS

In time domain identification of structural parameters the source signals picked up by acceleration sensors are often noise contaminated, from which the derived velocity or displacement signals may be further distorted through integration. Such a kind of seriously distorted input signals can worsen and even cause the failure of parameter identification. Hence, the first thing for improving parameter identification is to improve the quality of source signals. A filtering technique based on wavelet packet analysis is introduced. In connection with that wavelet packet analysis is featured with high precision and multi-level resolution, a widely applicable filter can be readily designed with specific resolution for different selections of filtering frequency band, so that the filtered source signal can meet the desired requirements. Numerical results of an illustrative example on parameter identification of a shear column model structure show the feasibility and effectiveness of this filtering technique.