大型公铁两用斜拉桥健康监测数据的时域峰值统计及频域融合处理

Peak Statistics in Time Domain and Fusion Processing in Frequency Domain for Health Monitoring Data of Large-Scale Highway-Railway Cable-Stayed Bridge

为实现基于健康监测数据的桥梁运营安全性评定,针对大型公铁两用斜拉桥健康监测信号的数据量大、干扰强、频谱复杂、信噪比低等特点,基于沪通铁路长江大桥的健康监测数据,建立监测数据质量状况评定的3项指标(信噪比,单测点异常数据占比和总测点异常数据占比)及其5个评判等级的区间评判体系,并据此进行监测数据降噪去异常值的初步处理;基于30~300s的时域分段及监测值峰值即时替代措施,给出健康监测数据时域分段峰值的提取方法及其均值和方差的统计方法;采用加权平均方法推导出基于快速傅里叶变换、小波变换、希尔伯特—黄变换的监测数据频域融合公式,以提高频域数据的可靠性及精度,获取有效开展桥梁运营安全性评判的评估参数。以监测到的某大型桥梁的振动加速度数据为例,采用给出的数据处理方法进行数据的去噪处理、时域峰值提取及统计、频域融合处理,验证了方法的工程实用性。

For the purpose to realize operation safety performance assessment based on health monitoring data, according to the special characteristics of health monitoring data of large-scale highway-railway cable- stayed bridge, such as huge number of monitoring data, strong interference, complex frequency spectrum, low signal-noise ratio and so on, based on the data analysis and processing of Shanghai-Nantong Yangtze River railway bridge health monitoring system, the method of assessment index was adopted to judge the quality of monitoring data with three indexes （signal noise ratio, single test point abnormal data ratio and total test points abnormal data ratio） and their according five assessment levels with assessment section system, then the data pretreatment including de-noising and abnormal value reduction was put forward. With the time domain segmentation based on 30~300 s and the immediate replacement of monitoring peak value, the method for extracting the peak value of time domain segmentation as well as the statistical method of mean and variance were put forward. The weighted mean method was adopted to derive the equation of frequency fusion based on FFT, WT and HHT to improve the reliability and precision of frequency domain parameters, so the assessment parameters obtained could be effectively used to assess the service performance of bridge. Taking an example of the acceleration monitoring data of one long span bridge, the monitoring data processing method was used for de-noising, time domain peak value extraction and statistics, frequency domain fusion processing. The practicability of the monitoring data processing was verified.