基于探索性数据分析的桥梁模态识别数据处理方法

摘要

桥梁健康监测系统以传感器测得的结构响应识别桥梁结构的模态参数,从而对损伤进行预警.然而,由于测试环境的限制,传感器老旧、损坏等情况时常发生,使得测试结果受到影响.结合探索性数据分析(ExploratoryData Analysis,EDA)方法,采用箱形图和QQ图对测试结果进行了跟踪,对分析结果异常的数据对应的传感器进行了修复,并对比了修复前后的模态参数识别结果.结果表明,经过本文所提方法处理后的测试结果能提供更为准确的模态参数识别结果,提高桥梁健康监测的准确度和效率.%Bridge health monitoring system can help to identify the modal parameter from the structure response data collected from the sensors installed at these structures,thus a pre-warning about the damages in the structure can be issued on time.However,due to the complexity in the testing environment,there is always something wrong with the sensors,which eventually leads to false measurements.For this purpose,the demand for an accurate data evaluation method once the data is collected form these sensors has increased manifoldly.In this paper an efficient data interpretation method called Exploratory Data Analysis (EDA) is introduced.Based on this method,boxplots and QQ plots were adopted to track the inner trends in the measured data,and then based on these results,those sensors having abnormal data were overhauled.Finally,a comparison of the modal parameter identification result before and after the overhauling of sensors were presented.The result shows that an accurate modal parameter identification can be achieved after the adoption of EDA as a preprocessing technique.Thus in the end it can be said that,EDA has drastically improved the accuracy and efficiency of bridge health monitoring process.