基于统计学习的试验测量数据筛选算法研究

Research on Test Data Screening Algorithm Based on Statistical Learning

全机结构静力试验的应变/位移测量规模庞大,试验中常常会出现因应变片失效、位移传感器干涉等问题导致的无效应变/位移数据,要求对无效数据及时筛选并剔除.为后续试验分析提供完整有效的测量数据库,进而提高测量效率。现有方法主要依据经验进行人工观察筛选,效率低且易出现疏漏。本文首先建立数据筛选算法数据库,提出了一种基于统计学习的数据筛选算法,并进行了对比验证试验,结果表明.该方法可有效提高测量数据筛选的准确性和效率.为测量数据分析的自动化和程序化提供初步实验结果,为后续海量应变/位移数据自动化处理软件的开发提供理论依据。

Full-scale Aircraft Structure Static Tests have a large amount of measurement data,but some data could be useless because of the problems of strain gauge in actual tests.Invalid data is required to be screened and eliminated in a timely manner to provide a complete and effective measurement database for subsequent test analysis.which can increase measurement efficiency.Nowadays,data analysis is done by observation,which is low efficient and error prone.In this paper.we built a dataset for data analysis based on existing measurement data,then a data analysis algorithm is proposed and tested on the built dataset.Results of tests show the proposed algorithm can effectively improve the accuracy and efficiency of measurement data analysis.It provides preliminary experimental results for the automation and programmatic analysis of measurement data,and provides the theoretical basis for the development of automatic processing software for massive straindata.