基于AR模型和因子分析的结构损伤预警研究

摘要

Structural damage causes changes of dynamic response feature parameters. The variation of environmental factors (temperature,humidity)makes dynamic response feature parameters change too,they even mask subtle changes of dynamic response feature parameters caused by damage to lead to false structural damage detection. Here,to eliminate the influences of temperature,the AR model of time series analysis and factor analysis of multivariate statistical analysis combined with statistical control charts were employed for structural damage early warning.Firstly,the AR model was used to model structural acceleration response data before and after damage,and the model coefficients were extracted.Secondly,the factor analysis was employed to remove the influences of temperature on the AR model coefficients.Finally,the standard deviation control charts were used to predict damage.The results of numerical simulations showed that the proposed method can be used to successfully perform the damage early warning of a four-floor steel frame under temperature variations.%结构损伤会引起动力响应特征参数的变化，而环境因素（温度、湿度）的改变同样也会导致动力响应特征参数发生变化，有时甚至会淹没损伤引起的动力响应特征参数的变化，从而导致基于振动的结构损伤识别方法失效。为消除环境因素（温度）的影响，利用时间序列分析中的AR模型和多元统计分析中的因子分析并结合标准差控制图进行结构的损伤预警研究。首先利用AR模型对结构损伤前后的加速度响应数据进行动态建模，并提取模型系数，其次利用因子分析去除温度对AR模型系数的影响，最后利用标准差控制图进行损伤预警。数值模拟结果表明，在温度变化的条件下，该方法成功地对四层钢框架结构进行了损伤预警。