论文部分内容阅读
大坝观测数据经常规回归分析后的残差序列一般并非为白噪声 .考虑将回归拟合与随机型时间序列方法结合 ,先对大坝位移数据按水位、温度、时效等物理因素作回归分析 ,再对回归残差作时序列建模处理 .实例采用Box- Jenkins法和由自相关、偏自相关函数及 AIC准则进行模型识别 ,建立时序列模型 .应用示例的计算表明 ,这样获得的回归 -时序列模型能很好拟合实测数据 ,提高精度 ,误差序列也符合白噪声要求 .
Dam observation data after routine regression analysis of the residual sequence is not generally white noise. Considering the combination of regression and random time series method, the first displacement of the dam data by water level, temperature, aging and other physical factors for regression analysis , And then model the regression residuals for time series modeling.Examples using the Box Jenkins method and the autocorrelation, partial autocorrelation function and AIC criteria for model identification, the establishment of time series model.Application example calculation shows that the obtained regression - Time series model can well fit the measured data to improve the accuracy, the error sequence also meets the requirements of white noise.