关键词: 多元线性回归, Bernstein多项式, RLS算法, 大坝安全监测

Abstract: The conventional Multiple Linear Regression model（MLR）is widely used in dam safety monitoring, but there is chaotic phenomena in residual series in this model, and the residual analysis is not implemented, which will affect the prediction accuracy. In this paper, a new MLR model combined with Bernstein polynomial was put forward by using RLS adaptive prediction algorithm that is based on Bernstein polynomial theory, namely it separates different influential components by MLR model, then uses RLS adaptive prediction algorithm to deal with the residual time series. Moreover, the model was validated by the observation data of horizontal displacement of dam crest of Si'nanjiang Hydropower Station. The validation results show that this model has higher prediction accuracy and can well reflect the actual value fluctuation.  

Key words: multiple linear regression, Bernstein polynomial, RLS algorithm, dam safety monitoring