| Based on analysing the characteristics in dam observation data treatment, Some problems, such as , correlation effect of independent variables, selection of the model variables and on—line analysis, are considered in this paper.1. Generalized weighted regression method is proposed for parameter estimation, and it bas been proved that this method can not improve the parameter estimation of the model. It is also proved that the Grey System model GM (o, m) and the differential successive regression method both are special cases of the generalized weighted regression method, and therefore they can not improve the estimation of the model parameter. Further, it is proved that the differential successive regression method can not treat problem of close correlation of independent variables effectively in regression analysis.2. The effect of the close correlation of independent variables on regrres-sion estimation of model parameter is analysied in this paper. Some formula of the mean sqarre error (MSE) and the relative MSE of every variables are first given by the author. On the other hand, the so—called "Theore lically constrained regression method" is proposed for analysis of dam observation data.3. The effect of selection of the model variables on regression analysis is analysied in this paper. Especially, the effects of the omittive variables and the combinced variables are investigated. Further more, in view of periodical character of dam observation data, a frequency spectrum analysis method for data treatment is suggested.4. This paper proposes an on—line analysis method of the dam observation data, introducing a time—varying model and two recursive estimation methods with fogetting factors and/or with variable fogetting factors.5. Based on the on—line analysis ,two monitorring methods are suggested.6. According to the methods proposed in this paper, the displacement observation data of Shen—Wo dam is analysied.
|
PDF Full Text Request