Inversion Analysis Of Material Parameters And Deformation Prediction For Large Aqueduct

As the safety problems of large-scale aqueducts become increasingly prominent,in order to accurately grasp the operating status of large-scale aqueducts,it is necessary to conduct theoretical research on the analysis methods of its observational data.This paper relies on the water-filling test data of the Kashi Kezi River in Xinjiang to carry out research from two aspects,namely,the inversion analysis of local structural material parameters and the deformation prediction analysis of the overall structure.The stresses on key structures such as the bottom longitudinal beams and side walls of large aqueducts are complex,and the aqueducts generally have problems such as cracks.The conventional inversion results have large deviations from the actual values.Therefore,using the deterministic model method combined with the finite element deformation analysis theory,the material parameters of the mid-span parts of the longitudinal beams of the aqueduct were inversely analyzed,and the influence of each component on the deformation of the aqueduct under the full load condition was discussed;In view of the common multicollinearity and poor extension of the aqueduct’s multi-point deformation prediction model,the deformation prediction model was established by a combination of partial least squares regression and BP neural network,which was verified by comparison with the calculation results of the conventional model.Model fitting and deformation prediction capabilities,provide theoretical basis for observational data analysis of actual engineering.The main research results of this paper are as follows:(1)Starting from the theory of finite element deformation analysis,various forms and assumptions of the water pressure,temperature and aging components of the monitoring model are analyzed and discussed,and a deterministic model of the aqueduct deformation observations is constructed.For the temperature component,when the monitoring data of the aqueduct is missing and short,it is proposed to calculate the temperature field with air temperature and water temperature data,and to construct the internal and external temperature field of the aqueduct with the transient temperature field;for the time-dependent component,the viscoelastic model is used to construct the constitutive relationship of the aqueduct It is more realistic to divide the aging component into two parts: elastic deformation and attenuation creep.(2)The calculation results obtained by conventional inversion methods for large aqueducts are usually difficult to guarantee accuracy.For this reason,the deterministic model method is used for material parameter inversion analysis.This paper is based on the inversion analysis method of aqueduct deterministic model,the physical meaning of the adjustment coefficients in the deterministic model was discussed and the time-effect component was appropriately revised;on this basis,the aqueduct single-point deterministic model was established,and Based on the field observation data of Xinjiang Kezi River Aqueduct Project,the parameter inversion analysis was carried out.The calculation results show that fitting different measuring points,the multiple correlation coefficients are all greater than 0.9,the analysis model itself has a high degree of fit;Comparing the calculated elastic modulus with the design data,it is found that the long-term operation has caused a certain degree of aging of the aqueduct,and the relative decrease of each measuring point is 15.5%,10.1%,and 6.1%;Analyzing the proportion of the components under the full load condition of the aqueduct,it is found that the water pressure component is the main influencing factor that causes the deformation of the aqueduct,and the influence of the three components on the vertical displacement of the tank body is: water pressure>temperature>aging.(3)The deformation prediction analysis is carried out by using partial least square regression-BP neural network model.In order to solve the common problems of multicollinearity and poor prediction ability in conventional multi-point deformation prediction models,a combination of partial least squares regression and BP neural network is proposed to jointly establish aqueduct prediction models.The analysis model takes the Kezi River Aqueduct as the research object,a comparative analysis of the model’s fitting ability and forecasting ability was carried out,the multi-correlation coefficient in the training stage of the model is0.912,and the mean square error in the prediction stage is 0.151.Compared with the conventional statistical model and the BP neural network model,the calculation results have better fit and short-term forecasting ability.