Dissertation > Excellent graduate degree dissertation topics show

The Study on the Surrogate Model of the Numerical Simulation Model of Groundwater

Author: YiYanPing
Tutor: LuWenXi
School: Jilin University
Course: Hydrology and Water Resources
Keywords: Groundwater numerical simulation model Alternative model RBF neural network Sampling methods Jinquan Industrial Park
CLC: P641
Type: Master's thesis
Year: 2011
Downloads: 191
Quote: 0
Read: Download Dissertation

Abstract

Proved by theoretical research and engineering practice, the simulation model can only solve the problem of forecasting and simulation optimization to resolve the problem of groundwater remediation process of decision-making program preferred. Therefore, increasing attention has been widespread attention numerical simulation-based optimization model. But the optimization model must be based on a simulation model for the optimization model optimization process must follow the inherent principles and laws of the groundwater system (by the simulation model to express) on the basis of. Hence the need for some way and means embedded in the simulation model to the optimization model, make it a part of optimization model, the simulation model and optimization model connection. Embedding method applied in the past, the response matrix method, the state transition equation method is used to solve the problem of how embedded in the optimization model and call simulation model, but they have their own scope of use is often limited. Alternative model method is an effective way to connect simulation models and optimization models. Create function can approach the approximate alternative model of groundwater flow numerical simulation model, optimization model iterative solution process can be called directly alternative model (ie, the alternative model to solve), without having to simulation model. This will not only be able to overcome the limitations of the method of coupling technology in the past, and can greatly reduce the optimization model for solving computational load calculation process caused by the simulation model is called to save a lot of time. Therefore, an alternative model approach - kind of useful to explore the potential and promising way to resolve its research and application will be increasing people's attention and recognition. The alternative model quality is good or bad depends on two key aspects of the modeling process, that is the kind of sampling methods and alternative models selected. Selection of an appropriate sampling methods combined simulation model input (pumping capacity) - Output (drawdown) sample data set, the basic premise is to create an alternate model. Jinquan Industrial Park water as a study area on the basis of the establishment of groundwater numerical simulation model, the application of Monte Carlo sampling and Latin Hypercube Sampling, respectively, in the region of five pumping wells pumping the extent practicable sampled input data set (pumping capacity), so that both meet the practical needs but also improve work efficiency. In the case of extracting the same number of samples, the sampling results of the comparative analysis of the two sampling methods Latin Hypercube Sampling data more representative. Selection of Latin hypercube sampling results cubic pumping test program data, and as a set of input data into the groundwater numerical simulation model, calculated from 10 observation wells drawdown deep data sets, the output data set. On the basis of input and output data sets, respectively, using regression analysis, artificial neural network, the establishment of a dual response surface model, an alternative model of the radial basis function (RBF) neural network model both groundwater numerical simulation model, and alternative The validity of the model can be tested, results showed that: the two alternative models of the results obtained with the simulation model fits these results are good, two alternative models can function approximation groundwater numerical simulation model. Finally, for dual response surface model, the application process of RBF neural network model of two alternative models, results were analyzed. The results showed that, compared with the dual response surface model, RBF neural network model advantages in terms of number of samples, and the results of the simulation model to calculate fitting accuracy, expression, better application prospect, indicating that the two alternative model RBF neural network model is more suitable for the the approximate alternative model of groundwater numerical simulation model.

Related Dissertations

CLC: > Astronomy,Earth Sciences > Geology > Hydrogeology and Engineering Geology > Hydrogeology (groundwater hydrology )
© 2012 www.DissertationTopic.Net Mobile