Application Of ANN For The Prediction Of Water Content In Geophysical Exploration

According to the statistics, since the founding of New China, Chinese underground water resources development and utilization of the amount is increasing rapidly. Only sporadic mining in the 20th century 50s, increasing to 57 billion cubic meters annually in 70s, and increasing to 75 billion cubic meters annually at present groundwater-exploitation has exceeded 100 billion cubic meters annually in 80s, accounting on the 1/5 of country's total water volume, but the exploitation of the country still has great potential for groundwater. For a long time, groundwater resource evaluation by the use of labor intensive and time-consuming and costly large-scale pumping test method, its problems are:First, for a long time a large number of abstraction of groundwater wasted; Second, working inefficient. However, with the scientific and technological progressing, the current geophysical technology has been developed enough to meet the requirements of hydrographic surveys, in order to change the traditional method to predict groundwater levels, using an integrated geophysical technology forecasting aquifer water content of the above problems can be avoided, in order to carry out water resource evaluation to provide a new technological means.Although the research of using geophysical methods to predict the underground water content has been carried out years, such as in 1992,Hebei geophysical exploration team adopted frequency sounding in the three northern governorates of Yemen Republic of the Gobi area to find water drilling project to find biological Jurassic fractured limestone aquifer structure, after well finished sunrised water 400t; another high artesian 300ms' deep wells, discharge date up to 1100t. Another example is the Department of Geology and Mineral methods Hydrogeology and Engineering Geology Institute of Technology to conduct geophysical exploration work to find water with a high success rate of wells in 1991 in Nigeria. There also was another case of Hebei Hydrological drilling work teams in the Somali foreign aid to fight for town water supply wells, geophysical and drilling of wells accuracy rate is very high; even in very complex geological conditions in pastoral areas, as well rate achieved satisfactory results. Various of geophysical examples of finding water shows that geophysical methods for hydrogeological has been recognized by the insiders, but still could not escape the problem of a large number of wells for pumping The research project is to solve the problem associated with conventional hydro-geophysical exploration work costs of water pumping test methods, time-consun laborious and other problems, try to make it easily able to solve groundwater wells on the evaluation.Traditional hydrology and water resources forecasting based on linear regression analysis, the prophet-type function of nonlinear analysis, and so on, the structure of these models are too simple and can not escape the scope of statistics, depend on the large number of samples makes prediction accuracy is difficult to guarantee. The nonlinear prediction based on artificial neural network technology in the development of geological modeling applications has been more mature, but the neural network used for the quaternary aquifer water content forecasting is still a blank in their applications. Therefore, this research through the quaternary pore aquifer lithology, the analysis of hydrogeological conditions, to carry out new technology new method research of ground geophysical techniques to identify the fine structure of the aquifer and hydrogeologic parameters of the distribution, to establish an integrated approach to the optimization of hydro-geophysical exploration model; to overcome the existing geophysical parameters of the water content based on lack of forecast models, namely:â‘ linear model instead of an objective non-linear relationship,â‘¡limited to function types of the Prophet nonlinear model,â‘¢built on statistical basis, with dependence on large sample statistics. Artificial neural network (ANN) technology to discover the geophysical methods to analyze a number of geophysical parameters, and a nonlinear aquifer water content of BP neural network prediction model, the model has good adaptability, showing a more robustness to promote and eventually set up a more porous aquifer system, water content of the comprehensive geophysical technology forecasting system for the evaluation of groundwater resources to provide a new practical skills.The result of this research is adopting neural network technology for the use of a number of known geophysical testing the well pumping test data to the geophysical parameters of aquifer discharge units with single-holc model composed of the training sample set, establishing the error BP Network (Back-Propagation Network, called BP network) forecasting model for regional geophysical methods unknown underground water content prediction, this will be a major improvement of the traditional water content prediction.The research relys on Ministry of Land and nonprofit industry special funds research project " Research of water layer moisture forecasted by the composite geophysical technology" the collection of a large number of previous forecast for the establishment of hydrological parameters of water model based on the work of data optimized combination of geophysical methods the relevant parameters of the work. The technology used routes are:first to obtain the corresponding geophysical geophysical work to filter the data consolidation; and then to choose the appropriate model parameters as input, using neural network prediction model of aquifer water content; Finally, some of the known area tested sample the predictive ability of models to promote the test in order to select the best model. Using ultimately forecast model on the unknown area,to make promotion of performance evaluation on the extend ability of the model. The research results is twofold:First, the various geophysical methods on the feasibility study and ultimately concluded effective geophysical methods combined mode; Second, the establishment of the aquifer water content of the nonlinear forecast model.Innovation of this research project are:â‘ established hydrogeological exploration optimization model by using integrated geophysical methods;â‘¡using BP artificial neural network under non-prophet nonlinear prediction model, compared with previous models to improve prediction the accuracy and generalization ability, to overcome the dependence on large sample statistics and the known conditions require excessive limitations.