Research On Coal Floor Water Inrush Prediction Based On Adaptive Robust Evidence Theory

Coal acts as the one of the foundation energies in our country, and occupies an important position in national economy. On the other hand, our country is also one of the countries suffering from the coal mine water disaster seriously, and the safety production of coal industry is indeed threatened. In recent years, with the increase of mining depth, the head pressure of aquifer is higher than before, the mining method under the pressure mining with the threat of backplane limestone confined water has become one of the main current, it makes the coal floor water inrush disaster become the most harmful and widest influence in form. Because coalmine hydrogeology conditions in our country are special, and the factors which caused the coal floor inrush are too more and complex. Many traditional coal floor water inrush prediction methods are unable to achieve ideal forecast effect in practical applications.Nowadays, information fusion technology, expanding in many application fields constantly and extending to water inrush prediction technology gradually, has a rapid development. The coal floor prediction theory based on multi-source information fusion, uses the multi-source information fusion technology to excavate fully the information connotation associated with the water inrush and their correlation, and makes a comprehensive analysis and utilization. The theory is expected to improve the accuracy, validity and reliability of water inrush prediction. So the information fusion theory in the mine water inrush prediction has high theoretical advantage and application prospect.D-S evidence theory, a classical information fusion technology, has an obvious advantage to distinct the "uncertainty" and "unaware" information, as well as deals with uncertainties caused by "unaware". It embodies the strong ability for dealing with uncertain information. Thus it has been widely used in the field of prediction. However, there are many defects such as poor robustness, fusion poor results when the conflict probability of evidence source is too large. This paper studies deeply the D-S evidence theory and its several improved forms, and analyzes their advantages and disadvantages. On this basis, this paper proposes a new theory of evidence-adaptive robust evidence theory. Adaptive robust evidence theory divides the synthetic formula into conjunction and disjunction. In the process of synthesis, it introduces conjunction weights function, the disjunction weights function and the concept of adaptive function. It distributes again the conflict probability and solvesthe poor fusion results which the conflict probability is too largeon the D-S evidence theory. Its robustness synthesis results are more great.In this paper, a prediction model for coal floor water inrush is established on the basis of the adaptive robust evidence theory and the experience of multiple neural network instead the traditional assignment as a source of evidence of the basic probability assignment way. After that, the model is tested by the measured data of oblique ditch coal mine, to testify the accuracy of water inrush prediction model, by comparing with the prediction results of water inrush coefficient method. Analysis and comparison of the results show that the water inrush based on the theory of the adaptive robust evidence is better than the traditional forecasting model in a higher accuracy and better stability. The model can be used to effectively judge whether coal floor water inrush will happen.