ARTIFICIAL NEURAL NETWORK PREDICTION OF PARAMETERS OF NON-LINEAR SEEPAGE MODEL WITH PLANAR INTERSECTING FRACTURES

The fluid flow in rock fracture network exists widely in underground engineering,which has important influences on the safety of engineering construction and operation. Therefore,the study of flow in fracture network has important theoretical and practical significance. This paper is based on the flow model of intersecting fractures. The model is derived based on the cubic law and Forchheimer equation. It studies the nonlinear parameters of the flow model of planar intersecting fractures by numerical simulation and artificial neural network. Using the artificial neural network genetic algorithm,the paper explores the relationship between the geometry of the intersecting fractures and the nonlinear coefficient in the flow model. It proves that the nonlinear flow model of the plane intersecting fractures can adequately describe the flow characteristics of the intersecting fractures. It verifies the feasibility and accuracy of the neural network method in predicting the nonlinear coefficient. At the same time,it obtains the characteristics of the two methods of fitting numerical expression and artificial neural network.