Infiltrates The Flow Field Inversion Research Ased On The BP Neural Network's Earth Stone Dam

The earth dam, the base of dam soil either the rock mass, both banks soil body or the rock mass penetration coefficient is analyzes the earth dam to infiltrate the flow field a key parameter. It relates the people whether can understand the dam truly the actual safe condition. These design variable tests through the laboratory experiment or the open country scene comes, as a result of many kinds of reason's influences, the actual value which and the actual situation the laboratory experiment will arrive will possibly have the big discrepancy; But the scene experiment has the varying degree the data separate, the representation not strong perhaps the cost expensive and so on shortcomings. Will therefore monitor and the inversion analysis on the spot unifies the inversion penetration coefficient, had the obvious reality to affect. This article from this angle embarking, in view of between monitoring data's in measuring tube flood peak and penetration coefficient's misalignment relations, introduces the BP nerve network method, the genetic algorithm, uses in the inversion earth stone dam penetration coefficient, the contrast reference actual design penetration coefficient.(1) uses between the penetration coefficient and the measuring tube flood peak misalignment relations, unifies the BP neural network algorithm the characteristic inversion 3 earth dam project example, monitors under the different water level measuring tube flood peak which safely the dam comes, as well as the actual design penetration coefficient, obtains the corresponding forecast seepage coefficient value through the training learning network, and compares mutually it with the original actual design penetration coefficient, should prove the BP neural network algorithm inversion earth dam to infiltrate the flow field the feasibility.(2) easy to fall into the partial optimum value in view of the BP algorithm, the restraining time long malpractice, we use the genetic algorithm to carry on the optimization to it, the BP neural network are quite sensitive to the network initial weight and the threshold value, therefore how to determine that the initial weight and the threshold value were the global optimum become the key to be. We carry on the optimization to the network weight and the threshold value, after will optimize before weight and threshold value counter-substitution, trains the good BP neural network, obtains the new forecast seepage coefficient value. These project example proved that after the optimized forecast seepage coefficient value must the forecast night-watch close original design parameter which obtains compared to the BP algorithm inversion. This indicated optimizes after the genetic algorithm the BP neural network inversion earth dam infiltrates the flow field is feasible.