The warning model of the early kick based on BP neural network

For the matter of low correct rate in early warning of kick,this paper proposes a method of early warning of kick based on bp neural network,this method select the appropriate network training samples and test samples according to the data of adjacent well,through the network training and network testing,different layers of warning network can be get,which are weights and threshold matrix; then put the actual drilling data into the warning model for training,obtain the training results and build training result vector; finally,based on the principle of maximum similarity,compare the calculated vector with the desired output vector,and the output that most similar to the desired result is the warning result. statistical experiments show that the method of early warning accuracy rate is higher,and the convergence speed is fast,which can guide the construction site. © 2014 pakistan journal of statistics.