Research And Application Of Drilling Design Algorithm Based On Neural Network

In recent years,with the rapid development of the Internet and the rise of interdisciplinary,many geological scientists have used computers to simulate the design and construction process of boreholes in in-situ immersion wells.In the construction of in-situ leaching wells,in-situ leaching drilling(hereinafter referred to as drilling)is the only channel that can extract uranium resources from rock strata,so only reasonable drilling design can ensure the stable operation of uranium mining in wells.But in the actual development of drilling,there are many processes and factors involved in the judgment of drilling and the arrangement of drilling filter.In order to solve these problems,a drilling construction system based on in-situ leaching well site is designed and developed to solve the problems of drilling qualification judgment and filter arrangement,so as to improve the efficiency of in-situ mining.In order to improve the accuracy of drilling judgment,this paper designs a model based on stack self-encoder network and gradient lifting tree(GBDT-SAE).This model normalizes the borehole data in the in-situ leaching well field,and then encodes the borehole data through the multi-layer stack self-encoder to obtain the drilling data after dimension reduction of the hidden layer.Then,the data are trained by the gradient lifting tree model to obtain the model for determining and classifying the drilling.The experiment was designed and the results were compared with those of K-Means,Adaboost,GBDT,KMeans-SAE and Adaboost-SAE.The experimental results show that GBDT-SAE has higher iterative rate and accuracy in the judgment of drilling qualification,which verifies the feasibility and correctness of GBDT-SAE algorithm in the judgment and classification of drilling qualification.In order to improve the accuracy of borehole filter layout,this paper proposes a model(EMD-AMGRU)based on empirical mode decomposition,attention mechanism and gate control loop unit network combination to achieve borehole filter layout.Empirical mode decomposition method can deal with many non-stationary factors such as lithology,grade,geological strata,permeability and zenith angle in borehole data,and these non-stationary sequences are adaptively decomposed into stationary sequences.AMGRU combines the attention mechanism in the gate control loop unit network to strengthen the network memory and the concentration of the target value,so as to improve the accuracy of the model.The experimental results show that compared with other models,the error of EMD-AMGRU prediction is the lowest,indicating that the prediction results are closer to the real value.The feasibility and correctness of EMD-AMGRU algorithm in the arrangement of drilling filters are verified.Finally,the proposed algorithm is empirically analyzed in the system.The GBDTSAE and EMD-AMGRU models are applied to the functional modules of drilling qualification judgment and filter batch layout in the drilling construction system,respectively.The accuracy and feasibility are verified and the expected results are achieved.It is verified that the method has strong applicability for the processing of drilling data.At present,the development of the system is basically completed,and the functional test of the practical application of in-situ leaching well field is being carried out.