神经网络及数值模拟技术在热熔钻中的应用

摘要

为了能准确把握影响热熔钻钻进的各个主要因素,揭示各因素的影响规律以及热熔钻钻进时周围地层温度场的变化情况和影响范围,首次将神经网络应用到热熔钻领域的研究中,采用神经网络及数值模拟技术分别对实验数据进行预测拟合及仿真分析.运用神经网络能够对热熔钻的实验数据进行很好的预测拟合,数值模拟可模拟地层温度场情况,利用有限元软件生成地层的温度场云图及曲线,使温度场的变化更为直观,便于分析研究.将训练好的网络模型保存下来,在下次实验中只需测量少数几个值将其导入训练好的网络模型,便可得出一组对应的数据,从而节省实验成本,提高实验效率.该模型可作为对应结论的预测模型使用,为热熔钻的研究提供了新思路.%In order to accurately grasp the various major factors of impact hot melt drilling to drill, the law of the various factors and the change of the temperature about the surrounding strata and the extent of it should be explore. Its the first application of neural network in the field of the hot melt drilling, and the neural network and numerical simulation to predict and simulation analysis about experimental data were used. The BP neural network can predict the experimental data fit well on hot melt drilling and numerical simulation can simulate the formation temperature conditions. The FEM software can generate the temperature distribution field and curved lines, and it makes the temperature field more intuitive and convenient for analytical study. To save the cost of experiments and improve the experimental efficiency, the trained network models can be saved, only a few values into the trained network model of it to arrive at a set of corresponding data need measured in the next experiment This model can be used as the prediction model for the corresponding conclusion. It provides a new idea for researching on the hot melt drilling.