基于有限元的钻进参数相互影响机理研究

在实际的岩土钻进过程中会经常遇到软弱岩土层、破碎带及断层、岩溶、高地应力等不良地质,因此给 钻进工程带来一系列问题和灾害。 众多文献资料和试验数据证明,此类问题与钻进的运行参数以及岩土的力学参数 有着密不可分的关系。 由此,本项目以 PDC 复合片钻头为研究对象并进行仿真模拟分析,设置不同的岩体参数以及 不同的钻进参数,将在不同条件下所得的作业数据进行分析研究,揭示钻头在不同岩石中钻进参数表现的区别和在 不同岩体在相同钻进条件下轴压和扭矩的区别和联系。 分析发现岩石性质越好的其 Mises 等效应力就越高;在一定 范围内钻头轴压和扭矩均与钻速呈现出正相关的关系,钻压的提高增加了钻齿的切入深度与岩石切削作用,扭矩继 而随之增加;在相同钻速条件下转速的提高并没有对轴压产生较明显的影响,扭矩则呈现正相关变化。 最后通过 Python 语言建立人工神经网络,经过机器学习对数值模拟数据和实际钻孔数据进行训练,验证利用钻进参数和机器学习 的方法来实现对岩石种类的判识具有可实现性,最终结果表明对模拟数据训练预测的准确率达到 90%左右,实际数 据的准确率最高达到 78%。

Study on Interaction Mechanism of Drilling Parameters Based on Finite Element Method

In the actual process of geotechnical drilling,weak rock and soil layer,broken zone,fault,karst,high and low stress and other bad geology are often encountered,so it brings a series of problems and disasters to the drilling project. Many literature and test data prove that such problems are closely related to the operation parameters of drilling and the mechanical parameters of geotechnical soil. Therefore,taking PDC composite drill as the research object and simulation analysis,setting different rock mass parameters and different drilling parameters,the operation data obtained under different conditions is analyzed,the difference and connection between the drilling parameters in different rocks and the axial pressure and torque under the different drilling conditions is revealed. The analysis shows that the better the rock properties,the Mises,the higher;the shaft pressure and torque of the drilling speed,the drilling pressure increases the drilling depth and the rock cutting,and the torque increases,and the same drilling speed did not affect the axial pressure,the torque showed a positive correlation. Finally,the artificial neural network was established through Python language,and the numerical simulated data and actual drilling data were trained to verify that the use of drilling parameters and machine learning methods to realize the identification of rock types. The final results show that the accuracy of simulation data reached about 90%,and the actual data reached 78%.