基于数值计算与遗传神经网络热熔钻研究

来源期刊:中南大学学报(自然科学版)2013年第12期

论文作者:温继伟 陈晨

文章页码:5051 - 5060

关键词:热熔钻;热熔体;玻璃状硬壳;数值计算;神经网络

Key words:hot melt drilling; hot melt; glassy hard shell; numerical calculation; neural network

摘    要:对热熔钻进过程中热熔体的相关问题展开研究,建立相应的数学模型,结合热熔钻进实验过程中的相关参数进行计算,得到了相应的解析解,并分析其规律与产生的原因,计算结果的变化规律与实际过程基本相符,表明所构建的数学模型是可靠的。此外,对热熔钻进过程中孔壁周围玻璃状硬壳的形成展开研究并给出了可靠的硬壳厚度计算的经验公式。将遗传算法与BP算法相融合构建GA-BP神经网络分别对有关热熔体与施加在热熔器上的有效热功率及钻压对热熔钻进速度的影响的数据进行预测。研究结果表明:基于GA-BP神经网络对热熔体在环空间隙中流速的分布,热熔体作用在热熔器外表面上的压力,热熔体作用在热熔器上的摩擦力,热熔器有效热功率在不同地层中对热熔钻进速度的影响和热熔器施加不同钻压在不同地层中对热熔速度影响的预测得到数据的均方误差比基于BP神经网络预测得到数据的均方误差分别小0.449×10-6,0.005 6,0.001 1,0.104和0.136,且GA-BP网络比BP网络的计算时间分别短7,15, 2, 9和15 s。

Abstract: Through the study on the series of problems of hot melt in the drilling process, the mathematical models were established, and by calculating the relevant parameters in hot melt drilling experiment, the corresponding analytical solutions was obtained, and its own laws and the produced reasons was analysed, the variation of the calculation results were basically the same with the actual process. It is shown that the constructed mathematical model is reliable. In addition, the study on the glassy crust formation around the hole wall in the process of the hot melt drilling was conducted and the hard shell thickness calculated reliable empirical formula was given. Genetic algorithm and BP algorithm integration were used to build the GA-BP neural network to predict the data of the hot melt and the influence on hot melt drilling speed of the effective thermal power and the WOB applied to the subterrene. The results show that the mean square errors based on GA-BP neural network to predict the velocity distribution of hot melt in annular clearance, the role of the hot melt pressure on the outer surface of the subterrene, the friction of the hot melt in the subterrene, the influence on the hot melt drilling speed of the effective thermal power on the subterrene, the influence on the hot melt drilling speed of the impose different WOB on the subterrene in different stratas are 0.449×10-6, 0.005 6, 0.001 1, 0.104, 0.136, respectively, which are lower than those based on BP neural network. And the computing time of GA-BP network are 7, 15, 2, 9 and 15 s, which are shorter than BP network.

相关论文

  • 暂无!

相关知识点

  • 暂无!

有色金属在线官网  |   会议  |   在线投稿  |   购买纸书  |   科技图书馆

中南大学出版社 技术支持 版权声明   电话:0731-88830515 88830516   传真:0731-88710482   Email:administrator@cnnmol.com

互联网出版许可证:(署)网出证(京)字第342号   京ICP备17050991号-6      京公网安备11010802042557号