基于机器学习的小尺寸涡轮钻具输出性能优化
Research on Performance Optimization of Small-Sized Turbine Drilling Tools Based on Machine Learning
胡子龙 1陈婷 1马卫国 1聂玲1
作者信息
- 1. 长江大学机械工程学院,湖北荆州 434023
- 折叠
摘要
为解决小尺寸涡轮钻具扭矩小带来的弊端,建立一种基于BP神经网络和非支配排序多目标遗传算法(NSGA-II)优化模型,通过对涡轮结构尺寸进行优化,得到高效率下扭矩更大的涡轮.采用权重估算和Garson算法,对涡轮的扭矩和效率进行敏感性分析,再通过比较多种机器学习算法构建回归模型的拟合度,选用反向传播神经网络(BPNN)建立扭矩和效率与设计参数之间的回归模型,最后结合非支配排序多目标遗传算法(NSGA-II)寻求Pareto最优解集.结果表明:安装角对输出扭矩的影响最大,叶片数对输出效率的影响最大;采用BP神经网络构建的回归模型最为准确;优化后的涡轮与初始涡轮相比,扭矩提高 1.2 倍,效率提升 1.35%.
Abstract
In order to solve the drawbacks brought about by the small torque of the small-sized turbine drilling tools,an optimiza-tion model based on BP neural network and non-dominated sorting genetic algorithm(NSGA-II)was established,the turbine structure size was optimized to obtain a turbine with greater torque under high efficiency.Sensitivity analysis of torque and efficiency was conduc-ted using weight estimation and Garson algorithm,multiple machine learning regression models were compared for their fitting perform-ance and back propagation neural network(BPNN)was selected to establish the regression model between torque&efficiency and de-sign parameters.Finally,the non-dominated sorting genetic algorithm II(NSGA-II)was used to search for the Pareto optimal solution sets.The results show that the installation angle has the greatest impact on the output torque,and the number of blades has the greatest impact on output efficiency.The regression model constructed with BP neural network exhibits the highest accuracy.The optimized tur-bine has 1.2 times more torque and 1.35%more efficiency than the initial turbine.
关键词
涡轮钻具/神经网络/敏感性/NSGA-II算法/Pareto最优解Key words
turbine drilling tools/neural network/sensitivity/NSGA-II algorithm/Pareto optimal solution引用本文复制引用
出版年
2024
NETL
NSTL
万方数据