基于多特征混合与GWO-SVM的气液两相流流型识别方法  

Flow pattern identification method based on multi-feature extraction and GWO-SVM for gas-liquid two-phase flow

作　　者：施艳艳 杨珍[1,2] 王萌 夏济根 SHI Yanyan;YANG Zhen;WANG Meng;XIA Jigen(College of Electronic and Electrical Engineering,Henan Normal University,Xinxiang 453007,Henan,China;Henan Key Laboratory of Optoelectronic Sensing Integrated Application,Henan Normal University,Xinxiang 453007,Henan,China;The 22nd Research Institute of China Electronics Technology Group Corporation,Xinxiang 453002,Henan,China)

机构地区：[1]河南师范大学电子与电气工程学院,河南新乡453007 [2]河南师范大学河南省光电传感集成应用重点实验室,河南新乡453007 [3]中国电子科技集团公司第二十二研究所,河南新乡453002

出　　处：《化工学报》2024年第10期3536-3547,共12页CIESC Journal

基　　金：国家自然科学基金项目(61903127,52277234)。

摘　　要：气液两相流流型识别对提高石油化工行业产能和生产效率具有重要作用。针对气液两相流电导波动信号的强非线性和非平稳特性导致特征提取困难、影响流型识别精度的问题,提出了一种基于多特征混合与灰狼算法优化支持向量机(GWO-SVM)气液两相流流型识别方法。研究中,分别采用统计分析法和熵分析法对电导波动信号的不同统计特征和归一化近似熵特征进行提取,并将两类特征混合构成数据集,再利用灰狼算法(GWO)对支持向量机(SVM)模型进行优化,以提高流型识别精度。气液两相流流型识别实验表明,所提方法比SVM、PSO-SVM和GA-SVM方法具有更高的识别精度,流型平均识别率达到98.45%。Flow pattern identification of gas-liquid two-phase flow plays an important role in the improvement of production capacity and efficiency of petrochemical industry.The nonlinear and non-stationary characteristics of conductance fluctuation signal lead to the difficulty of feature extraction and affect the accuracy of flow pattern identification of gas-liquid two-phase flow.Aiming at this problem,a new flow pattern identification method which combines multi-feature extraction with a grey wolf algorithm optimized support vector machine(GWO-SVM)is proposed in this work.In the research,statistical analysis and approximate entropy analysis are used to extract statistical features and normalized approximate entropy features of the conductance fluctuation signals.And the two different kinds of features are combined to form a dataset.Then,to improve the accuracy of flow pattern identification,grey wolf optimization(GWO)algorithm is used to optimize the support vector machine(SVM)model.The gas-liquid two-phase flow pattern recognition experiment showed that the proposed method has higher recognition accuracy than the SVM,PSO-SVM and GA-SVM methods,and the average flow pattern recognition rate reached 98.45%.

关 键 词：两相流 流型识别 多特征提取 支持向量机 

分 类 号：O359.1[理学—流体力学] TP183[理学—力学]