基于EEMD与分形盒维数的气固两相流流型识别  

Identification of gas-solid two-phase flow regimes using EEMD and fractal box dimension

作　　者：张肖[1,2] 李林[1] 胡红利[1] 

机构地区：[1]西安交通大学电气学院,陕西西安710049 [2]西安市产品质量监督检验院线缆室,陕西西安710065

出　　处：《西北大学学报（自然科学版）》2016年第2期178-183,共6页Journal of Northwest University（Natural Science Edition）

基　　金：国家自然科学基金资助项目(51177120);国家"863"计划基金资助项目(2009AA04Z130);电力设备电气绝缘国家重点实验室主任基金资助项目(EIPE14132)

摘　　要：针对气固两相流气力输送管道中测控装置后常见的几种过渡流型,如层流、环流与核心流,文中通过静电传感器获取静电波动信号,利用总体经验模式分解和分形盒维数相结合的方法进行流型识别。首先使用总体经验模式分解对不同流型的静电信号进行自适应分解从而获得位于不同频段的本征模式函数,然后利用分形技术对所分解的IMF求取盒维数,并将其作为特征量。最后通过提取的特征量对BP神经网络进行训练和测试从而识别流型。实验结果表明,该方法能够有效识别气固两相流流型,识别率达94%。For three common transitional flow regimes behind the detection and control devices in the pneumatic conveying pipeline of gas-solid two-phase flow,namely stratified flow,annular flow and core flow,the electrostatic fluctuation signals were obtained through electrostatic sensor.In this paper,Ensemble Empirical Mode Decomposition（EEMD） and fractal box dimension are combined to identify the flow regimes.Firstly,via EEMD algorithm,the electrostatic signals of different flow regimes were decomposed into a series of Intrinsic Mode Function（IMF） in different frequency channels correspondingly.Then fractal box dimensions of IMFs were calculated and treated as the feature parameters.At last,the BP neural network was trained and tested using the feature parameters to identify the flow regimes.The experiment shows this method of combing EEMD and fractal box dimension can identify the flow regimes effectively,and the identification rate reaches 94%.

关 键 词：气固两相流 总体经验模式分解 盒维数 流型识别 

分 类 号：O359.2[理学—流体力学]