基于多尺度特征向量的密相气力输送两相流流型划分及识别方法研究

Flow patterns information of the gas-solid two-phase flow in dense phase pneumatic conveying system has great significance on the optimization design of the facility, prediction and control of the transportation stability, etc. Meanwhile, the flow pattern identification of the gas-solid two-phase flow plays an important role in the field of multiphase flow measurement. In the presented project, a novel method based on multi-scale characteristic value vector for flow patterns diversion and identification in dense phase pneumatic conveying pipeline is proposed to solve problems in the field of flow pattern characterization extraction and flow pattern identification. Based on the multi-scale method, the extraction of the flow pattern characterization is carried out in multiple scales independently, then the multi-scale characteristic value vector is constructed as flow patterns indicator. The multi-scale characteristic vectors of traditional flow patterns are acquired by fusing the output signals of an electrostatic sensor and a pressure sensor installed on the conveying pipeline. Then, the identification standard of traditional flow patterns is derived based on the multi-scale characteristic value vector，while the relationship between the new patterns and the multi-scale characteristic value vector is studied to identify flow pattern criterion for flow patterns identification. Based on the multi-scale characteristic vector and Fuzzy Support Vector Machines, a method of flow patterns identification in the dense-phase gas-solid two-phase flow is proposed. The project is expected to promote the development and application of the online automatic flow pattern identification technology in the dense phase pneumatic conveying system.

管道中的流型信息，对密相气力输送系统的优化设计、运行控制及输送稳定性等具有极其重要意义，气固两相流流型识别是多相流检测研究的热点和难点。本申请针对流型特征值提取及流型识别中存在的问题，提出基于多尺度特征向量的密相气力输送两相流流型划分及识别的新方法。采用多尺度方法分尺度从气固两相流系统波动信号中提取流型特征值并构造多尺度特征向量，进而对流型进行表征；以静电和压力传感器采集信号作为多尺度信息源并融合，从中获取传统流型的多尺度特征向量；明确基于多尺度特征向量的流型划分标准，对传统流型进行归类划分，建立划分后的流型与多尺度特征向量之间的对应关系，为流型识别提供理论判据；建立基于多尺度特征向量和模糊支持向量机的流型在线识别方法，实现对流型的判别，为研制高效准确的流型在线识别仪器提供理论基础和技术支持。该项研究成果有望推动流型在线识别技术在密相气力输送系统中的发展和应用。

管道中的流型信息，对气力输送系统的优化设计、运行控制及输送稳定性等具有极其重要意义，气固两相流流型识别是多相流检测研究的热点和难点。该项目提出了基于多尺度能量比重的气力输送两相流流型划分的新方法。建立了以多尺度能量比重为参数的流型划分准则,气固两相流存在5种流型，分别定义为“微”、“微介”、“介”、“介宏”和“宏”流型；利用静电传感器获取气力输送实验装置水平管中的流动信号，采用信号处理方法从静电信号中分解得到多尺度能量比重。根据流动状态对应的多尺度能量比重，对其进行流型划分，同时明确了这5种流型所对应的流动状态。 另外，建立了基于阵列式静电传感器和BP神经网络的流型在线识别方法，将阵列式静电传感器的输出信号幅值的平均值和方差分别作为流型特征值和BP神经网络的输入参数，训练后的神经网络对“微”、“介”和“宏”3种流型的识别率均为100%。该项研究成果有望推动流型在线识别技术在密相气力输送系统中的发展和应用。