【作者】 李洪伟；

【导师】 周云龙；

【作者基本信息】 华北电力大学 ， 热能工程， 2012， 博士

【摘要】 两相流动现象广泛存在于自然界和现代工业生产过程中,与人类的生活及生产密切相关,两相流流动参数检测对生产过程及工艺优化具有重要意义。但由于两相流存在复杂相间界面效应及相对运动,尤其是流型转变动力学机理至今尚未十分清楚,加之理论与技术上的欠缺,两相流参数检测问题,特别是复杂流体流量测量和流型识别问题长期未能得到很好解决。图像处理技术是一门新兴的检测技术,近年来图像处理在工业检测中的应用越来越受到人们的重视。在两相流系统的参数检测中,图像处理技术也得到了广泛的应用。采用图像处理技术与现代信息处理技术融合的检测理论与方法,是解决复杂两相流流动参数检测的有效途径。本文采用图像处理与多尺度非线性分析等相结合的方法研究了气液两相流的流动特性,应用特征挖掘及信息融合的软测量方法实现了两相流流型参数测量。取得了创新性研究成果。在气液两相流型图像边缘分割的诸多算法中,对等高线方法进行了改进,首先对图像进行二维高斯核梯度模的求取,在梯度模对图像边缘增强的基础上进行经验模态分解,取得高频信息层子图像并进行叠加,后应用开闭运算进行细节噪声去除与边缘修补,最后应用等高线理论进行特征提取和表达。通过对比发现,这种图像特征表达方法不仅能够对图像的边缘特征有一个很好的检测,而且能够保留一些经典算法所保留不了的细节信息,考虑到气液两相流复杂的动力学机理,此方法应用于此,能够为气液两相流图像特征提取提供一个新思路。应用了一种新的时间序列的提取方法,将气液两相流视频信号的每一帧图像分块成小的区域,每相邻两帧图像的相邻两区域进行灰度相似值与最大差值计算,分别组成时间序列。对时间序列分别求取最大李亚普诺夫指数,组成最大李亚普诺夫指数矩阵。利用李亚普诺夫指数的特征将气液两相流视频划分为混沌特性不同的区域,分别对其从整体以及细节两部分进行分析,分别采用0,1分布图谱、二维和三维等高线图谱以及概率密度谱结合最大李亚普诺夫指数矩阵的分形盒维数、香农熵以及均值等特征,对气液两相流型流动机理进行辅助分析。实验结果表明对图像小区域分块结合最大李亚普诺夫指数提取的方法能够区别出不同流型的流动特性,气液两相流视频图像的背景与变化相界面具有强度不同的混沌特性,是一种有效的分析气液两相流图像信号的新方法。将EMD分解与HURST指数、递归图特征相结合对气液两相流型从整体到细节进行了剖析,并提取递归特征中的典型的对角线平均长度值与香农熵值两个特征值,观察其随气相表观速度增加的变化情况,对气液两相流型的动力学特征进行定量分析。研究发现随着气相表观速度的增加,双分析特征的变化规律更为明显。在高频与低频段,三种流型存在单分形特征,而在中频段则呈现出双分形摸样,从整体与局部两方面对三种流型的流动机理进行了分析。在递归分析中,对角线平均长度与递归香农熵两个特征值的变化体现出了气泡与气塞乃至整个泡群的运动规律。在EMD分解后的高频信息层中段塞流较泡状流与雾状流存在明显峰值。总之,基于图像灰度波动信号的多尺度非线性分析方法是理解与表征气液两相流动力学特性的有效途径。将应用于桥梁建筑性能分析的随机子空间方法应用到气液两相流型图像灰度波动信号的分析中,而用于确定系统阶次的一种比较新颖的方法稳定图也被应用到气液两相流灰度波动信号的分析中,并结合直线度特征值对复杂气液两相流时间序列进行定量表征。建立了基于随机子空间方法进行流型辨识的新标准,雾状流相角在1.35-2.07之间,泡状流相角在4.01-4.36之间,段塞流相角在-0.52-0之间,其不受实验工况的变化所影响,完全由流型内部模态结构所决定,能够准确的对流型信号进行区分。更多还原

【Abstract】 The phenomenon of the two-phase flow is widly found in nature and modern industrial production process, and it is closely related to the human life and production. The measurement of the flow parameters is of great significance for industrialproducing process and the optimization of process. Because of complex interfacial effect and relative motion between the phases, it is difficult to identify accurately the two-phase flow patterns. In particular the mechanism of the transition from flow patterns to dynamics has not yet been very clear. And the lack of theory and technology, the measurement of two phase flow parameters failed to get very good solved, especially the complex fluid flow measurement and flow patterns’ identification problem.Image processing technology is a new measurement technology, in recent years the image processing was applicated to the industrial measurement more and more by people. Image processing technology was also be taked a wide range of applications in the two phase flow system.Therefore it’s an effective approach to solve the measurement of complex two phase flow parameters by the fusion of image processing and modern information processing method and theory.the nonlinear dynamic characteristics of two phase flow patterns were studied by using multiscale nonlinear analyzing method and image processing.The flowrate of two phase flow was measured by soft measurement method based on characteristics extracting and information fusion.And the innovative research fruits are shown as below:In so many image edge division algorithms, the paper improved the contour method and applicated in the images of gas-liquid two-phase flow patterns. To get effective extraction from the main objective of the gas-liquid two-phase flow images, in this paper, the2-Dimensional Gaussian kernel gradient, the multi-scale image decomposition techniques, the opening and closing operations, and the contour theory are combined together to put forward a way of the expression of the image characteristics. The new method can be described as follows:first, the decomposition of the empirical mode is based on the gradient mode enhancing image edge to get the high-frequency information sub-images and superpose them; then, the application of the opening and closing operation is to reduce the details’noise and repair the edge; finally extracting and expressing the features are by using the contour theory. By comparison, this way of expression can not only get the edge features of the image very well, but also retain some details, which can not be reserved by some classical algorithm. This method applied here can provide a new idea for extracting features from gas-liquid two-phase flow images.Each frame of the video signal is divided into smaller areas by a new method for extracting time series.The gray similar values and the maximum gray scale difference of each two adjacent frame are calculated,then formed the time series.The largest Lyapunov exponents of time series are respectively extracted,and the largest Lyapunov exponent matrix is composed.The videos of gas-liquid two-flow patterns are divided into different chaotic characteristic areas by the characteristics of Lyapunov exponent.Then they are respectively analyzed from overall and details by zero and one distribution map and3D map. The flowing mechanism of gas-liquid two-phase flow is analyzed,combined the fractal box dimension and Shannon entropy of the largest Lyapunov exponent matrix.The results show that the method of the gray similar values series of small areas combined extracting the largest Lyapunov exponent can distinguish the flowing characteristics of different flow patterns;the background and changed phase interface of gas-liquid two-phased flow video have chaotic charateristics of different intensity,which is an effective method for analyzing the gas-liquid two-phase flow signals.Empirical Mode Decomposition (EMD) method, HURST exponents and Recurrence Plot (RP) were combined together to analyze the gas-liquid two-phase flow from the whole to the details. At the same time the various signals are checked in the chaotic recursion chart by which the two typical characteristics (diagonal average length’s and Shannon entropy) are obtained. The results show that the change regularity of the characteristic of dual-fractal is more obvious with the increase of gas surface measured velocity. In the high frequency section and low frequency section, all of the three flow patterns have simple fractal characteristic. But in the middle frequency section, the flow patterns appear to be dual fractal characteristic. The flow mechanism of the three flow patterns is analyzed partly and wholly.The Movement Law of bubble, gas slug and the whole bubble group present the change of two characteristics that are diagonal average length’s and recursive Shannon entropy’s. After the decomposition by the EMD method the slug flow and the mist flow in the high frequency section have obvious peaks.Anyway, it is an effective way to understand and characterize the dynamic characteristics of two-phase flow patterns that the multi-scale non-linear analysis method is based on image gray-scale fluctuation signals.The Stochastic subspsace parameter identification(SSI) which was used in the analysis of the construction of the bridge will be applized in the analysis of the grey fluctuation signals of gas-liquid two-phase flow patterns’images. And the stability graph which can determine the order of the system also be used in the analysis of the fluctuation signals of gas-liquid two-phase flow patterns’images.Complex time series can be quantitative characterized combining straightness eigenvalue.The feature vector of the pressure fluctuation sequence of three typical gas-liquid two-phase flow were extracted by application of random sub-space. It can be distinguished by amplitude and phase angle, of which the phase angle feature is the most reliable. The phase angle of mist flow is between1.35to2.07. The phase angle of bubbly flow is between4.01to4.36. The phase angle of slug flow is between-0.52to0. Changes of experimental conditions will not bring any effects. Determined entirely by flow pattern internal modal structure, the flow pattern signal can be accurately distinguished.更多还原