基于小波包多尺度信息熵和HMM的气液两相流流型识别方法

为了研究垂直上升管中气液两相流的流型,利用自制的多电导探针测量系统采集了四种典型流型的电导波动信息.由于气液两相流电导波动信号的非平稳特征,提出了一种基于小波包多尺度信息熵(Wavelet Packet Multi-scal2e Information Entropy)和隐马尔可夫模型(Hidden Markov Model,HMM)的流型识别方法.该方法首先对采集到的电导波动信号进行3层小波包分解,得到了8个不同频带的信号,提取各频带信号的小波包多尺度信息熵特征作为流型的特征向量,然后将其转换为观测序列输入到各种状态的隐马尔可夫模型进行训练并识别流型.结果表明:与BP神经网络相比,采用隐马尔可夫模型进行流型识别可以获得更高的识别率,表明该方法是有效和可行的.     

基于灰色神经网络和支持向量机的两相流流型辨识

两相流流型直接影响两相流的流动特性和传热传质性能。利用小波分析对气液两相流压降实验数据进行处理,提取不同频率的小波系数。以小波能量为特征,输入BP神经网络进行训练,进行流型的初步辨识。将灰色神经网络模型应用于气液两相流的辨识,同时创立将压差波动数据和小波能量数据输入Lib-SVM机分类器的方法,分别对流型进行辨识。结果显示,这三种方法均可进行流型的辨识,小波能量支持向量机的判别结果比灰色神经网络和BP神经网络的判别结果准确。支持向量机对压差信号直接进行流型辨识时准确率达到95.2%。     

基于奇异值分解和最小二乘支持向量机的气-液两相流流型识别方法

针对气-液两相流压差波动信号的非平稳特征和BP神经网络学习算法收敛速度慢、易陷入局部极小值等问题，提出了一种基于奇异值分解和最小二乘支持向量机（LS-SVM）的流型识别方法。该方法首先采用经验模态分解将气-液两相流压差波动信号分解为多个平稳的固有模态函数之和，并形成初始特征向量矩阵；对初始特征向量矩阵进行奇异值分解，得到矩阵的奇异值，将其作为流型的特征向量，根据LS—SVM分类器的输出结果来识别流型。对水平管内空气-水两相流4种典型流型进行识别，结果表明，与神经网络相比，该方法具有更高的识别率和识别速度。     

垂直上升矩形流道内气液两相流流型图的数值模拟

两相流流型在分析换热、流动不稳定性以及临界热流密度方面具有基础性作用.本文基于VOF(Volume of Fluid)多相流模型,对垂直上升矩形流道内气液两相流动进行数值模拟,表观气速0.1～110 m/s,表观液速0.1～3.2 m/s.得到了流道内气液两相流的主要流型:泡状流、弹状流、搅混流和环状流,分析了流道内截面含气率分布与流型的对应关系,以及截面含气率与气液两相流容积含气率的关系;分析了各种流型下的压降分布特性,并绘制了基于气液表观动能通量的不同流量下气液两相流的流型图,直观的表示出各种流型的分布区域及各流型间的流型转换边界,与已发表文献的实验结果对比符合较好.     

基于递归定量特征的气-液两相流型融合识别

为了进一步提高流型识别的准确率,针对气-液两相流压差波动信号的非平稳特征,提出了一种基于递归定量分析(RQA)和多传感器数据融合技术的流型识别方法.该方法首先采用RQA方法提取压差波动信号的非线性特征参数,对3个不同取压间距压差波动信号的特征参数进行特征层融合,构成融合特征向量,并运用融合的特征向量对支持向量机进行训练并识别流型.对水平管内空气-水两相流4种典型流型的识别结果表明,经过多传感器数据融合,识别结果的可信度明显提高.     

气-液两相流图像灰度脉动信号的多尺度双分形特性研究

应用高速摄影技术拍取气-液两相流水平管中3种典型流型的动态图像视频,对每一帧图像的平均灰度脉动信号进行提取;将提取的信号进行多尺度固有模态函数分解,然后与极差/标准偏差(R/S)分析方法相结合,提取各尺度的HURST指数和双分形特征.对气-液两相流的3种典型流型进行了气泡群和单个气泡2种形式的动力学行为分析,应用峭度系...     

摇摆对水平管内气液两相流流型的影响

本文对水平放置在摇摆台架,随台架做偏离水平位置的左右往复运动的管内空气-水两相流流型进行了实验研究.实验发现,低流速时,摇摆状态下水平管内流动变得很小稳定,流型发生周期性的改变:当水平管处于倾斜向上或倾斜向下状态时,管内流型分别有些近似于非摇摆的稳态倾斜上升或倾斜下降管内流型,并且流型转变要经历一个发展的过程,发展快慢与气相和液相流速大小有关;而在高液相或高气相流速时,摇摆状态下与非摇摆稳定状态下的两相流流型相近,主要有泡状流、间歇流(弹状流和准弹流)和环状流.根据可视观察以及气液界面在一个摇摆周期内的整体特征和部分时间段的局部特征,定义了不同流动条件下气液两相流的流型,给出了摇摆状态下水平管内气液两相流流型图.     

多孔介质通道内气-液两相流动阻力特性实验

基于新型水冷球床反应堆,以水和空气为工质,分别在直径为2、5、8 mm的玻璃球填充圆管形成多孔介质通道中,对竖直向上气-液两相流动阻力特性进行了实验研究.结果表明,阻力压降随着气液流量的增加而增大,并且与流型存在一定的对应关系;在相同流动条件下,颗粒直径和孔隙率对压降有明显影响.结合实验所得的234组实验点,对两类阻力...     

摇摆状态下水平管内气-水两相流流型的判定方法

对摇摆状态下水平管内气-水两相流流型进行了实验研究.实验发现,摇摆状态下两相流的压差波动有明显的周期性.本文根据各流型压差波动的差异判断摇摆状态下水平管内气-水两相流的流型.结果表明:通过与可视化观察和高速摄影观察的流型相比,利用压差特性曲线可以很好地判断摇摆状态下气-水两相流的流动型式.     

摇摆状态下水平管内气-水两相流的流型研究

摇摆产生的惯性力以及水平管路发生向上和向下倾斜,会使管道内两相流的流动型式发生变化.本文对直径25mm管内气-水两相流在摇摆周期为15s、摇摆角度为10°状态下的流型进行了实验,研究了气-水两相流在摇摆状态下的流动型式,并给出了流型图.实验结果表明,在一些气-水流量区域,两相流体在一个摇摆周期内存在两种流动型式.     

基于PCA-GA-SVM的竖直下降两相流流型预测

为提高小样本条件下的流型识别精度和时效性,提出了一种融合小波包分解（WPD）、主元分析（PCA）、遗传算法（GA）和支持向量机（SVM）的优化识别模型,并成功应用在竖直下降两相流流型辨识工作中。利用WPD对非平稳电导波动信号进行分解、重构,提取小波包能量构造特征向量;通过PCA对特征向量进行降维,降低特征输入的复杂性;同时采取GA全局迭代寻优的方式确定SVM的关键参数惩罚因子（C）和核函数参数（g）。对PCA-GA-SVM识别效果进行验证后与SVM、PCA-SVM、GA-SVM网络进行对比。结果表明,经过PCA和GA优化后的SVM网络在流型识别精度和时效性方面均提升显著,对泡状流、弹状流、搅拌流和环状流的总体预测精度达到了94.87%,耗时仅3.95 s,可满足在线识别需求。      

Classification of two-phase flow regimes via image analysis and a neuro-wavelet approach

A non-intrusive method of two-phase flow identification is investigated in this paper. It is based on image processing of data obtained partly from dynamic neutron radiography recordings of real two-phase flow in a heated metal channel, and partly by visible light from a two-component mixture of water and air. Classification of the flow regime types is performed by an artificial neural network (ANN) algorithm. The input data to the ANN are some statistical moments of the wavelet pre-processed pixel intensity data of the images. The pre-processing used in this paper consists of a one-step multiresolution analysis of the 2-D image data. The investigations of the neutron radiography images, where all four flow regimes are represented, show that bubbly and annular flows can be identified with a high confidence, but slug and churn-turbulent flows are more often mixed up in between themselves. The reason for the faulty identifications, at least partially, lies in the insufficient quality of these images. In the measurements with air-water two-component mixture, only bubbly and slug flow regimes were available, and these were identified with nearly 100% success ratio. The maximum success ratio attainable was approximately the same whether the raw data was used without wavelet preprocessing or with a wavelet preprocessing of the input data. However, the use of wavelet preprocessing decreased the training time (number of epochs) with about a factor 100.     

摇摆流动不稳定性的遗传算法优化神经网络预测

摇摆工况下自然循环系统的流动不稳定性现象对船用核动力系统的安全性有着显著影响。结合神经网络和遗传算法,对复杂不稳定性行为的预测进行了优化。采用小数据量法计算了流量时间序列的最大Lyapunov指数,得到了时间序列的最大可预测时间。应用单隐层BP神经网络对流量变化进行了多步滚动预测,在步数较少时预测结果与实验结果符合较好。但由于BP神经网络存在陷入局部最优解的问题,为此采用遗传算法对神经网络的初始阈值和权值进行优化,从而改善了BP神经网络的非线性预测性能。本文结果为流动不稳定性的实时预测提供了一种易于实际应用且准确度较高的途径。     

基于能量分布特征的地震事件自动识别

研究了地震信号在小波包变换下的特性,依据地震事件识别中“历史事例对比法”的思想,根据不同震源地震信号频率时变特性的不同．提出了基于“能量分布特征”的特征值,同时采用该特征值用神经网络方法对地震事件进行识别分类。该方法不依赖于系统的数学模型,而是直接利用各频率成分能量的变化提取特征值作为神经网络的输入特征向量来进行事件的识别,避免了对地震信号、传播途径准确建模的困难,简便、直观地完成了事件的识别。实验证明,该方法的事件识别率可达到99％以上．是一种有效的地震事件识别方法。     

拌摩擦焊焊缝缺陷超声检测信号特征分析与神经网络模式识别

以搅拌摩擦焊（FSW）焊缝的包铝层伸入、未焊透、隧道孔缺陷为对象，将小波分析理论应用于缺陷超声检测信号特征提取问题的研究，使用小波包分解重构节点能量、小波包分解节点系数、缺陷信号的功率谱密度小波分解这三种方法对缺陷的超声检测信号进行特征提取。利用类别可分离性判据和BP神经网络分别对提取的特征量进行评估和识别。结果表明，缺陷信号的功率谱密度小波分解这一特征提取方式具有最好的类别可分性，并且以该特征量为网络输入的BP神经网络具有85.71%缺陷识别率。      

A general regression artificial neural network for two-phase flow regime identification

Supplementing the collection of artificial neural network methodologies devised for monitoring energy producing installations, a general regression artificial neural network is proposed for the identification of the two-phase flow that occurs in the coolant channels of boiling water reactors. The utilization of a limited number of image features derived from radiography images affords the proposed approach with efficiency and non-invasiveness. Additionally, the application of counter-clustering to the input patterns prior to training accomplishes an 80% reduction in network size as well as in training and test time. Cross-validation tests confirm accurate on-line flow regime identification.     

Flow regime identification methodology with neural networks and two-phase flow models

Vertical two-phase flows often need to be categorized into flow regimes. In each flow regime, flow conditions share similar geometric and hydrodynamic characteristics. Previously, flow regime identification was carried out by flow visualization or instrumental indicators. In this research, to avoid any instrumentation errors and any subjective judgments involved, vertical flow regime identification was performed based on theoretical two-phase flow simulation with supervised and self-organizing neural network systems. Statistics of the two-phase flow impedance were used as input to these systems. They were trained with results from an idealized simulation that was mainly based on Mishima and Ishii's flow regime map, the drift flux model, and the newly developed model of slug flow. These trained systems were verified with impedance signals measured by an impedance void-meter. The results conclusively demonstrate that the neural network systems are appropriate classifiers of vertical flow regimes. The theoretical models and experimental databases used in the simulation are shown to be reliable.     

Non-invasive on-line two-phase flow regime identification employing artificial neural networks

A novel non-invasive approach to the on-line identification of BWR two-phase flow regimes is investigated. The proposed approach receives neutron radiography images of coolant flow recordings as its input and performs feature extraction on each image via simple and directly computable statistical operators. The extracted features are subsequently used as inputs to an ensemble of self-organizing maps whose outputs demonstrate swift and accurate classification of each image into its corresponding flow regime. The novelty of the approach lies in the use of the self-organizing map which generates the different classes by itself, according to feature similarity of the corresponding images; this contrasts traditional artificial neural networks where the user has to define both the number of distinct classes as well as to supply separate training vectors for each class.