Liqnet: A real-time monitoring network for two-phase flow patterns

Real-time identification of gas-liquid two-phase flow can help fluid systems maintain safe operating conditions. A flow pattern identification method based on a convolutional neural network (CNN) algorithm (after this referred to as liqnet) is proposed in this paper to realize automatic detection and real-time identification of two-phase flow patterns. This paper mainly focuses on solving two problems of CNN algorithm flow pattern identification (1): the experimental samples for two-phase flow classification are few, and (2): the existing methods do not fully consider the real-time nature of two-phase flow identification. Therefore, this paper constructs a two-phase flow database containing 6242 images using data enhancement, proposes a lightweight network liqnet, and compares it with six mainstream CNN models. The results show that liqnet can achieve the highest accuracy (98.65%), has the least amount of parameters (1.3708 M), and can achieve the purpose of real-time prediction (32.11FPS).     

Flow pattern identification for gas-oil two-phase flow based on a virtual capacitance tomography sensor and numerical simulation

Gas-oil two-phase flow is widely encountered in oil exploitation and transportation pipelines. It's complex and transient changes of flow regimes present a great challenge for accurate and real-time measurement. As a non-invasion and real-time measuring method, electrical capacitance tomography (ECT) is suitable for the transient measurement of non-conductive gas-oil flow. However, the highly random and nonlinear nature of multiphase flow make it difficult and limited to investigate the flow parameters based on either static or dynamic measurement. In this research, the whole process of dynamic measurement of ECT applying in gas-oil two-phase flow is thoroughly studied, including simulation calculation, experimental validation and comprehensive data analysis. A simulation approach by coupling the flow and electrostatic field is proposed based on a virtual ECT sensor, in order to monitor the gas-oil two-phase flow characteristics. Based on FLUENT and COMSOL platform, the numerical simulation under six typical flow patterns in a horizontal pipe is carried out. Combining the visualized image generated by ECT measurement and the theory of flow pattern transition, the formation mechanism and structural characteristics of different gas-oil flow patterns are analyzed in detail. Furthermore, this research attempts to analyze the signal fluctuation characteristics caused by flow pattern change, in order to access more in-depth flow information implied in the original capacitance data, via time-series analysis as well as frequency domain analysis based on Flourier Transform. At last, a series of dynamic experiment is conducted to verify the feasibility of the simulation and data analysis approach. The experiment focuses on the flow pattern transition, gas-liquid dynamic characteristics and noise influence in the actual process. It can be concluded from the results of simulation and experiment tests, combining the visualized images and the dynamic characteristics of capacitance signals can make it more effective and intuitive for flow pattern identification, which might be used for the online measurement in real-industry process.     

Intelligent identification of two-phase flow patterns in a long pipeline-riser system

Accurate and rapid identification of multiphase flow patterns in long-distance pipelines is an important means for flow assurance. In this paper, an experiment of air-water two-phase flow has been carried out on a pipeline-S-shaped riser system with a length of 1687 m. Based on the amplitude of riser differential pressure, the working conditions are categorised into three typical flow patterns, severe slugging flow, oscillatory flow, and stable flow. Only two difference pressure signals that are most practical near the offshore platforms are used. The data of liquid phase accumulation of severe slugging is adopted as samples, and the signal features are extracted by wavelet multiresolution analysis. The parameters of six common classifiers are optimized, and the effects of different classifiers with the optimal hyperparameters on flow pattern recognition are compared and analyzed. For neural networks, decision trees, support vector machines and k-Nearest Neighbors classifier with the optimal hyperparameters, the recognition rate of severe slugging is higher than 95.8% and the average recognition rate of the three flow patterns is higher than 94.2% when the sample duration is only 6.2 s. On the premise of achieving the highest recognition rate, the number of features is substantially reduced to 15.6% of the original number by principal component analysis.     

基于 CEEMDAN 和概率神经网络的 起伏振动气液两相流型识别

起伏振动气液两相流型准确识别对漂浮核动力平台安全稳定运行有重要意义。通过对比静止和起伏振动管道的压差信号以及对应的频谱图发现,起伏振动管道内的压差信号波动幅度更大且包含更多的频率分量,两种流型均含有主频率,该频率为起伏振动频率。针对起伏振动状态气液两相流压差信号的复杂性,分别采用自适应白噪声的完备总体经验模态分解(CEEMDAN)和集合经验模态分解(EEMD)对小波降噪后的压差信号进行模式分解,发现CEEMDAN能够在减少模式分量的同时获得更多有效的分量。通过计算spearman相关系数选择具有表征意义的IMF分量进行Hilbert变换计算能量作为特征值,采用概率神经网络对流型进行识别。结果表明,采用CEEMDAN进行模式分解结合概率神经网络的识别方法准确率达到95.83%,能够用于起伏振动下气液两相流型识别。     

基于经验模态分解和BP神经网络的油气两相流流型辨识

基于经验模态分解（empidcal mode decomposition，EMD）BP神经网络，提出了油气两相流流型辨识的新方法。应用EMD将差压信号分解成不同频率尺度上的单组分之和，并提取组分的归一化能量作为流型辨识特征量。BP神经网络以这些能量特征量为输入对油气两相流不同流型（包括泡状流、塞状流、层状流、弹状流和环状流）进行分类。实验结果表明，本文提出的流型辨识方法是有效的，其中泡状流、塞状流、层状流、弹状流和环状流的辨识精度分别为100％、89．4％,93．3％、96.3％和96.9％。     

Development of noise pattern map for predicting refrigerant-induced noise in refrigerators

In refrigeration systems, flow state in evaporators is two-phase. Acoustic characteristics of noise generated in pipes with two-phase flow depend on flow patterns and instantaneous cycle conditions. Refrigerant-induced noises are irregular and have a band frequency. Therefore, it is difficult to predict the occurrence conditions of refrigerant-induced noise in evaporators. Hence, the development of a design guide that anticipates the occurrence of refrigerant-induced noise is important. Flow pattern maps have been proposed by Baker, Hewitt, Taitel-Dukler, etc. However, these do not consider the acoustic viewpoint and cannot predict refrigerant-induced noise based solely on the estimated flow patterns. Therefore, in this study, we suggest a noise pattern map for directly anticipating refrigerant-induced noise from various flow patterns. We performed experiments using R600a refrigerant and refrigerant-supply equipment designed for flow visualization and flow regime transition. Based on experimental data, we suggest a noise pattern map for determining the level of refrigerant-induced noise.     

基于ECT投影数据的两相流流型辨识

提出一种基于电容层析成像（ECT）系统的两相流流型辨识新方法。ECT投影数据经特征参数提取前处理环节后馈人事先训练好的神经网络实现流型辩识。文中介绍了特征参数的提取、神经网络的构成以及训练方法。研究结果表明，该方法辨识精度高、辨识速度快，是两相流流型在线辨识的一种有效手段。     

An adaptive CEEMD-ANN algorithm and its application in pneumatic conveying flow pattern identification

The accurate measurement of dust concentration using electrostatic sensor is serious affected by two-phase flow patterns in practice. In this paper, the electrostatic sensor signals of flow in a pneumatic conveying pipeline were collected, and the electrostatic fluctuation signals of three typical flow patterns of gas–solid two-phase flow in the horizontal pipe were obtained. By combining complementary ensemble empirical mode decomposition (CEEMD) and a back propagation (BP) neural network, an algorithm for flow pattern identification is proposed. This algorithm can adaptively determine the number of layers of the intrinsic mode function (IMF) decomposition and the number of input vectors for the neural network, ensuring the minimum size vector is used. The selected IMF energy feature as the input of the BP neural network can effectively ensure that an accurate flow pattern discrimination rate is obtained. The experimental results show that the algorithm proposed in the paper can guarantee the recognition rate of the flow pattern to reach more than 99%, yet through adaptive adjustment ensure that the size of trained BP neural network input is as small as possible, and the guaranteed algorithm calculation is kept at a minimum.     

汽车液压制动系统气液两相流流型的识别

建立了汽车液压制动系统中气液两相流流型检测装置,根据压差波动信号,利用Hilbert-Huang变换(HHT)对制动液两相流流型进行识别,并利用高速摄像机采集不同工况下制动液的气液两相流流型图像。结果表明,制动时车轮转速越高,压差信号幅值越大,幅值主要集中在0～50Hz区域;识别制动时的制动液流型为一种泡状流。高速摄影的结果验证了液压制动管路中制动液为泡状流;制动转速越高,气泡越小。结论揭示了制动时汽车制动液的气液两相流流型,说明利用测量制动液的压差波动信号进行HHT就可以识别其流型。     

Identification of oil–gas two-phase flow pattern based on SVM and electrical capacitance tomography technique

The correct identification of two-phase flow patterns is the basis for the accurate measurement of other flow parameters in two-phase flow measurement. Electrical capacitance tomography (ECT) is a new visualization measurement technique for two-phase/multi-phase flows. The capacitance measurements obtained from the ECT system contain flow pattern information, and then six feature parameters are extracted. The support vector machine (SVM) has a desirable classification ability with fewer training samples. The inputs of the SVM are extracted feature parameters of different flow patterns. Simulation and static experiments were carried out for typical flow patterns. Results showed that this method is fast in speed and can identify these flow patterns correctly.     

Doppler spectrum analysis and flow pattern identification of oil-water two-phase flow using dual-modality sensor

Horizontal oil-water two-phase flow widely exists in petroleum and chemical engineering industry, where the oil and water are usually transported together. As one of most importance process parameters to describe the two-phase flow, the flow pattern can reflect the flow characteristics of inner flow structure and phase distribution. The identification of flow pattern will contribute to develop more accurate measurement model for flow rate or phase fraction and ensure the safety and efficiency of operation in industry. A dual-modality sensor combining with continuous wave ultrasonic Doppler sensor (CWUD) and auxiliary conductance sensor, was proposed to identify flow patterns in horizontal oil-water two-phase flow. In particular, the oil-water flow characteristic was analyzed from Doppler spectrum based on the CWUD sensor. Besides, the dimensionless voltage parameter based on conductance sensor was applied to provide the information of continuous phase in the fluid. Several statistical features were directly extracted without any complicated processing algorithm from Doppler and conductance signals. The extracted features are put into a multi-classification Support Vector Machine (SVM) model to classify five oil-water flow patterns. The results show that the overall identification accuraccy of 94.74% is satisfactory for horizontal oil-water two-phase flow. It also demonstrates that the noninvasive ultrasonic Doppler technique not only can be used for flow velocity measurement but also for flow pattern identification.     

Pattern recognition and signal processing of ultrasonic echo signal on two-phase flow

This paper presents a novel measurement method using ultrasonic echo signals on the flow of air–water mixtures. This method has the capability of measuring an instantaneous echo intensity profile along an ultrasonic beam, so it is expected to apply to pattern recognition of two-phase flow. Additionally, this method has an advantage compared with conventional techniques because of the clump-on type. The principle of the flow pattern recognition is based on the delay time and strength of the pulse echo. In this paper, first of all, the transmission of ultrasound through solid plates, which are made of plexiglass and carbon steel, has been investigated and the effective incidence angles for these materials were found. Then, echo signals reflected off a boundary between water and air in a vertical pipe, having a diameter of 50 mm, were obtained using an ultrasound system, and the effects estimated of a two-phase pattern, from bubbly to slug flow, on the signals. In addition, water flows down the inner surface of a pipe as annular flow, and the echo signals then also investigated.     

Identification of boiling flow pattern in narrow rectangular channel based on TFA-CNN combined method

In this paper, an indirect flow pattern recognition method based on time-frequency analysis and neural networks is proposed to investigate the flow patterns in the narrow rectangular channel under heating and non-inertial conditions. Firstly, the adaptive optimal kernel algorithm is utilized to analyze on the typical pressure signal and convert it into time-frequency spectrograms. Then based on the concept of transfer learning strategy, convolutional neural networks are applied as feature extractors to classify flow patterns by the spectrogram images. The proposed method is verified by the visualized flow boiling experiment data. The results show that the adaptive time-frequency algorithm can effectively reflect the characteristics of different flow pattern signals, and several chosen neural network models show high recognition accuracy after training. Among them, VGG-16 network with small convolution kernels and strong transferability has the highest recognition rate. In addition, the network based on data of static conditions remains identifying more than 75% spectrograms of rolling conditions, exhibiting the generalization ability of the method under different flow conditions.     

Sectional void fraction measurement of gas-water two-phase flow by using a capacitive array sensor

The sectional void fraction measurement for multiphase flow is usually influenced by flow patterns. Inspired by electrical capacitance tomography (ECT) devices applied to flow imaging (whose measured capacitance data contain both the flow pattern and sectional void fraction information), a capacitive array sensor is developed to realize two functions, flow pattern recognition and void fraction measurement, simultaneously; so that the void fraction measurement can be conducted for a certain flow pattern and the measurement accuracy can be expected to be improved. The main idea of the proposed method can be described as: firstly, the proper feature vectors are extracted from the electrical signal to identify the flow pattern (the BPNN model with GDX learning algorithm is used for flow pattern identification); and then the average of electrical signal is applied to estimates the void fraction by the corresponding calibration curve. An experimental platform of air/water two-phase flow is built (on which 3 flow patterns can be generated stably) to test the performance of the proposed method. The results support the correctness and effectiveness of the proposed method.     

基于遗传算法/神经网络组合技术的气液两相流型辨识

通过采用遗传算法训练BP神经网络、优化网络权值的技术,对气液两相流的流型进行了辨识研究,在此基础上建立了基于遗传算法/神经网络组合技术的气液两相流流型的预测模型,从而为发动机轴承腔内润滑油气液两相流流型识别提供了技术支持,也为考虑轴承腔气液两相流的相关设计和实验工作提供了技术条件。     

Phase discrimination and a high accuracy algorithm for PIV image processing of particle–fluid two-phase flow inside high-speed rotating centrifugal slurry pump

PIV technology is an efficient and powerful measurement method to investigate the characteristics of fluid flow field. But for PIV particle image post-processing, some problems still exit in two-phase particles discrimination and velocity field algorithm, especially for high-speed rotating centrifugal slurry pump. In this study, through summarization and comparison of the various phase discrimination methods, we proposed a two-phase identification method based on statistics of gray-scale level and particle size. The assessment of performance through experimental PIV images shows that a satisfying effect for particle identification. For high speed rotation of the impeller, a combination of adaptive cross-correlation window deformation algorithm and multistage grid subdivision is presented. The algorithm is applied to experimental PIV images of solid–liquid two-phase flow in a centrifugal slurry pump, the results show that the algorithm in the present study has less pseudo vector number and more matching particle pairs than those of fixed window and window translation methods, having the ability to remove pseudo vector efficiently. It confirmed that the algorithm proposed in the present study has good performance and reliability for PIV image processing of particle–fluid two-phase flow inside high-speed rotating centrifugal slurry pump.     

Identification of two-phase flow pattern in porous media based on signal feature extraction

The statistical analysis methods based on differential pressure signals of two-phase flow are employed in the present study to identify the flow patterns in packed porous bed. The typical flow pattern images of two-phase flow in the packed porous beds are recognized and the corresponding differential pressure signals are recorded based on the visualization experiments. Then the statistical analysis methods, including probability density function (PDF), power spectral density (PSD), and wavelet energy spectrum (WES), are employed to extract the features of differential pressure signals in the time domain, frequency domain, and time-frequency domain respectively. The dimensionless parameters are proposed as the evaluation index to quantify the differences among flow patterns. The results show that the PDF, PSD, and WES methods can effectively characterize different flow patterns in the time, frequency, and time-frequency domain, respectively. The comprehensive recognition efficiency is about 88.5% using the introduced dimensionless parameters.     

Flow visualization of oscillation characteristics of liquid and vapor flow in the oscillating capillary tube heat pipe

The two-phase How patterns for both non-loop and loop type oscillating capillary tube heat pipes (OCHPs) were presented in this study. The detailed flow patterns were recorded by a high-speed digital camera for each experimental condition to understand exactly the operation mechanism of the OCHP. The design and operation conditions of the OCHP such as turn number, working fluid, and heat flux were varied. The experimental results showed that the representative flow pattern in the evaporating section of the OCHP was the oscillation of liquid slugs and vapor plugs based on the generation and growth of bubbles by nucleate boiling. As the oscillation of liquid slugs and vapor plugs was very speedy, the How pattern changed from the capillary slug flow to a pseudo slug flow near the annular flow. The flow of short vapor-liquid slug-train units was the flow pattern in the adiabatic section. In the condensing section, it was the oscillation of liquid slugs and vapor plugs and the circulation of working fluid. The oscillation flow in the loop type OCHP was more active than that in the non-loop type OCHP due to the circulation of working fluid in the OCHP. When the turn number of the OCHP was increased, the oscillation and circulation of working fluid was more active as well as forming the oscillation wave of long liquid slugs and vapor plugs in the OCHP. The oscillation flow of R-142b as the working fluid was more active than that of ethanol and the high efficiency of the heat transfer performance of R-I42b was achieved.     

Identification of two-phase flow regime using ultrasonic phased array

Flow regime is one of the key characteristics of gas-liquid two-phase pipe-flows and its identification is essential for several industrial applications. In this paper, the ultrasonic phased array technology is used to identify flow regimes of two-phase (air-water) vertical flow. The ultrasonic phased array can perform multi-point, omnidirectional detection to obtain high-resolution data suitable for image processing. The scanned images, which have distinctive features, are subjected to a series of image-treatment techniques, such as principle component analysis, to extract information necessary for flow regime identification. The K-nearest neighbors (KNN) classification algorithm is then used to identify flow regimes with high accuracy.     

Hilbert–Huang transform based signal analysis for the characterization of gas–liquid two-phase flow

This paper reports the application of the Hilbert–Huang Transform (HHT) to the dynamic characterization of gas–liquid two-phase flow in a horizontal pipeline. A differential pressure fluctuation signal of gas–liquid two-phase flow is adaptively decomposed into Intrinsic Mode Functions (IMFs) through the use of Empirical Mode Decomposition (EMD) methods. Based on the EMD, the associated time–frequency–energy distribution, i.e., the Hilbert spectrum, is obtained for the analysis of the differential pressure fluctuation signal and subsequent identification of its corresponding energy characteristics. The relationship between the energy distribution of the signal and the flow pattern is established. In order to assess the effectiveness of the approach, the results obtained using the HHT are compared with those from Fourier analysis and wavelet based methods. It is found that the extracted energy characteristics give a good indication of the dynamic state of the gas–liquid two-phase flow and thus can be used for flow pattern recognition. The proposed method is a useful tool for the in-depth understanding and subsequent quantitative characterization of gas–liquid two-phase flow.