基于深度学习的气液固三相反应器图像分析方法及应用

doi:10.11949/0438-1157.20191255

化工学报 ›› 2020, Vol. 71 ›› Issue (1): 274-282.DOI: 10.11949/0438-1157.20191255

基于深度学习的气液固三相反应器图像分析方法及应用

黄正梁^1,³(),王超^1,³,李少硕^1,³,杨遥^1,³(),孙婧元^1,³,王靖岱^1,²,阳永荣^1,²

1. 浙江省化工高效制造技术重点实验室，浙江杭州 310027
2. 浙江大学化学工程国家重点实验室，浙江杭州 310027
3. 浙江大学化学工程与生物工程学院，浙江杭州 310027

收稿日期:2019-10-23 修回日期:2019-10-30 出版日期:2020-01-05 发布日期:2020-01-05
通讯作者: 杨遥
作者简介:黄正梁（1982—），男，助理研究员，huangzhengl@zju.edu.cn
基金资助:
国家自然科学基金项目(91834303);国家杰出青年科学基金项目(21525627);国家自然科学基金创新研究群体项目(61621002)

Development and application of image analysis method based on deep-learning in gas-liquid-solid three-phase reactor

Zhengliang HUANG^1,³(),Chao WANG^1,³,Shaoshuo LI^1,³,Yao YANG^1,³(),Jingyuan SUN^1,³,Jingdai WANG^1,²,Yongrong YANG^1,²

1. Zhejiang Provincial Key Laboratory of Advanced Chemical Engineering Manufacture Technology, Hangzhou 310027, Zhejiang, China
2. State Key Laboratory of Chemical Engineering, Zhejiang University, Hangzhou 310027, Zhejiang, China
3. College of Chemical and Biological Engineering, Zhejiang University, Hangzhou 310027, Zhejiang, China

Received:2019-10-23 Revised:2019-10-30 Online:2020-01-05 Published:2020-01-05
Contact: Yao YANG

摘要/Abstract

摘要：

气液固三相反应器中复杂的颗粒背景给流动参数的图像检测带来巨大挑战。发展了一种基于深度学习的气液固三相反应器图像分析方法，包括采集图像、制作训练集、建立图像识别模型和提取流动参数四个步骤。采用全卷积神经网络，在学习率为0.005、训练次数为2000次、训练集大小超过400张图像的条件下，图像识别误差小于5%。利用该方法可以获取三相反应器中局部相含率（气相分数和液相分数）及其空间分布、时间序列等信息，再采用时域分析、频率分析、小波分析等分析方法提取二次参数，可用于流型识别、压降预测和气液分布的均匀性判别等。将该方法用于涓流床中流动参数的检测，结果表明，局部液相分数的时间序列信号及其功率谱、概率密度分布均能清晰地区分涓流、脉冲流、鼓泡流等典型流型；时间序列信号的均值、标准差、极差和概率密度分布曲线半峰宽等特征参数可用于确定流型边界；平均液相分数可以用于预测涓流区的压降，计算值与实验测量值的平均相对偏差约为15%；液相分数空间分布的标准差可用于表征涓流床中不同流型的气液分布均匀性。该方法为气液固三相反应器的研究提供了新的工具。

关键词: 多相流, 深度学习, 神经网络, 成像, 涓流床, 相含率

Abstract:

In gas-liquid-solid three-phase reactors, the great challenges for image detection of flow parameters have been caused by the complex background with particles. An image analysis method of gas-liquid-solid three-phase reactor based on deep-learning is developed, which includes four steps: collecting image, making training set, establishing image recognition model and extracting flow parameters. The full convolutional neural network algorithm was chosen to build the identification model, and when the learning rate was set as 0.005, the number of epochs was set as 2000 and the training set size was set as 400 images in this model, the identification error of the model is less than 5%. The flow parameters such as local phase fractions (gas fraction and liquid fraction) and its spatial distribution, time series signals were obtained by the method. Then time domain analysis, frequency analysis, wavelet analysis and other analysis methods were used to extract the quadratic parameters from the spatial distribution and time series signals. And the quadratic parameters could be used to identify flow regimes, predict pressure drop and detect the uniformity of gas-liquid distribution. Furthermore, the method was applied to the detection of flow parameters in a trickle bed. The results showed that the time-domain variation, power spectrum density and probability density distribution of liquid fractions were used to distinguish the trickle flow, pulse flow and bubble flow. Meanwhile, the flow regime boundary was detected by characteristic parameters such as mean, standard deviation, range and half-width of probability density distribution curve of liquid fractions. The pressure drop in the trickle flow regime was predicted by the average liquid fraction, and the average relative deviation between the theoretical and experimental measurements was about 15%. In addition, the standard deviation of liquid fractions in the spatial distribution was used to detect the uniformity of gas-liquid distribution of different flow regimes in the trickle bed. This method provides a new tool for the research of gas-liquid-solid three-phase reactors.

Key words: multiphase flow, deep-learning, neural networks, imaging, trickle bed, phase fraction

中图分类号:

TQ 021.9

黄正梁, 王超, 李少硕, 杨遥, 孙婧元, 王靖岱, 阳永荣. 基于深度学习的气液固三相反应器图像分析方法及应用[J]. 化工学报, 2020, 71(1): 274-282.

Zhengliang HUANG, Chao WANG, Shaoshuo LI, Yao YANG, Jingyuan SUN, Jingdai WANG, Yongrong YANG. Development and application of image analysis method based on deep-learning in gas-liquid-solid three-phase reactor[J]. CIESC Journal, 2020, 71(1): 274-282.

图/表 13

图1 基于深度学习的气液固三相反应器图像分析流程

Fig.1 Procedure of image analysis method based on deep-learning in gas-liquid-solid three-phase reactors

图2 图像的典型处理过程

Fig.2 Typical processing of original images

图3 气液固三相反应器图像分析方法提取的流动参数

Fig.3 Flow parameters extracted by image analysis method in gas-liquid-solid three-phase reactors

表1 不同学习率下图像识别模型训练准确率/%

Table 1 Training accuracy of model at various learning rates/%

学习率	训练次数
学习率	50	100	200	500	1000	2000
0.0001	62.0	72.4	73.3	77.8	84.1	88.8
0.001	78.0	81.2	85.7	89.9	91.6	91.9
0.005	86.2	87.7	90.4	92.9	94.1	95.3
0.01	84.4	85.4	86.7	87.0	90.1	91.4
0.1	61.8	62.3	61.3	63.1	61.5	61.9

表2 图像识别模型的平均相对偏差

Table 2 Relative deviation of model

训练集大小	平均相对偏差/%
50	20.6
100	16
150	12.1
200	9.1
250	7.2
300	5.1
350	4.4
400	4.1

图4 气液固三相反应器冷模实验装置1—水槽；2—泵；3—空气压缩机；4—缓冲罐；5—转子流量计；6—电磁流量计；7—固体料仓；8—固体进料管；9—气体分配器；10—液体喷嘴；11—矩形床；12—光源；13—高速相机；14—图像采集系统

Fig.4 Cold-model experimental system for gas-liquid-solid three-phase reactor

图5 涓流床中3种典型气液流型

Fig.5 Three typical gas-liquid flow behaviors in trickle bed(a) uG = 0.069 m·s-1, uL = 0.003 m·s-1; (b) and (c) uG = 0.069 m·s-1, uL = 0.022 m·s-1; (d) uG = 0.007 m·s-1, uL = 0.028 m·s-1

图6 涓流床中3种典型流型下局部液相分数随时间的变化

Fig.6 Variations of local liquid fraction with time under three typical flow regimes in trickle bed

图7 涓流床中3种典型流型下局部液相分数信号的功率谱图

Fig.7 Variations of PSD of local liquid fractions under three typical flow regimes in trickle bed

图8 涓流床中3种典型条件下局部液相分数信号的概率密度分布曲线

Fig.8 Variations of probability density curve of liquid fractions under three typical flow regimes in trickle bed

表3 涓流床中3种典型流型对应的特征参数

Table 3 Characteristic parameters corresponding to three typical flow regimes in trickle bed

流型

液相分数

均值/ %

液相分数

标准差/ %

液相分数

极差/ %

半峰宽/%

涓流（u_G = 0.069 m·s^-1, u_L = 0.003 m·s^-1）

脉冲流（u_G = 0.069 m·s^-1, u_L = 0.022 m·s^-1）

鼓泡流（u_G = 0.007 m·s^-1, u_L = 0.028 m·s^-1）

图9 涓流床压降的实验值与理论预测值的比较

Fig.9 Comparison of experimental pressure drop with theoretical predictions from model in trickle flow

图10 涓流床中3种典型流型下局部液相分数空间分布的标准差随时间的变化

Fig.10 Variation of standard deviation of spatial distribution of local liquid fractions under three typical flow regimes in trickle bed

参考文献 40

1	Nigam K D P, Larachi F. Process intensification in trickle-bed reactors[J]. Chemical Engineering Science, 2005, 60(22): 5880-5894.
2	Kantarci N, Borak F, Ulgen K O. Bubble column reactors[J]. Process Biochemistry, 2005, 40(7): 2263-2283.
3	Ng K M. A model for flow regime transitions in cocurrent down-flow trickle-bed reactors[J]. AIChE Journal, 1986, 32(1): 115-122.
4	Ranade V V. Trickle Bed Reactors: Reactor Engineering and Applications[M]. Chaudhari R V, Ed. Amsterdam: Elsevier, 2011: 285.
5	Sai P S T, Varma Y B G. Flow pattern of the phases and liquid saturation in gas-liquid concurrent downflow through packed beds[J]. The Canadian Journal of Chemical Engineering, 1988, 66(3): 353-360.
6	Sato Y, Hirose T, Takahashi F, et al. Flow pattern and pulsation properties of cocurrent gas-liquid downflow in packed beds[J]. Journal of Chemical Engineering of Japan, 1973, 6(4): 315-319.
7	Fukushima S, Kusaka K. Gas-liquid mass transfer and hydrodynamic flow region in packed columns with cocurrent upward flow[J]. Journal of Chemical Engineering of Japan, 1979, 12(4): 296-301.
8	Melli T R, Desantos J M, Kolb W B, et al. Cocurrent downflow in networks of passages-microscale roots of macroscale flow regimes[J]. Industrial & Engineering Chemistry Research, 1990, 29(12): 2367-2379.
9	Benkrid K, Rode S, Pons M N, et al. Bubble flow mechanisms in trickle beds—an experimental study using image processing[J]. Chemical Engineering Science, 2002, 57(16): 3347-3358.
10	Jo D, Revankar S T. Bubble mechanisms and characteristics at pore scale in a packed-bed reactor[J]. Chemical Engineering Science, 2009, 64(13): 3179-3187.
11	Gunjal P R, Kashid M N, Ranade V V, et al. Hydrodynamics of trickle-bed reactors: experiments and CFD modeling[J]. Industrial & Engineering Chemistry Research, 2005, 44(16): 6278-6294.
12	Urseanu M I, Boelhouwer J G, Bosman H J M, et al. Induced pulse operation of high-pressure trickle bed reactors with organic liquids: hydrodynamics and reaction study[J]. Chemical Engineering and Processing: Process Intensification, 2004, 43(11): 1411-1416.
13	Kolb W B, Melli T R, Desantos J M, et al. Cocurrent downflow in packed beds. Flow regimes and their acoustic signatures[J]. Industrial & Engineering Chemistry Research, 1990, 29(12): 2380-2389.
14	Boelhouwer J G, Piepers H W, Drinkenburg A A H. Nature and characteristics of pulsing flow in packed bed reactors[J]. Chemical Engineering Science, 2002, 57(22/23): 4865-4876.
15	Zhao T, Eda T, Achyut S, et al. Investigation of pulsing flow regime transition and pulse characteristics in trickle-bed reactor by electrical resistance tomography[J]. Chemical Engineering Science, 2015, 130: 8-17.
16	Singh B K, Jain E, Buwa V V. Feasibility of electrical resistance tomography for measurements of liquid holdup distribution in a trickle bed reactor[J]. Chemical Engineering Journal, 2019, 358: 564-579.
17	Basavaraj M G, Gupta G S, Naveen K, et al. Local liquid holdups and hysteresis in a 2-D packed bed using X-ray radiography[J]. AIChE Journal, 2005, 51(8): 2178-2189.
18	Anadon L D, Sederman A J, Gladden L F. Mechanism of the trickle-to-pulse flow transition in fixed-bed reactors[J]. AIChE Journal, 2006, 52(4): 1522-1532.
19	Sederman A J, Gladden L F. Transition to pulsing flow in trickle-bed reactors studied using MRI[J]. AIChE Journal, 2005, 51(2): 615-621.
20	Hamidipour M, Larachi F, Ring Z. Cyclic operation strategies in trickle beds and electrical capacitance tomography imaging of filtration dynamics[J]. Industrial & Engineering Chemistry Research, 2010, 49(3): 934-952.
21	Honda G S, Lehmann E, Hickman D A, et al. Effects of prewetting on bubbly- and pulsing-flow regime transitions in trickle-bed reactors[J]. Industrial & Engineering Chemistry Research, 2015, 54(42): 10253-10259.
22	Schneider C A, Rasband W S, Eliceiri K W. NIH Image to ImageJ: 25 years of image analysis[J]. Nature Methods, 2012, 9(7): 671-675.
23	LeCun Y, Bengio Y, Hinton G. Deep learning[J]. Nature, 2015, 521(7553): 436-444.
24	Voulodimos A, Doulamis N, Doulamis A, et al. Deep learning for computer vision: a brief review[J]. Computational Intelligence and Neuroscience, 2018, 2018: 1-13.
25	Schmidhuber J. Deep learning in neural networks: an overview[J]. Neural networks, 2015, 61: 85-117.
26	Shaw G L. Donald Hebb: the organization of behavior[J]. Brain Theory, 1986, 17(3): 231-233.
27	Arulkumaran K, Deisenroth M P, Brundage M, et al. Deep reinforcement learning a brief survey [J]. IEEE Signal Processing Magazine, 2017, 34(6): 26-38.
28	Rumelhart D E, Hinton G E, Williams R J. Learning representations by back-propagating errors[J]. Nature, 1986;323(6088): 533-536.
29	Hinton G E, Osindero S, Teh Y W. A fast learning algorithm for deep belief nets[J]. Neural Computation, 2006,18(7): 1527-1554.
30	Ma R H, Zeng X Q. A review of deep learning research[J]. KSII Transactions on Internet and Information Systems, 2019, 13(4): 1738-1764.
31	Ciresan D, Meier U, Schmidhuber J. Multi-column deep neural networks for image classification[J]. IEEE Conference on Computer Vision and Pattern Recognition, 2012: 3642-3649.
32	Litjens G, Kooi T, Bejnordi B E, et al. A survey on deep learning in medical image analysis[J]. Medical Image Analysis, 2017, 42: 60-88.
33	Taigman Y, Yang M, Ranzato M A, et al. DeepFace: closing the gap to human-level performance in face verification[C]//IEEE Conference on Computer Vision and Pattern Recognition.2014: 1701-1708.
34	Ouyang W, Chu X, Wang X. Multi-source deep learning for human pose estimation[C]//IEEE Conference on Computer Vision and Pattern Recognition. 2014: 2337-2344.
35	Paisitkriangkrai S, Sherrah J, Janney P, et al. Semantic labeling of aerial and satellite imagery[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2016, 9(7): 2868-2881.
36	Maggiori E, Tarabalka Y, Charpiat G, et al. Convolutional neural networks for large-scale remote-sensing image classification[J]. IEEE Transactions on Geoscience and Remote Sensing, 2017, 55(2): 645-657.
37	Shelhamer E, Long J, Darrell T. Fully convolutional networks for semantic segmentation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(4): 640-651.
38	Levec J, Grosser K, Carbonell R G. The hysteretic behavior of pressure drop and liquid holdup in trickle beds[J]. AIChE Journal, 1988, 34(6): 1027-1030.
39	Holub R A, Dudukovic M P, Ramachandran P A. Pressure drop, liquid holdup, and flow regime transition in trickle flow[J]. AIChE Journal, 1993, 39(2): 302-321.
40	Macdonald I F, El-Sayed M S, Mow K, et al. Flow through porous media-the ergun equation revisited[J]. Industrial & Engineering Chemistry Fundamentals, 1979, 18(3): 199-208.

[1]	金伟其, 吴月荣, 王霞, 李力, 裘溯, 袁盼, 王铭赫. 化工园区工业气体泄漏气云红外成像检测技术与国产化装备进展[J]. 化工学报, 2023, 74(S1): 32-44.
[2]	宋瑞涛, 王派, 王云鹏, 李敏霞, 党超镔, 陈振国, 童欢, 周佳琦. 二氧化碳直接蒸发冰场排管内流动沸腾换热数值模拟分析[J]. 化工学报, 2023, 74(S1): 96-103.
[3]	温凯杰, 郭力, 夏诏杰, 陈建华. 一种耦合CFD与深度学习的气固快速模拟方法[J]. 化工学报, 2023, 74(9): 3775-3785.
[4]	诸程瑛, 王振雷. 基于改进深度强化学习的乙烯裂解炉操作优化[J]. 化工学报, 2023, 74(8): 3429-3437.
[5]	闫琳琦, 王振雷. 基于STA-BiLSTM-LightGBM组合模型的多步预测软测量建模[J]. 化工学报, 2023, 74(8): 3407-3418.
[6]	杨越, 张丹, 郑巨淦, 涂茂萍, 杨庆忠. NaCl水溶液喷射闪蒸-掺混蒸发的实验研究[J]. 化工学报, 2023, 74(8): 3279-3291.
[7]	尹刚, 李伊惠, 何飞, 曹文琦, 王民, 颜非亚, 向禹, 卢剑, 罗斌, 卢润廷. 基于KPCA和SVM的铝电解槽漏槽事故预警方法[J]. 化工学报, 2023, 74(8): 3419-3428.
[8]	董明, 徐进良, 刘广林. 超临界水非均质特性分子动力学研究[J]. 化工学报, 2023, 74(7): 2836-2847.
[9]	徐野, 黄文君, 米俊芃, 申川川, 金建祥. 多源信息融合的离心式压缩机喘振诊断方法[J]. 化工学报, 2023, 74(7): 2979-2987.
[10]	郑志航, 马郡男, 闫子涵, 卢春喜. 提升管射流影响区内压力脉动特性研究[J]. 化工学报, 2023, 74(6): 2335-2350.
[11]	高学金, 姚玉卓, 韩华云, 齐咏生. 基于注意力动态卷积自编码器的发酵过程故障监测[J]. 化工学报, 2023, 74(6): 2503-2521.
[12]	黄磊, 孔令学, 白进, 李怀柱, 郭振兴, 白宗庆, 李平, 李文. 油页岩添加对准东高钠煤灰熔融行为影响的研究[J]. 化工学报, 2023, 74(5): 2123-2135.
[13]	贠程, 王倩琳, 陈锋, 张鑫, 窦站, 颜廷俊. 基于社团结构的化工过程风险演化路径深度挖掘[J]. 化工学报, 2023, 74(4): 1639-1650.
[14]	吴心远, 刘奇磊, 曹博渊, 张磊, 都健. Group2vec：基于无监督机器学习的基团向量表示及其物性预测应用[J]. 化工学报, 2023, 74(3): 1187-1194.
[15]	李新亚, 邢雷, 蒋明虎, 赵立新. 倒锥注气强化井下油水分离水力旋流器性能研究[J]. 化工学报, 2023, 74(3): 1134-1144.

基于深度学习的气液固三相反应器图像分析方法及应用

Development and application of image analysis method based on deep-learning in gas-liquid-solid three-phase reactor

RichHTML

PDF (PC)

可视化

被引次数

摘要/Abstract

引用本文

使用本文

图/表 13

参考文献 40

相关文章 15

编辑推荐

Metrics

本文评价