Study On Bubbles Segmentation Reconstruction And Flow Pattern Identification Of Gas-Liquid Two-Phase Flow Based On Concave Detection And PCA With Mixed Dimension Reduction

The gas-liquid two-phase flow widely exists in industrial process,especially in the application of gas-liquid mass transfer or reaction equipment,such as distillation column,bubble column,spray column.The ratio of gas and liquid is one of the most important parameters of gas-liquid two-phase flow system.The gas-liquid ratio directly affects the quality and performance of product.Once the ratio changes,it will cause changes in flow resistance and flow stability,and thus lead to bad heat transfer crisis,and even affect the working performance of the system,resulting in damage to equipment.Therefore,it is necessary to study the gas-liquid ratio to control it.In order to solve these problems,a method for bubbles segmentation reconstruction and flow pattern of gas-liquid two-phase flow based on concave detection and PCA with mixed dimension reduction is proposed,and it is applied in the experiments to verify the validity.In the first chapter,the related research is summarized,including the image segmentation and reconstruction,multiple features extraction and dimension reduction,and image recognition and classification.This paper analyzes the problems existing in the image reconstruction and image recognition are analyzed,and the research content and framework are described.In the second chapter,a method for overlapped bubbles contour segmentation based on edge information is proposed by analyzing the characteristics of gas-liquid two-phase flow image with uneven gray level.The image is pre-processed based on Otsu method.The contour of bubbles is extracted by the Canny edge detection algorithm.Then,the region of interest of the image is extracted based on the Hough linear transform.The image is processed by median filtering and contrast enhancement,which lays the foundation for subsequent feature extraction.In the third chapter,a method for overlapped bubbles segmentation and contour reconstruction based on curve concave-convex transformation and circle fitting is proposed.Firstly,the bubble image is pre-processed.The Gauss filter is used to reduce noise.The foreground is separated by the adaptive threshold.So the pre-segmented image is obtained.Then,the contour is extracted by the Canny algorithm.The curve is smoothed by the Gaussian function.The distance between contour points and the centroid is calculated.The curve is analyzed by concave-convex transformation characteristic.And the segmentation path is selected for each of the two concave points.Finally,the bubble contour is reconstructed by using the least square circle fitting interpolation algorithm,and the bubble segmentation and reconstruction image is obtained.In the fourth chapter,a method for the flow pattern identification of gas-liquid two-phase flow based on PCA mixed features fusion is proposed.Through analyzing the feature of the image after pre-processing,three features are extracted including invariant moment,gray-level co-occurrence matrix and LBP.Three features are fused and reduced dimensionality based on the PCA technology.To a large extent,and the original information is not lost.The data dimension is reduced effectively.In the fifth chapter,the proposed methods are applied to the study on gas-liquid two-phase flow,containing overlapped bubbles segmentation and contour reconstruction,as well as the flow pattern identification.The performance comparison between the proposed algorithm and the Hough transform algorithm is realized.The comparison experiments are carried out based on multiple classifiers,between the mixed features and the other three separate features.The validity of the proposed method is verified by experiments.In the sixth chapter,the research results are summarized,and the future research work is prospected.