Study On Flow Monitoring For Gas-liquid Two-phase Flow In Small Channel Based On Curvature Domain Optical Flow Algorithm

With the development of small-scale industrial equipment,the application demand for gas-liquid two-phase flow in small channel is increasing gradually.The flow pattern of gas-liquid two-phase flow in small channel can influence the mechanical characteristics such as mass and heat transfer,flow dynamics,etc.,and be the basis for other parameters accurate measurement.The image-based method is a common method for parameters measurement of gas-liquid two-phase flow in small channel.Most of the conventional methods perform digital image processing and machine learning technology on single image to analyze and identify the flow pattern of gas-liquid two-phase flow in small channel.However,the dynamic images are seldom utilized to analyze the motion information of the flow.Optical flow method is an effective method for dynamic image analysis.At present,most optical flow algorithm frameworks are based on the assumption of brightness invariance,which make the algorithm less robust to environmental illuminance,and the calculation accuracy of the motion of the phase interface of the two-phase flows is insufficient.In view of this,this paper proposes a new method for monitoring the gas-liquid two-phase flow in small channel based on optical flow in the curvature domain.The main work and contributions of this paper are as follows:1.Based on the characteristics of gas-liquid two-phase flow in small channel and the characteristics of optical flow calculation framework,a curvature domain optical flow new algorithm is proposed.The main idea of the method is as follows:first,the Nadaraya-Watson kernel regression method is used to smooth the dynamic image,then the curvature domain information is calculated according to the geometric meaning of the second basic form of the surface,and finally the optical flow field is calculated based on curvature domain basic structure invariance.Experiments on the Middlebury optical flow evaluation dataset demonstrate that the average angular error of our proposed algorithm is less than 5.93,and the standard angular error is less than 1.57.Experiments on dynamic images of gas-liquid two-phase flow in small channel show that our proposed algorithm can emphasize the motion information of the phase interface.2.A method of flow monitoring for gas-liquid two-phase flow in a small channel based on the curvature domain optical flow algorithm is proposed.The principal component analysis is utilized to extract flow features from the optical flow field which quantitatively describes the flow information of the gas-liquid two-phase flow in small channel,and then K-means clustering algorithm is performed on the flow features to complete flow monitoring effectively.The optical flow field obtained by the curvature domain optical flow algorithm with strong anti-noise and emphasis of motion details,quantitatively describes the flow pattern of gas-liquid two-phase flow in small channel for the convenience of flow monitoring.Utilizing principal component analysis to extract flow characteristics from the optical flow field and then implement K-means clustering algorithm,flow patterns including bubble flow,slug flow,laminar flow,and annular flow and the transition between these classic flow patterns can be identified effectively.The overall recognition accuracy is over 96%.3.A small-channel gas-liquid two-phase flow monitoring system is designed and developed,including a flow unit,a dynamic image acquisition unit,and a flow analysis unit.Flow monitoring experiments of the gas-liquid two-phase flow were carried out in a small horizontal channel with an inner diameter of 5.0mm.Experimental results show that the proposed method and the developed system can effectively monitor bubble flow,slug flow,laminar flow,annular flow and the transition patterns between the above classic flow pattern,and the overall identification accuracy is over 96%.