The Parameter Detection Of Gas-Solid Two-Phase Flow Based On Images Processing And Optical-flow Technique

As one of most important forms in multiphase flow, gas-solid two-phase flowwidely exists in modern industrial process, such as non-ferrous metal, metallurgy,building material, electric power, chemical engineering and food industry, etc. For thecomplex characters of two-phase flow, it’s really difficult to accurately measure thoseflow parameters. Many detection techniques and methods of two-phase flow are stillat laboratory research stage, which is extremely incompact with universality oftwo-phase flow in the engineering field. Therefore, developing the new technique forthe measure and analysis of multi-phase flow parameters are vital to the analysis ofmoving mechanism of gas-solid two-phase flow, and to the design and operation ofrelated instruments.High-speed photography method is applied to the flow parameters measurementin fluidized bed, and achieves a visual, non-contact measurement, which does notinterfere with the normal operation of the production equipment. Moreover, in thehigh speed gas-solid two-phase flow, we can adjust the frequency of snapping toobtain more clear flow images. Therefore, it has a larger scope of application. Basedon a large amount of experimental data, image processing, artificial neural network,optical flow analysis method and MQD method have been employed in flowparameters detection. Flow image-based multi-parameters detection method isdiscussed systematically from the aspects of theory and experiment.Firstly, the experiments were conducted on gas-solid fluidized bed system andflow images were captured by a high speed photography system. The flow imagesseparately are five typical regimes of gas-solid two-phase flow fluidized bed whichbubbling bed, sluggling bed, turbulent bed, wall pressing flow and thin phase convey.First, combine filter method was used to eliminate noises in those separately images.Then, use optical flow technique to get continuous two frames images optical flow field, extract images dynamic textures by gray level co-occurrence matrix, regardedas input variable, and separately send those samples to elasticity BP neural net, Elmanneural net and BP neural net work optimized. Thus the image texture eigenvectors offlow regime were identified. The test results show the combination between dynamictextures and elasticity BP neural net can more effectively identify five typical regimesof gas-solid two phase flow fluidized bed. The whole identification accuracy is98%,opening up a new avenue of the flow pattern recognition.Furthermore, we introduce the analytic optical flow method into measurement offlow field, velocity field and vorticity field in gas-solid fluidized bed, and discussvelocity distribution of typical flow regime in fluidized bed. Quantitative draw spacedistribution curves of average rising, falling and overall velocity of different flowregime in vertical direction, we provide an effective diagnostic tool to analysemoving mechanism of gas-solid two-phase flow, and guide the design of relatedequipments. Therefore, flow filed map which acquired by optical flow method ismore approach to the reality than by MQD cross-correlation method, and also costless time. It can be applied to detecting the flow filed map of gas-solid two-phaseflow and analyzing its flow rules, we develop a new method in the measurement offlow parameters of gas-solid flow form the views of theory and technique.