Experimental And Simulation Study On Multiphase Flow In Quench Chamber Of Dry Pulverized Coal Gasifier

Dry pulverized coal entrained bed gasification is currently the most widely used type of coal gasification bed with the highest gasification efficiency.As a component of the gasification furnace,the quench chamber has the function of washing and cooling the synthesis gas,and is a key equipment to ensure the smooth progress of the subsequent process.When the quench chamber is in industrial operation,problems such as excessive water taken out,false liquid level,serious ash accumulation,and poor washing effect will occur.Parameters such as the apparent gas velocity,the height of the separation space,the distribution of gas holdup and the velocity distribution of gas and liquid phases affect the operation of the quench chamber.This paper analyzes the influence of the above parameters on the operation of the quench chamber with internal components through two methods:cold model experiment and numerical simulation.It will also try to use data-driven methods to solve the flow field of the quench chamber.The specific work is as follows:(1)Using the cold-state modeling experiment method,a cold-mold experimental platform was built according to the ratio of 1:7 to the industrial device,and the influencing factors of droplet entrainment were quantitatively measured.The study found that for the bubble-breaking strip structure quench chamber,the bubble-breaking strip has little blocking effect on droplets,and the droplet entrainment fraction E of the gas at the outlet has an exponential relationship with the dimensionless parameter u'g/h',E(h,ug)=1.756E10(u'g/h')3.569;when the wind speed is the same,the rising tube deflects The droplet entrainment rate E of the plate structure is only that of the bubble-breaking strip structure.There is a critical liquid level and critical gas velocity in the bubble-breaking strip structure.When the gas velocity is 4.68m/s,the gas velocity exceeds the critical gas velocity,and the main body of the liquid phase splashes directly to the outlet;with the change of the gas velocity,the internal components of the bubble-breaking strip The liquid level fluctuates between 3-10cm,and the external liquid level of the riser's internal components fluctuates between 2-4cm;when the static liquid level is constant,under three wind speed loads(1.17m/s,2.34m/s,3.51m/s)The washing effects of the two internal components are relatively close,both are between 94%and 98%,and the washing efficiency of the riser baffle structure is slightly higher.(2)Use Fluent to simulate the gas-liquid two-phase flow in the quench chamber of the bubble-breaking strip structure.It is simulated that the first airflow hits the farthest distance at 0.3s,and then the airflow moves upwards to form a huge bubble.After 0.4s,the gas-liquid interface contacts the bubble breaker,and the bubble breaker breaks the large bubble,and the gas phase Diffusion in the radial direction,the whole process lasts for a total of 0.5s;after 0.9s,the impact depth is basically maintained at about 17mm;the gas-liquid two-phase is between the four-layer bubble-breaking strips,and the degree of mixing gradually increases as the axial height increases;Below 0.5m,the higher the wind speed,the higher the gas holdup.Above 0.7m,the higher the wind speed,the lower the gas holdup;the gas speed gradually increases along the height of the quench chamber and changes at the position of the broken bubble Severely,below the third layer of bubble-breaking strips,the gas basically does not diffuse radially;the axial liquid velocity is obviously lower than the axial gas velocity,and the liquid phase movement is mainly radial.(3)Constructed two neural network models:BP neural network and physical information neural network.Using the Simple difference method,discrete Navier Stokes equations,extracting 15 inputs and 2 outputs of each step,a total of 800,000 sets of data are used for training,and the square cavity flow is successfully calculated using the trained BP neural network;Add the step flow data to the training,use the BP neural network to simultaneously calculate the two flow fields of the square cavity flow and the step flow.It is found that the network has a certain degree of robustness and can calculate the variable boundary step flow,but it cannot calculate the complicated The flow field of the quench chamber.The Navier Stokes equation was embedded in the neural network,a PINN with 10 neural layers was constructed,and 1%of the flow field data was used as the training set to successfully reconstruct the local velocity field of the quench chamber,the velocity field u,The average relative errors of v are 4.14%and 2.83%,respectively.