Model of a flow in intersecting microchannels Denis Semyonov LUT 2012
Content Objectives Motivation Model implementation Simulation Results Conclusion
Objectives A flow and a reaction model is required to support development of the novel plate microreactor, consisting of numerous intersecting channels. Geometry of the microreactor is subject to optimization Flow conditions inside the reactor should be known
Motivation Development of a new equipment faces multiple challenges Comprehensive outlook of the process is desired Model based on a Computational Fluid Dynamics (CFD) can give an information about a real system behavior. CFD model can include other more simple models, such as kinetics.
Model Implementation Task is very challenging, so none of the standard commercial CFD packages was able to handle it well The flow model is based on the open source Finite Volumes library OpenFOAM Built in standard solver interfoam was modified to make a solver suitable for the described problem Complex geometry was created by Fortran routine Computational mesh was made in ICEM-CFD package available from CSC Solver modifications aimed to reduce the dimension of the model (2D instead of 3D) Precise surface force resolution was also added to adequately resolve multiphase flow with free surface Reynolds number Re 20...2000 the flow can be considered laminar
Model implementation. Equations Physical formulation of the model includes Navier-Stokes equations: Continuity and momentum V =0 V 1 1 V V = p V F st V 1 1 V p+ν t = + F st ΔV + V t ρ ρ Volume of Fluid (VOF) method has been used to model two-phase flow with free surface. According to this method, one more equation for the indicator variable α has been added to the formulation: α + ( α V ) + ( α (V ) V r ) =0 1 α 1 V r =0 t t α=1 for the liquid phase and α=0 for the gas phase VOF method allows to use only one set of Navier-Stokes equations (as for a single phase) by introducing additional variable and the equation for it. This significantly reduces the computational cost.
Model implementation. Equations Density and viscosity were calculated as: ν=α νl +(1 α) ν g Surface tension force was defined as: F st =σ κ n ρ=α ρl +(1 α) ρ g where σ is the surface tension coefficient n κ= is the interface curvature α n = α is the unit vector normal to interface
Model Implementation. Geometry and mesh Only part of the real reactor was computed, because of the computational resources and time limitations The flow is assumed to be fully developed in a computational domain. Only ideal case can be computed In the real world the flow is not fully developed everywhere Only hydrodynamics is tested, no reaction and mass transfer considered Roughly 1/8th of the real reactor was computed Computational mesh consists of 780.000 to 1.000.000 elements
Simulation Simulation was carried out on a supercomputer cluster (louhi.csc.fi) 1 to 1.5 seconds of real time was simulated to get reliable flow statistics One simulation run takes 4 14 wall-clock days Parallel computing was used in simulation Computational domain was decomposed into 32, 64, or 128 partitions Parallelization was handled purely by OpenFOAM libraries
Results. Flow pattern Qualitative results comparison indicates that the model is able to resolve all flow phenomena with dimensions not smaller than the computational mesh length scale Mesh length scale was chosen accordingly to capture all major phenomena of gas-liquid interaction
Results. Velocity field Typical velocity is in the range 0.1...0.5 m/s Streamers with very high velocity often appear in the flow Velocity in the core of the streamer can be as high as 4 m/s or even higher Experimental measurements confirm these facts
Results. Candidate geometries A number of candidate geometries was tested to choose the best according to the optimization criteria A new prototype was built and tested based on the simulation results Optimization criteria maximal gas-liquid interfacial area
Results To ensure comparability with the experimental values, the same image processing technique was used to process both. Image processing was implemented in Matlab Gas volume fraction and gas-liquid interface were extracted from the images extract edges convert to binary Select appropriate region Direct gas-liquid interface only Assuming thin liquid film on catalyst elements
Results. Interfacial area
Results. Gas holdup
Results. Residence time ratio
Conclusion The model of a 2D non-stationary two-phase flow in a microreactor was developed and tested. The model is able to compute and predict macroscopic hydrodynamic parameters of twophase flow, such as the interfacial area and the gas holdup. Due to computing challenges, the model has been simplified, and coarse computational mesh was used, which leads to large deviations from the experimental data in some cases. Nevertheless, the model is applicable for optimization and gives correct trends. Model was successfully implemented for the shape optimization of the microreactor prototype Reaction will be added to the model when the experimental data will be available