Dr. Chung: can you bring some graphs next time
Running on Steady mode
Using Pressure difference
Dr Chung: how is your grid independence going?
Dominic: It is taking a long time, maybe a more course mesh should be used
Adrian: You can save time by saving the result from the last grid independence test
Then refine the mesh and simulate again.
Dr. Chung: can you send around your results to the 3D group
Done a medium flow analysis to see the jet
Done cross sections around the narrowest part to compare how the speed changes
Waiting for it to properly converge
This is an unsteady simulation
Lost about 4 hours of simulation in Star CCM
Wanted to do one run on each of the upper airways models
To see how the flow changes across the three
And then go back to the particle simulation
Dr Chung: So the simulations should be the same between you and dominic
Dr Chung: have you worked out the Reynolds number?
Dr Chung: Pressure difference might be better for you
Need to choose the target speed to choose our target Reynolds number
Then you can increase the pressure difference and eventually we will have about 1000 reynolds number and hence turbulent
Getting some interesting grid independence tests on the trachea and bronchioles
Whilst I am varying the base size from a minimum to a maximum, some of the intermediate base sizes output an error during the simulation
Dr. Chung: Start from your fine mesh simulations and work back from that
CFD crash course: You solve partial differential equations. We give up trying to find the true solutions as they are nearly impossible (very hard) to find.
Instead we choose discrete points in the domain. Say I don’t want to know what is inbetween those points, but if you connect the values, you will have a clear idea of what is going on. If you have enough points you can assume it will change linearly between the points
This is why you need the grid independence test to see whether that certain amount of points is enough
By choosing discrete points you change partial differential equations to a series of simultaneous equations. I.e. ten points gives you 10 equations.
With three or four equations you can re-arrange the equations with the gauss-seidel method. However with like 4 million equations you cannot do this. And so we have to use course mesh solutions to get started and then refine to get convergence.
Adrian: Using inlet velocity boundary conditions now.
Dr Chung: Because we don’t know the velocity profile, it is better to use a pressure inlet rather than a velocity inlet.
Adrian: Prism layer to base size 200%
Dr. Chung: Plot your mesh on the cross sectional areas and then see how much is covered by the prism layers. Then plot the velocity profile there and see if there are any sharp changes.
So just plot your mesh at many different locations
A paper that Tom found last term will be used to model the affect of angle changes on pressure drop.
Tom will also manually find the angle changes for select pathways of the stl file.
Tom will make a matlab model that encorporates all of the mathematical relations that we have found so far
I have been working on the algorithm
It is now all coded, however had trouble plotting the surface of the airways
I am currently trying to find the problem in the code
Charlie and Tom have been thinking about how the angle change data can be recorded to preserve the all important connectivity of the model
Daniel has been doing research
Daniel will be working with Tom to create this Matlab model