hi guys, I´m investigating radial fans and have problems matching my experimental data. One problem is that when I start the same def-file with different (interpolation) res-files, e.g. pressure rise is totally different calculated. How can this be??

I have ICEM-hexa grids and use SST, steady, isothermal. Timescales, boundarys and so on are identical.

best wishes adam

This simply means that you have multiple possible stable solutions. Which one you end up at depends upon the initial conditions. Basically it is hysteresis in your model. Look at how the two solutions differ. There is probably a separation zone in one which is not induced in the other.

-CycLone

Thanks, but this can´t be. I can reproduce my experimentally data (pressure rise) in one (or more) operating points (OP), whether I´m coming from a higer or lower level one the pressure curve. So CFX can´t calculate different results (25% pressure difference) for one OP depending on the initial OP, right?! So are there other possibilities?

best wishes adam

Is your solution converged? If it is not, then the solution (in this case - pressure) may show fluctuation between two result files.

both cases reach the same level of convergence. and the other monitorpoints (torque, force, density,...) differ only sparsely.

It is still hysteris; perhaps more numerical, not physical. Your modeling assumptions (periodic symmetry for instance), mesh (is it fine enough?) and convergence level can all contribute.

Note that physically you can see similar behavior, where the performance will depend on whether you sped up or slowed down the system.

-CycLone

I don´t have symmetries, but massflow-inlet, static pressure opening (entrain) and frozen-rotor-interfaces between the stationary and rotating domains. What is fine enough? Grid has ~2.6 million nodes in the rotating and ~1.4 million in the stationary domain. y+ range is 1 to 40. convergence level is the same.

Try clocking your interface (there is a clocking angle setting on the interface). You should do this when running Frozen Rotor anyhow. Alternatively you could run transient rotor stator and store the time average velocity field.

-CycLone

clocking interface? you mean rotational offset? anybody else an idea??