I have done a transient run with up to 24,000 timesteps to calculate pressure fluctuations on an aeroacoustic noise source. To complete the run, I had to reduce the number of coeff loops to 2 and set a memory integer stack to 3.5. Now what had happened is that toward the tail end of my run, the pressure fluctuatons get smaller and smaller until it looks like the whole thing has settled down to a smooth line. This should not be the case and i'm wondering if the stack function averages out the results causing me not to read further pressure differences. Please let me know if anyone out there knows if this is a side effect of the stack function. It shouldnt be the coeff loops as the cod is still converging to 10e-5.

It's hard to justify that memory stack function smears pressure fluctuation. Which model you use in your simulation, URANS, DES, or LES? When you talk about pressure fluctuation, what's the fluctuation magnitude compared with overall averaged pressure value? By the way, do you use small enough time step to capture noise signal?

Using DES simultaion, the magnitude of pressure fluctuation to start is approx +/- 10pa for the first 10000 timesteps with an average of 25pa once the run settles down. (the first few timesteps have very large pressure fluctuations but are not converged to my stated accuracy). Then in the last 10000 timesteps, fluctuations decrease to as low as 10e-5. I know its not correct as i have experimentally validated this run. And i am using small enough timesteps, it is duplicating my practical experiment using a sampling frequency of 48KHz.

What's the accumulated time of your simulation? In other words, what's the time elapsed in 24,000 timesteps? Bascially, the question comes back to whether the flow field arrives at a quasi-steady or developed state or not. If flow field is still developing, fluctuation may increase or dies down.

The 24,000 timesteps accumulates to 0.5 seconds. the flow becomes steady toward the end, in other words the transient part dies out. What i'm wondering is if it is caused by the averaged out Reynolds No.solution, or if the K-Epsilon model is killing it. as it is a very simple model, the flow is fully developed, that much i do know.

Since DES uses RANS to cover boundary layer, it's very possible pressure fluctuation dies out if it is orignated from BL and too much dissipation introduced by RANS. Have you checked that the switch from RANS to LES occurs at right location and LES is turned on at intended region?