Thank you for your responses - I will look into the tutorials - good tip. Here is a bit more context. I am using the WALE-model - there is no k_sgs right? So regarding the mesh - i tried a mesh with double as many cells in my bend setup and it didnt change result significantly - but i will run a finer mesh for the straight pipe as for consistency. Regarding the sampling period. The length is 10D (D=0.02)- ive simulated around 70 passes throughs for this (150s). I will look into time-integral scale as well it alone will not fix the problem. For Re=20.000, Pr=0.5, we would expect Nu=42,5. https://cdn.discordapp.com/attachments/598785402391887872/1475797948209172561/image.png?ex=699ecb71&is=699d79f1&hm=d71c31e46e773670c03eee5f9e2be95215684fe453b78c5b7b9b7a8c6f999fc7&