Dropping pressure turns negative in front of shock

JoshS · May 9, 2012, 10:01

Hey all, I have a fairly advanced problem. I am using a customized version of OVERFLOW to run ILES on a shock ramp. I'm using 5th-order HLLC on the RHS with WENO shock capturing, and a 4th-order Pade filter to handle aliasing. In front of the shock, which moves around due to the unsteady, turbulent flow, I invariably get bubbles of low pressure that eventually turn negative.

This is not a convergence issue--I've knocked the time step down low enough that I'm getting stable residuals way, way below the usual tolerance in my Newton iterations. Of course the problem can be fixed with artificial dissipation, but this completely screws up the location of the separation, the turbulent boundary layer thickness, and the shock wave, which are obviously not things I can sacrifice!

DOes anyone with experience in moving shockwaves know how to handle this?

JoshS · June 5, 2012, 20:06

I figured out my own problem. I'll post the answer here in case anyone googles this, because this is the first result for "iles 'pade filter' cfd."

It turns out that a high-order filter introduces Gibbs' phenomena across a shock wave. This should have been obvious, I guess, but I wasn't thinking too carefully about the math. What I did was introduce a fairly simple shock-detection routine to turn off the filter at the shock. This solved my problem.

May 9, 2012, 10:01	Dropping pressure turns negative in front of shock	#1
JoshS New Member Josh Strodtbeck Join Date: Apr 2012 Posts: 6 Rep Power: 14	Hey all, I have a fairly advanced problem. I am using a customized version of OVERFLOW to run ILES on a shock ramp. I'm using 5th-order HLLC on the RHS with WENO shock capturing, and a 4th-order Pade filter to handle aliasing. In front of the shock, which moves around due to the unsteady, turbulent flow, I invariably get bubbles of low pressure that eventually turn negative. This is not a convergence issue--I've knocked the time step down low enough that I'm getting stable residuals way, way below the usual tolerance in my Newton iterations. Of course the problem can be fixed with artificial dissipation, but this completely screws up the location of the separation, the turbulent boundary layer thickness, and the shock wave, which are obviously not things I can sacrifice! DOes anyone with experience in moving shockwaves know how to handle this?

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Boundary condition in SIPHONS - negative pressure	-	FLUENT	1	September 19, 2018 11:19
Pressure Outlet Guage pressure	Mohsin	FLUENT	36	April 29, 2016 18:16
Pressure BC for combustion chamber	Giuki	FLUENT	1	July 19, 2011 12:35
negative static pressure at inlet	nino	FLUENT	7	July 27, 2007 09:22
negative pressure	mAx	FLUENT	0	January 25, 2006 15:31

June 5, 2012, 20:06		#2
JoshS New Member Josh Strodtbeck Join Date: Apr 2012 Posts: 6 Rep Power: 14	I figured out my own problem. I'll post the answer here in case anyone googles this, because this is the first result for "iles 'pade filter' cfd." It turns out that a high-order filter introduces Gibbs' phenomena across a shock wave. This should have been obvious, I guess, but I wasn't thinking too carefully about the math. What I did was introduce a fairly simple shock-detection routine to turn off the filter at the shock. This solved my problem.