Convergence Issue of Rotor67 Simulation under Turbo Mode in SU2

LinShiyan · April 14, 2024, 07:27

I am writing to seek assistance regarding my endeavors with the Turbomachinery mode in version 8.0.1 of SU2. Specifically, I am attempting a single-channel RANS/LES simulation on Rotor67 using the EDDES method I have utilized Autogrid to generate a structured grid with approximately 11 million cells.

Currently, I am encountering an issue. Upon reaching the peak efficiency point of the rotor, I conducted a RANS calculation initially, which seemed normal as depicted in Figure 1(Iso-surface of Q criterion = 1×10^8). However, when transitioning to an unsteady simulation using EDDES, after 5000 steps, the blade surfaces exhibit significant non-physical fluctuations, as shown in Figure 2(Iso-surface of Q criterion = 1×10^8). In an attempt to troubleshoot this problem, I started by examining the convection scheme. I have employed the JST scheme with second and fourth-order coefficients set to 0.5 and 0.05, respectively. Unsure if this is the cause, I attempted to reduce dissipation by employing the Roe+MUSCL format. Strangely, under this format, RANS calculations fail to converge despite adjusting all parameters I deemed adjustable, including:

1. Reducing CFL to 0.1
2. Adjusting the Entropy_fix_coeff to 0.1~0.15
3. Setting SLOPE_LIMITER_FLOW to either VENKATARISHNAN_WANG or VAN_ALBADA_EDGE and adjust the VENKAT_LIMITER_COEFF to 0.05~0.1
All of these adjustments resulted in divergence.

Hence, I seek guidance on the following aspects:

1. In unsteady simulations, particularly in high-precision numerical simulations, how can I adjust the coefficients of JST to minimize non-physical dissipation?
2. Does someone completed convergence studies with ROE+MUSCL? If so, how can I adjust parameters to enhance the convergence of the Roe+MUSCL format specifically for RANS?

Figure1

Figure2

3. I have attached my cfg file in the attachment.
https://github.com/Linnnnnn23/SU2-Question/issues/1

Tongtong1919 · April 27, 2024, 04:27

Hi Lin, I looked at the uploaded cfg file and in CUSTOM_OUTPUTS, there is an item VELOCITY_ZDENSITY which seems to be used to define the import/export mass flow rate, I would like to ask if VELOCITY_ZDENSITY is a built-in expression in SU2, it seems that I can't recognize this when I am using the v8.0.0 turbo branch.

April 14, 2024, 07:27	Convergence Issue of Rotor67 Simulation under Turbo Mode in SU2	#1
LinShiyan New Member LinShiyan Join Date: Jun 2023 Posts: 1 Rep Power: 0	I am writing to seek assistance regarding my endeavors with the Turbomachinery mode in version 8.0.1 of SU2. Specifically, I am attempting a single-channel RANS/LES simulation on Rotor67 using the EDDES method I have utilized Autogrid to generate a structured grid with approximately 11 million cells. Currently, I am encountering an issue. Upon reaching the peak efficiency point of the rotor, I conducted a RANS calculation initially, which seemed normal as depicted in Figure 1(Iso-surface of Q criterion = 1×10^8). However, when transitioning to an unsteady simulation using EDDES, after 5000 steps, the blade surfaces exhibit significant non-physical fluctuations, as shown in Figure 2(Iso-surface of Q criterion = 1×10^8). In an attempt to troubleshoot this problem, I started by examining the convection scheme. I have employed the JST scheme with second and fourth-order coefficients set to 0.5 and 0.05, respectively. Unsure if this is the cause, I attempted to reduce dissipation by employing the Roe+MUSCL format. Strangely, under this format, RANS calculations fail to converge despite adjusting all parameters I deemed adjustable, including: 1. Reducing CFL to 0.1 2. Adjusting the Entropy_fix_coeff to 0.1~0.15 3. Setting SLOPE_LIMITER_FLOW to either VENKATARISHNAN_WANG or VAN_ALBADA_EDGE and adjust the VENKAT_LIMITER_COEFF to 0.05~0.1 All of these adjustments resulted in divergence. Hence, I seek guidance on the following aspects: 1. In unsteady simulations, particularly in high-precision numerical simulations, how can I adjust the coefficients of JST to minimize non-physical dissipation? 2. Does someone completed convergence studies with ROE+MUSCL? If so, how can I adjust parameters to enhance the convergence of the Roe+MUSCL format specifically for RANS? Figure1 Figure2 3. I have attached my cfg file in the attachment. https://github.com/Linnnnnn23/SU2-Question/issues/1

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
SU2 Simulation of Rotor in Hover	Bakhshi	SU2	0	November 11, 2023 08:07
Convergence issue on NACA0012 simulation	sandrobrad	OpenFOAM	3	October 4, 2021 06:17
Issue with turbulent viscosity contour - SU2 v7.0.6	GA_Onera	SU2	4	June 28, 2021 09:11
Solver parameters to solve convergence issue	WilliamH	SU2	5	February 5, 2021 06:19
Restart CFX simulation in batch mode using ccl	luk1	CFX	2	December 29, 2020 15:37

April 27, 2024, 04:27		#2
Tongtong1919 New Member Tongtong Join Date: Jan 2024 Posts: 4 Rep Power: 2	Hi Lin, I looked at the uploaded cfg file and in CUSTOM_OUTPUTS, there is an item VELOCITY_ZDENSITY which seems to be used to define the import/export mass flow rate, I would like to ask if VELOCITY_ZDENSITY is a built-in expression in SU2, it seems that I can't recognize this when I am using the v8.0.0 turbo branch.