Ogee Spillway

luisbrandao · November 25, 2014, 12:55

Hello every body,

I have simulated 2D flow over an Ogee spillway by Flow-3D v11.
To validate the Flow-3D results (pressure values), I compared Flow-3D results with Hydraulic Design Chart (HDC) 111-16 prepared by U.S. Army Engineer Waterway Experiment Station. Please see attached files.
I have to mention that I generated a Neutral file to print the pressure values.

As you can see in attached file, There is a fluctuation in pressure values and also pressure values before the crest are much more than HDC results.

I would be thankful if anybody can help me to fix mentioned problems.

Best regards,
Luis

JBurnham · November 26, 2014, 17:55

Hi Luis - I've taken a look at your files. Your basic setup looks good, as does your .stl weir and analysis. There are possibly a couple of points that may be important:

1) The USACE pressure vs. distance chart doesn't make clear if the actual head H refers to the total head (H0 + Ha) or the head given only by the depth (H0). H0 is the depth from the crest to the free surface at point 4x or 5x P upstream of the crest. Ha is the approach velocity head = V^2/2g. The difference might make a difference. Check the definitions in the design sheets to make sure that H = H0 + Ha, because at the boundary the fluid is at rest (Hbc = H0 + 0). Once the fluid starts to accelerate, Ha > 0 and H0 < H.

2) The USACE pressure vs. distance chart doesn't make clear what the confidence of the line is. If the line is fit from many data sets, then it has inherent scatter that is not shown. It is not necessarily correct to say the pressure values before the crest are "much more" than HDC results without knowing the uncertainty of the HDC data.

3) My suggestion is to minimize numeric sources of error:

a) You could easily make this problem 2-D. In the mesh tree on Meshing and Geometry, under the Y direction, set Total Cells = 1. That will speed up your run times significantly. 2 cells in Y is not enough for a 3-D analysis anyway.

b) Go to Numerics > Convergence Controls. Set the Multiplier for Dynamic Convergence Criteria EPSADJ = 0.001 (1000x tighter pressure convergence). Set the minimum and maximum numbers of iterations ITMIN = 2 and ITMAX = 500. This will give a more precise pressure field. Does it change the results? If yes, keep the settings. If no, decide whether to keep or drop them.

c) Now still on the Numerics tab try 2nd-order, monotonicity-preserving momentum advection. Does it change the results for better or worse? If better, keep it. If worse, go back to 1st order.

d) Once you have picked pressure criteria and momentum advection, now start refining the grid. Make another run w/ the cell size 2x smaller in all directions. Then do it again. Keep doing it until the pressure results change by a very small amount: a few percent, probably.

When the pressure results stop changing, you can consider the mesh dependency study complete. Hopefully you will have more believable results. Note that this is not a true 'convergence study', as defined by ASME or ASCE, but it's enough to have confidence in your results.

Please let us all know what the results are after you've done a series of grid refinements and the other numerical tests. Better? Worse?

Cheers, - Jeff

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Flow over stepped spillway	ftiago	OpenFOAM Pre-Processing	1	March 15, 2016 11:35
scale difference in spillway modeling	flow_CH	FLOW-3D	5	May 27, 2013 04:48
Calculating flow rate through a section of a spillway	Indus	FLOW-3D	1	November 5, 2012 19:54
Flow Over a Spillway	Nicopep	OpenFOAM	1	October 21, 2009 05:48
Mesh for ogee spillway	a.etesam	FLOW-3D	1	August 3, 2009 16:39

November 26, 2014, 17:55		#2
JBurnham Senior Member Jeff Burnham Join Date: Apr 2010 Posts: 204 Rep Power: 17	Hi Luis - I've taken a look at your files. Your basic setup looks good, as does your .stl weir and analysis. There are possibly a couple of points that may be important: 1) The USACE pressure vs. distance chart doesn't make clear if the actual head H refers to the total head (H0 + Ha) or the head given only by the depth (H0). H0 is the depth from the crest to the free surface at point 4x or 5x P upstream of the crest. Ha is the approach velocity head = V^2/2g. The difference might make a difference. Check the definitions in the design sheets to make sure that H = H0 + Ha, because at the boundary the fluid is at rest (Hbc = H0 + 0). Once the fluid starts to accelerate, Ha > 0 and H0 < H. 2) The USACE pressure vs. distance chart doesn't make clear what the confidence of the line is. If the line is fit from many data sets, then it has inherent scatter that is not shown. It is not necessarily correct to say the pressure values before the crest are "much more" than HDC results without knowing the uncertainty of the HDC data. 3) My suggestion is to minimize numeric sources of error: a) You could easily make this problem 2-D. In the mesh tree on Meshing and Geometry, under the Y direction, set Total Cells = 1. That will speed up your run times significantly. 2 cells in Y is not enough for a 3-D analysis anyway. b) Go to Numerics > Convergence Controls. Set the Multiplier for Dynamic Convergence Criteria EPSADJ = 0.001 (1000x tighter pressure convergence). Set the minimum and maximum numbers of iterations ITMIN = 2 and ITMAX = 500. This will give a more precise pressure field. Does it change the results? If yes, keep the settings. If no, decide whether to keep or drop them. c) Now still on the Numerics tab try 2nd-order, monotonicity-preserving momentum advection. Does it change the results for better or worse? If better, keep it. If worse, go back to 1st order. d) Once you have picked pressure criteria and momentum advection, now start refining the grid. Make another run w/ the cell size 2x smaller in all directions. Then do it again. Keep doing it until the pressure results change by a very small amount: a few percent, probably. When the pressure results stop changing, you can consider the mesh dependency study complete. Hopefully you will have more believable results. Note that this is not a true 'convergence study', as defined by ASME or ASCE, but it's enough to have confidence in your results. Please let us all know what the results are after you've done a series of grid refinements and the other numerical tests. Better? Worse? Cheers, - Jeff