CAE simulations. The art of the possible.

What factors contribute to making CAE projects successful? In the company I work for, in 15 years, we have delivered hundreds of CAE projects to industrial companies (most of them in the turbomachinery field). Some projects were highly successful, while others were not. In this article, I would like to share a few thoughts and lessons learned about project-success patterns and engineering productivity in general.

The Game

In engineering, we don’t simulate reality. We merely simulate our understanding of reality with currently available resources. In other words, the game we play here is about turning our knowledge and effort (and talent) into the best possible results. CAE projects, like software, are never really finished. They only have their versions. At any point, you can make them better. When chasing accuracy, the effort grows exponentially, so it’s smart to recognize the point when the results are good enough. Having clear, measurable goals is critical to prevent infinite drift, resource waste, and disappointment. Well-defined goals help to navigate, focus, and prioritize. The game we play is all about goals and their feasibility with the available knowledge and effort.

Continue reading on CFD SUPPORT website >>> https://www.cfdsupport.com/cae-simulations-the-art-of-the-possible/

Component 5, (Nacelle_Top) and Component 7 (Nacelle_Bottom) are experiencing failure to mesh. Can you take a look? It might be obvious to you.

Here's a download link to the compressed project file for your review. Please see my comment for a better Component map image for reference.

Edit: I've even tried performing the meshing operation using Salome-Meca. Meshing just one stl fails with it too, but unlike TCAE, takes several hours to fail completely. The component 5 (fuchsia) is Nacelle_Top, the patch in question is named Solid_Top_Bell-Mouth_inlet_and_Duct_Surface_08_21_25a.

I suspect Nacelle_Bottom is going to be just as bad, or worse, than Nacelle_Top. It is slightly smaller in diameter, but it's a more complicated surface.

Thanks in advance,

David

The previous iteration of this TCAE project experimented with omitting a central component from the topology, the Duct_Separator, going from the Top_Fan rim (sometimes called a hub, in this case an open outlet interface), through which air not used by the Top_Fan flows from the Top_System_Suction outlet interface, and continues directly to the inlet interface of the Bottom_Fan open rim, skipping the Duct Separator. Confusing, I know, but the Bottom_Fan is upside-down compared to the Top_Fan, so it takes in air not from the shroud inlet interface, as in the case of the Top_Fan, but from the Bottom_Fan's open rim inlet interface.

This iteration is more like the actual setup, Suction, Top_Fan, Duct_Separator, Bottom_Fan, Nacelle, Outlet_Extension. Please see my comment for a more complete image of the Component Map.

Who knows? It might eliminate some setup errors seen previously.

Working through the Check for Errors exercise finally presented only this orange border in the Components Map. Other than TCAE Output messages, there's no textual description of what or where things have gone awry. I'm assuming it's displayed somewhere in the TCAE Output messages, but what should I be looking for? Is it because the component Duct Separator (Co6) is without a partner, present in the group, barely visible, but no interaction with any other?

Still working through the Check exercises. Narrowing the errors down (hopefully!).

Hello everyone,

We are a team of university students currently working on a project to replicate the CFD simulation of the DJI 9450 propeller. First of all, we would like to thank the content creators for the excellent "Acoustic Benchmark Propeller DJI 9450" study. It has been an invaluable resource for our work.

As we try to reproduce the simulation results, we've encountered some confusion regarding the benchmark target values and would be very grateful for any clarification from the community or the official team.

Issue 1: Discrepancy Between Two Reported Results

We have been referencing two main documents:

 [1] The formal PDF research report (Acoustic Benchmark of DJI 9450.pdf).

 [2] The web page report, available at: https://www.cfdsupport.com/download/report/DJI-9450-Propeller-Benchmark/tcfd/TCFDReport.html

We noticed a significant difference in the thrust and torque values at 6000 RPM between these two sources. Specifically:

 * The PDF Research Report [1] lists a target Thrust of 4.256 N and a target Torque of 0.066 N·m.

 * The Web Page Report [2]shows a computed Thrust of 3.438 N and a computed Torque of 0.058 N·m.

Furthermore, the simulation result graph in the PDF report (approx. 4.2 N) also differs from the web page report's result (3.44 N).[1] We suspect this might be due to differences in mesh density. Could anyone confirm if this understanding is correct, and which set of values should we consider the definitive target for our validation?

Issue 2: Question About the Source of the Benchmark Target Value

To delve deeper, we located the original experimental paper cited in the PDF report (Intaratep, N., et al., AIAA 2016-2873).[1, 1] However, upon reviewing this paper, we found that the measured experimental thrust for a single rotor at 6000 RPM is approximately 6.1 N (as read from Fig. 6).[1,1]

This experimental value is considerably different from the benchmark target of 4.256 N listed in the PDF report, which has caused us significant confusion.

Our specific questions are:

 * Could anyone clarify the source of the benchmark target values (Thrust = 4.256 N, Torque = 0.066 N·m) used in the PDF report? Why does this target differ from the ~6.1 N we found in the original experimental paper (AIAA 2016-2873)?

 * To help us perform a more accurate validation and learn from this work, would it be possible to get access to the mesh file (e.g., in OpenFOAM or MSH format) or the complete geometry file (STL/STEP) used for the high-fidelity simulation presented in the PDF report?

Any clarification would be immensely helpful for our academic research.

Thank you for your time and assistance.

Best regards

That I've gotten this far, in my estimation, is cause for celebration. 5-lime margarita time!

(they only yielded 4 oz. of juice from those small organic limes, but it'll do).

I'm amazed at the Check functionality. I've been eliminating one similar error after another until now.
From the directory for the component named "Top_Fan", I copied and pasted an interface patch (.stl) into the directory for the component named "Hull__Inlet__Outlet", same file name, same type: "outletInterface" with the intention to prevent this from happening. It seems to me that in that instance it should be an inletInterface (just saying, I tried it that way...didn't help), Static Reference frame, another, different component.
Setup Problem is displayed upon selecting Check, listing the source, the copied patch in Top_Fan as the problem.

When I go into the Top_Fan component to connect the patch to the patch of same name in the Hull__Inlet__Outlet component, I don't see it listed as a choice, nor do any of the available listed options seem to work.
Changing the name slightly of the file in Hull__Inlet__Outlet had no effect.

How can I fix this? I'm stuck. Your help is appreciated in advance.

David

Hi everyone,

I have a question concerning the mixing plane interface method : it seems to me that the method can generate numerical errors when reverse flow occurs on the interface.

To be more specific, I have a radial interface between the outlet of a centrifugal impeller and the inlet of its diffuser. On a specific zone of the interface, reverse flow occurs with a radial circumferential angle -180° < circAng < -90°, whereas on the rest of the interface, there is no reverse flow and 90° < circAng < 180°. In one of my former post on the forum, you confirmed to me that the weighted surface average of the circumferential angle might be distorted in this situation. It seems to me that the mixing plane method at the interface also encounter issues in this situation to estimate the value of circumferential angles for the inlet patch of the interface, in the strips where circAng oscillate between 90° < circAng < 180° and 180° < circAng < -90°. Indeed, I can see this kind of numerical errors when post-processing my simulations.

Do you have some recommendations to avoid this kind of problems ? For example, use specific configuration options, or different interface conditions ?

Thanks in advance !

Zacharie

A common inlet at the top feeds two (identical) centrifugal fans. I tried to connect Top_System_inlet_interface (identical to but different in name as suction outlet interface) to serve double-duty as outlet interface to both the Top_Fan (rotating) AND Bottom_Fan (rotating).

It seemed that wasn't working so it occurred to me that connecting a copy of the object with another name might. Sure enough! It seems to have worked.

How would you have done it?

Hi contributors,

I sincerely appreciate your team's efforts in making your work publicly available online—it has been immensely helpful for OpenFOAM users and developers alike. After reading your article (Node266 - CFD SUPPORT), I followed the provided instructions to configure OpenFOAM to generate a cellZone using snappyHexMesh.

Unfortunately, the attempt was unsuccessful. The number of cells in both the cellZone and faceZones files turned out to be zero, and checkMesh.log confirms this issue. When I check the result via paraFoam, I also find that net0 ~ net4 contain no mesh—they are empty. This is not only contrary to my expectations, but also differs from the result described in the article.

I am using OpenFOAM v2012, and the net0 section in my snappyHexMeshDict was set up following the example from your webpage.

I’ve uploaded the relevant case below. I would greatly appreciate any guidance.

--------------------------------------------------------------------------------------------------

snappyhexmeshdict:

/*--------------------------------*- C++ -*----------------------------------*\
| =========                 |                                                 |
| \\      /  F ield         | OpenFOAM: The Open Source CFD Toolbox           |
|  \\    /   O peration     | Version:  v2012                                 |
|   \\  /    A nd           | Website:  www.openfoam.com                      |
|    \\/     M anipulation  |                                                 |
\*---------------------------------------------------------------------------*/
FoamFile
{
    version     2.0;
    format      ascii;
    class       dictionary;
    object      snappyHexMeshDict;
}
// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
#include "../varsSettings" //Include global variables
castellatedMesh true;
snap            true;
addLayers       false;//true;
geometry
{
    floater
    {
        type 	triSurfaceMesh;
        file 	"DeepCwindPlatform_heave.stl";
    }
    
    net1
    {
        type 	triSurfaceMesh;
        file 	"Net1_improved_heave.stl";
    }
    
    net2
    {
        type 	triSurfaceMesh;
        file 	"Net2_improved_heave.stl";
    }
    
    net3
    {
        type 	triSurfaceMesh;
        file 	"Net3_improved_heave.stl";
    }
    
    net4
    {
        type 	triSurfaceMesh;
        file 	"Net4_improved_heave.stl";
    }
    
    net0
    {
    	type	searchableBox;
    	min		(-100	-50	-120);
    	max		(100	50	-80);
    }
    
    refineFreeSurface_L1
    {
        type searchableBox;
        min ($minX  $minY -10);
        max ($maxX  $maxY  10); 
    }
    
    refineFreeSurface_L2
    {
        type searchableBox;
        min (#calc "$minX+$inLen+20"  -70 -8);
        max (#calc "$maxX-$outLen-20"  70  8); 
    }
    
    refineFloater_L1
    {
        type searchableBox;
        min (-51  -49	#calc "-30 + $decay_z + $offset_z");
        max ( 37   49 	#calc "22 + $decay_z + $offset_z"); 
    }
};
castellatedMeshControls
{
    maxLocalCells 100000;
    maxGlobalCells 2000000;
    minRefinementCells 0;
    maxLoadUnbalance 0.10;
    nCellsBetweenLevels 1;
    features
    (
        {
        	file    "DeepCwindPlatform_heave.eMesh";
            level   4;
        }
        
		{
        	file	"Net1_improved_heave.eMesh";
        	level	4;
		}
		
		{
        	file	"Net2_improved_heave.eMesh";
        	level	4;
		}
		
		{
        	file	"Net3_improved_heave.eMesh";
        	level	4;
		}
		
		{
        	file	"Net4_improved_heave.eMesh";
        	level	4;
		}
    );
    refinementSurfaces
    {
		floater
		{
			level (4 4);//(5 5);
			patchInfo
        	{
          		type wall;
 
          		// Optional group
          		// inGroups (meshedWalls);
        	}
		}
          
		net1
    	{
			level (4 4);
			cellZone net1Zone;
/*			cellZoneInside inside;*/
			faceZones net1fZone;
    	}
    	net2
    	{
        	level (4 4);
			cellZone net2Zone;
/*			cellZoneInside inside;*/
			faceZones net2fZone;
    	}
    	net3
    	{
        	level (4 4);
			cellZone net3Zone;
/*			cellZoneInside inside;*/
			faceZones net3fZone;
		}
    	net4
    	{
        	level (4 4);
			cellZone net4Zone;
/*			cellZoneInside inside;*/
			faceZones net4fZone;
    	}
    	
    	net0
    	{
        	level (4 4);
			cellZone net0Zone;
/*			cellZoneInside inside;*/
			faceZones net0fZone;
    	}
    }
    resolveFeatureAngle 45;//30;
    planarAngle 30;
    refinementRegions
    {
        refineFreeSurface_L1
        {
            mode inside;
            levels ((1e-5 1)); //First entry here is distance, which will be ignored unless mode = distance
        }
        
        refineFreeSurface_L2
        {
            mode inside;
            levels ((1e-5 2)); //First entry here is distance, which will be ignored unless mode = distance
        }
        
        refineFloater_L1
        {
            mode inside;
            levels ((1e-5 2)); //First entry here is distance, which will be ignored unless mode = distance
        }
        
/*        net1*/
/*        {*/
/*            mode inside;*/
/*            levels ((1e-5 3));  // STL 内部 refinement 到 level 4*/
/*        }*/
/*        net2*/
/*        {*/
/*            mode inside;*/
/*            levels ((1e-5 3));*/
/*        }*/
/*        net3*/
/*        {*/
/*            mode inside;*/
/*            levels ((1e-5 3));*/
/*        }*/
/*        net4*/
/*        {*/
/*            mode inside;*/
/*            levels ((1e-5 3));*/
/*        }*/
    }
    limitRegions
    {
    }
        locationInMesh (-150 0 0);
        allowFreeStandingZoneFaces true;//false;
        faceZoneControls
        {
        }
    locationsOutsideMesh ((100 100 100));
}
snapControls
{
    nSmoothPatch 5;//3
    tolerance 2.0;
    nSolveIter 50;//30;
    nRelaxIter 10;//5;
        nFeatureSnapIter 20;//10;
        implicitFeatureSnap false;//true;
        explicitFeatureSnap true;//false;
        multiRegionFeatureSnap false;
}
addLayersControls
{
    relativeSizes true;
        expansionRatio 1.25;
        finalLayerThickness 0.64;
    	minThickness 0.1;
    layers
    {
		floater
		{
			nSurfaceLayers    2;//3;
		}
    }
    nGrow 0;
        featureAngle 180;//180;
        maxFaceThicknessRatio 0.5;//0.5;
        nSmoothSurfaceNormals 1;
        nSmoothThickness 10;
        minMedialAxisAngle 90;
        maxThicknessToMedialRatio 0.3;
        nSmoothNormals 3;
        slipFeatureAngle 30;
        nRelaxIter 5;
        nBufferCellsNoExtrude 0;
        nLayerIter 50;
        nRelaxedIter 20;
}
meshQualityControls
{
    #include "meshQualityDict"
    relaxed
    {
        maxNonOrtho 75;
    }
        nSmoothScale 4;
        errorReduction 0.75;
}
mergeTolerance 1e-6;
// ************************************************************************* //

Thank you very much in advance!

Best regards,

Liu Jiaming

The instructions are as follows:

To create two meridional slices, one for each fan in TCAE, follow these step-by-step instructions:

1. Load the Simulation Results: Open your TCAE project and load the results of your simulation for the fans.

2. Select the First Fan:

   • In the Pipeline Browser, select the mesh or results corresponding to the first fan.

3. Add a Meridional Average Filter:

   • Go to the Filters menu.
   • Select "Turbomachinery" and then choose "Meridional Average."

I am not presently seeing any geometry in the Pipeline Browser, so, I tried double-clicking the individual STLs making up the component in CFD Mesh, (later realizing selecting multiple elements is not supported) then going to the Filters menu, selecting Turbomachinery,  the option "Meridional Average."  is greyed out, please see image.

I'm still at the beginning setup phase; loading geometry and tweaking types and interfaces. I have no results of any simulation to load. I have not tried to create a mesh yet. Is that (or something else) I need to do first besides read the manual?

Thanks in advance

David

After reimporting surface geometry, I see that the legend for the color bar in Reader View has multiple color bar legends superimposed. Can I, should I hide one or the other?

If so, how?

Thanks,

David

I'm not wanting to re-import geometry, although that's most certainly what I'm preparing for at the moment. All the STLs were created using the _ExportWithOrigin command with World origin as basepoint in Rhino, so I'm scratching my head, as usual. Importing the offending object back into a "top" drawing, an assembly-of-all-parts model, and aligning it to it's exact position, then re-exporting it and replace the file having the wrong origin is what I'll try next.

Is there any easy way to manipulate objects once they've been imported?

Please see my comment, button at bottom right. Thanks.