74 faces is much better than 147 
Don’t see watertight message in the geometry event log tho…
Actually, I booleaned together, very easily, a lot of your model (after exploding them somewhat) in Rhino and that is how I found that nose cowl problem. The nose cowl did not boolean with aft fuselage.
I really like Rhino for things like this and I always import Rhino .3dm files to SimScale. I think it is very similar format to STL (maybe an extension of STL) but that is only my feeling without any way to confirm that.
Unless I have a specific reason not to, I import my aircraft as 1 watertight solid…
Better to find the problems in CAD than bring it into SimScale and hope SimScale can fix less than perfect CAD geometries.
1 Like
Vexing. The Solidworks model is clean and beautiful. Coincident surfaces wherever there should be. I tried an STL export and it’s apparently not watertight either.
Aw, what the heck, I will probably never build this anyway, so I just turned this into a Public Unfinished project…
I am pretty proud of my geometry file (37 watertight faces 
):
1 Like
Hi Both,
So is there a link to the adequately “fixed” geometry I can attempt the cfMesh run on?
Cheers.
Regards,
Barry
Well currently not one is watertight and it seems to be due to the fact that Solidworks sucks at complex surfaces… took it into Rhino and found multiple meshing errors. But this one is as close as it gets.
I played few minutes with free Autodesk MeshMixer (only STL) and it has set of optons for repairing meshes. Target user are 3D printing guys, trying to make sens of their 3D scans.
Cheers,
Restam
Looks like you will have to follow/help on a new topic I created about watertight problem… sorry.
EDIT: I guess you saw that answer this morning (what a great system we have here and comforting to know that real masters are lurking when really needed), all solids are considered watertight and the need for that message apparently disappeared, I still would like to see it there…
Before I discovered that issue, I started playing with your CAD files, I found a couple naked edges in Rhino, even though the spinner Booleaned to the ‘cowl’ I saw that somehow the rear spinner plate was still in the solid. Then I saw that the tip of the nodes in the spinner’s curve of revolution caused the spinner to have a concave nose (likely does not matter) and that the spinner centerline was not on the x-axis by ~.8" so model could not left/right symmetrical. Fixing all that did not help but I did anyway. Never ended up getting cowl to Boolean with rear fuselage, not seem to be lofted from same curve I guess.
I guess none of that matters to watertight issue, just thought I would pass along those minor points…
NOW, to move on, I guess someone should make those EVR plots on the successful 25.6M simulation and then make further decisions on mesh refinements for MK4 project. 
ing for a volunteer who knows Paraview to rescue me …
Thank you so much for that other topic… I was pulling my hair out trying to figure out how to get it to be watertight. There really aren’t many knobs in Solidworks to tweak if the geometry in the model is well defined but there are still errors on export.
You’re seeing the thrust vector offset of the engine modeled both unnecessarily and poorly in this case. Knowing what I know now I would omit this geometric detail and save some pain.
I’m wondering: other than your cute little table for layer %, do you ever use the added cells percentage as a measure of quality? For instance, here’s one that layered up pretty well (Y+ about 125, so using wall functions… much more consistent than full res BLs)
Detected 0 illegal faces (concave, zero area or negative cell pyramid volume)
Extruding 473080 out of 475489 faces (99.493363674%). Removed extrusion at 0 faces.
Added 3631913 out of 3803912 cells (95.4783654301%).
That 95% added seemed to be a fair measure of retained BL cells, so I was using it. By comparison, my best full res layering attempts (using the modified Kramer method with absolute BL definition and variable surface refinement) yielded around 72%.
Aha, you may have found place that better summarizes their layering table and save having to pull and analyze that chart…
Lets see if we can verify that better with more comparisons… and maybe find an actual confirmation of what the 95.478365…% is…
Extruding 1880522 out of 1902176 faces (98.8616195347%). Removed extrusion at 0 faces.
Added 16503166 out of 19021760 cells (86.7594060697%).
Variable surface refinement and 0.0005 face twist:
Extruding 852103 out of 891278 faces (95.604626166%). Removed extrusion at 3 faces.
Added 6988916 out of 10695336 cells (65.3454552526%).
Never could find one that was as low as 47%
I’m still not convinced it is better metric than mine, need to really find out what that embedded metric is that you found…
Are those ‘out of’ numbers actually the number of prism cells in a perfect layering???
EDIT: By that I mean, for 6988916 out of 10695336, are there 10695336 prism cells in the perfect layer mesh 
I believe that is the % of cells out of the original “perfect” BL extrusion that it is going to retain after doing displacement and applying the quality metrics. You’ll see that number go down with each iteration.
But I think that is just an assumption right now.
I think my calculation (sum of (faces*layers/face))/(total faces) is the correct weighted average of layers/face for all chart data.
EDIT: More research tells my gut that your new discovery is not what you think it is, more to do with individual iterations of layering process, not end result (but I could be wrong) 
No doubt yours is a better metric, but this one is free and fast. 
Can’t resist this one… maybe you get what you pay for here 
1 Like
@jhartung , wow this is now the 5th longest thread ever in the SimScale environment with respect to # of posts (I think) …
You deserve some sort of reward for hanging in there
I have learned a lot and I hope you do summarize what you have learned at the end here, so that those who follow may be pointed to the end post because I don’t think that a lot of others will get through the whole thing (but I know I would if I happened upon it )
If you do summarize here, please be sure to edit your 1st post with BOLD text that links to the end summary post 
But, be ready for some continuation after that summary
1 Like
Good idea Dale!
I think Josh wanted to write a summary anyway once he finished his studies 
Best,
Jousef
Hey all, sorry for the radio silence. Having gotten to some results I mostly like, I had to do some work on the project actually using them. I’ll be back to transform this into a great project and post, fear not!
1 Like
@Get_Barried AHA, I finally found Dylans explanation that I think sortof relates to the Absolute tolerance 1e-15 and for sure, the larger relative tolerances that I use, basically for quicker convergence in steady state analysis’s :
’
’
’
’
EDIT on 11-2-2023: The project where this WHOLE topic is mostly summarized is located here: 'Crop duster lateral stability' simulation project by jhartung
’
2 Likes