OpenFOAM

tahanson43206 · 2025-05-03 15:12:03

Update after a long day chasing bugs ...

Our attempt to add 25 new cells to a structure with thousands of existing ones has revealed dozens of ways to break the existing structure.

The process of resolving the error messages is challenging, because (it appears) a single point in the wrong place can cause a cascade of errors that adds up to many thousands.

My guess is that ChatGPT4o and I are close to finding that last pesky point, but OpenFOAM isn't much help. It tells us we have a point out of place, but it doesn't say where. Diagnosis is (apparently) 90% guess work and 10% ferreting hints out of the available data.

I sure am glad I didn't try to do anything more ambitious than adding a single ring of 25 cells to the end of the intake.

Update later:

In reviewing our steps, I went back to the beginning, and discovered that in the very first step, we somehow lost track of the count of points to be added. We added 51 instead of 52, so of course everything after that was off by that one point.
I've recreated the points and stored them in the points file.

(th)

tahanson43206 · 2025-05-04 12:44:59

If there is someone in the readership who enjoys working a numerical puzzle, here is a data file that ought to be a challenge:

https://www.dropbox.com/scl/fi/lgu7thih … 9potv&dl=0

This is the output of a Python program written by ChatGPT4o to find faces in a file for the Merlin Engine model.

The faces of interest are those associated with the points that comprise the leftmost set of points in the model.

The purpose of the exercise is to replicate the 25 cells that comprise the left edge of the structure (left = -X direction).

ChatGPT4o found 177 faces with at least one of the 52 points connected.

Of those a number are duplicates because more than one point will be associated with many of the faces.

I'm looking for a number like 101 faces that are unique, and are not part of the prior set of cells.

Each cell has six sides (top, bottom, East, West, North and South).

Update: Overnight, I figured out that the faces definitions must include both inside and outside.

25 cells have a total of 150 faces if they are standing alone. In combination, some faces are shared.

In addition, if the faces are defined as both "inside" and "outside" then we get counts that look closer to the 177 we found:

top: 25 cells so 50 faces
running total: 50 of 177
+X side: 25 cells so 50 faces
running total: 100 of 177
-X side: 25 cells so 50 faces
running total: 150 of 177
+ and - Y (with shared walls) 26 (no mirror because shared?)
running total: 176 of 177

By this accounting, I am short one face.

Perhaps one side facing out in Y is mirrored? But if one, why not both?

At any rate, the mystery of the 177 count seems almost resolved, if we assume OpenFOAM needs both inside and outside of each face.

Also, in thinking about this puzzle, I realized that the faces we found must have either four points (the top layer) or two points (all others) in common with the outside points. The list we have can be analyzed for the number of points shared with the top layer.

(th)

tahanson43206 · 2025-05-05 10:11:33

Work on the attempt to create a new layer of cells to be prepended to the existing Merlin Engine model is continuing.

Wording revised 2025/05/05:
In early study today, it has been confirmed that most of the 177 faces identified as connected to the leftmost points are mirror sets. This is consistent with the OpenFOAM structure of "inside" faces and "outside" ones. It appears that faces that do not have mirrors are faces that are shared between two cells. Study so far is still delivering mysteries. For example, a study to find mirrors in the 177 faces found earlier resulted in identification of six faces that do not meet the minimal criteria of a mirror. There may be another category of faces that do not have mirrors.

Investigation continues. I cracked open the 2011 documentation I have for about an hour today. I saw a lot that is familiar do to working on this project in 2025. The core features (the architecture) of OpenFOAM appears to have changed very little, but the software that performs the work sure has.

One thing I was surprised to find is that an "stl" file (such as created by Blender) can be fed into the OpenFOAM mesh create utility. And that was back in 2011, so I expect that feature is still available. That's an option, if the current effort to extrude the existing intake proves too difficult.

(th)

tahanson43206 · 2025-05-10 12:17:39

This post is to report a mini-milestone.

Setup:

The goal of the Optical Plane OpenFOAM project is to simulate the new engine gas heating system.

We are using an inherited Merlin Engine model, which works with modern OpenFOAM software.

The immediate challenge is to learn how to grow or extrude the left facing edge of the intake to create a pipe for heating.

A first attempt came close but failed, because neither ChatGPT4o or I understand what we are doing.

checkMesh is an OpenFOAM utility that looks at mesh created by automated systems (such as the Merlin engine) and human ones.

It likes the Merlin engine model.

It went blooey over the human created glob I was trying to stick on the front.

In an effort to try again, I've decided to print the points and faces on business cards.

The first phase of this effort is complete. I now have 52 business cards printed on Avery business card stock.

The next part of this process is to try to print the identical information on the backs of the existing business cards.

ChatGPT4o wrote a little line swapper utility for me, and it swapped the lines in the CSV input file so now the printout should appear on the back of the existing sheets with the columns swapped so the printout matches what is on the front. I'm hoping to complete that process today.

I'm hoping to have the 52 cards ready to show attendees of tomorrow's Google meeting.

Update at 21:20 New Hampshire time:

All 52 cards are printed front and back. The sequence of steps, and the specific steps themselves, are factors that lead to success or failure.

The default font is 12 so if you want a larger font (such as 16) you have to set ** that ** variable on the generated page of business cards ** before ** you start adding fields. There are several ways to add fields but only one seems to work (insert from top menu > fields > more fields)
You have to remember to newline each entry from the database. You have to remember to insert the Next Record after the last field, before you click Synchronize). To make things even more complicated, the default page configuration is for a book with left and right pages, so you have to remember to change that setting to Right Only. Fortunately, if you forget, at Print time, the Preview page will allow you to confirm you are about the print the correct page.

I got six good pages of business cards and only "wasted" two (used for practice). I wasted a ** lot ** of 8.5x11 paper!

The next print phase is 177 face cards, and ** that ** will be more complicated because many of these (if not all) must be printed with different information on front and back. The points cards were easy because they were printed the same on front and back. The only complication there was to swap columns when printing front and back, so the cards match front and back.

ChatGPT4o and I studied the cards I printed, and compared them the intake. it turns out the order of the cards is opposite to what I expected, because I sorted them (inadvertently) in positive Y instead of negative Y. At the time of the sort, my first concern was to insure that we got the + and - Z values adjacent to each other. I don't think the order with which we lay the cards on the table makes any difference to OpenFOAM. What ** is ** important is the ID of each point, because ** that is crucial for creating new faces that will eventually comprise the six sides of 25 new cells.

Here is what we were looking at:

(th)

tahanson43206 · 2025-05-11 10:01:02

This post is to report progress toward the next mini-milestone...

ChatGPT4o was able to create a Python program that collects the 25 faces we see in the end-on view of the Merlin intake.

The collected data is formatted for printing on business cards, so I will attempt to complete that operation today. The data includes both sides of the 25 cell outside faces, so there are a total of 50 lines of data. My intention is to attempt to print front and back.

(th)

tahanson43206 · Yesterday 19:29:59

This post is to report on another mini-milestone...

The printed business cards served their purpose.

The cards helped to create a mental structure that delivered an insight about what data to request.

ChatGPT4o generated a program that delivered all 150 faces of the outermost 25 cells of the existing model.

I'll print the cards tomorrow, and hope/expect to be able to visualize how to use that data to create the data that must be added to the existing mesh to prepend a layer of 25 cells. It is ** still ** going to be tricky to figure out how to create the new data, but this new result is definitely a step in the right direction. I won't print all 150 lines, but here is what a bit of it looks like:

cell_id,face_id,point_ids,face_type,rotation,orientation_hint,position_hint
0,0,"(0, 14117, 28307, 14190)",internal,CCW,card face angled,inside share
1,0,"(0, 14117, 28307, 14190)",internal,CCW,card face angled,inside share
0,1,"(0, 14190, 28379, 14189)",internal,CW,card face -X,inside share
1,2,"(14118, 28308, 14239, 49)",internal,CCW,card face angled,inside share
2,2,"(14118, 28308, 14239, 49)",internal,CCW,card face angled,inside share
1,3,"(0, 49, 14239, 14190)",internal,CW,card face -X,inside share

(th)

New Mars Forums

Announcement

#51 2025-05-03 15:12:03

Re: OpenFOAM

#52 2025-05-04 12:44:59

Re: OpenFOAM

#53 2025-05-05 10:11:33

Re: OpenFOAM

#54 2025-05-10 12:17:39

Re: OpenFOAM

#55 2025-05-11 10:01:02

Re: OpenFOAM

#56 Yesterday 19:29:59

Re: OpenFOAM

Board footer