Spent at about 1 hour on trying to model the mostly accurate shape of the blade of this gpu fan but can’t get (imagine) what kind of shape it really has because as you see my tries doesn’t match to the reference image at all .
Thanks for the reply. But I’m interesting in a way to make a blade that fits to the reference image’s blade shape. I can model generic form blade doing it by feelings, but as I mentioned in my post I get not the result I want, my blade shape and blade shape in the reference image differs.
I didn’t say that I think like there is something has to be not curved or I even have never seen how does fans look like. I said that I’d like to model this gpu card. But before starting modeling the main parts of the card I decided to understand is there (in reference images) enough information about the actual (mostly accurate) blade profile in order my model will look like original one.
So I created the blade’s profile with the help of “helix along the curve” plugin. Then I have added the faces with “curviloft” plugin. After that I extruded it with “joint push pull” and finally used “subD” to add some organic details.(you can see final results in my 1st message of this topic)
But after that I decided to check my fan (blades) with the reference images and I saw that it doesn’t match at all. So after that I created this topic to see if it’s really possible to make the blade mostly accurate to the reference images because that is all we’ve got, I don’t know actual profile of the blade, I have only work with that image or even other images of this gpu where we will not have such a view where we will exactly see the real profile of the blade.
You might also be asking yourself why am I asking you about things that certainly not visible enough (hidden from our eyes) to recreate their shapes. But I’m not an engineer and do not know if there are some worldwide standards exist when factories create such kind of fans. I thought what if this image enough to determine to what kind of blade it refers like aha, according to the vector of the line of one blade it says that it refers to the blade type called “11111” so the real profile of it can be found by this link “22222”, something like that. hope you understand.
Sorry I had misunderstood you, I had roughed out an example without opening your actual file as I thought at first you were looking for a general technique. When I look at your file i can see it looks good with the use of subd and assumed curviloft. So I wasn’t sure if you had made this or had the model and wanted to know how to replicate. Now I get it.
Coincidentally I have this exact GPU card, so when I power down I’ll take a look at the blades!
thanks. by the way my first mistake was the fact I didn’t count the number of blades correct from 1st sight. It was 12 but it has 13 that’s why my tries to deformate a mesh in other ways didn’t get any success. now it looks fine I think and visually matches the image.
It will take a ton of trial and error to get it as exact as you seem to want without a part print showing the dimensions and geometric development of the shape. In my career I have had to take 2d client part prints where no model was created, for things like fans and fan shrouds, they generally develop the shape with a set of equations and possibly also translating that into a huge point chart.
I worked for a mold maker and I had to use those prints to make the 3d model that was used to produce the injection molds, so it had to be absolutely precise. I was working on site with full blown CAD software worth between $5k and $10k, and so there were lots of tools to automate the process, but still it can’t pull a rabbit out of a hat, your precision is only as good as the precision of the data, and a picture just isn’t precise data.
If the fan already exists, a 3d scan can replicate the point chart fairly accurately, but that is expensive unless you have the capability in house. What do you need this model for? At the end of the day, if you aren’t using this 3d data to actually cut or print a functioning fan, you might just need to accept a bit of error in the visual comparison.