Any recommendations for reverse-engineering impellers?

Hi everyone,

We are in the process of purchasing a 5-axis machine to manufacture impellers (I have attached a sample photo). We have access to a Faro scanner that generates a point cloud, but we are a bit lost at what to do after that in order to create a simple machinable model. Any advice for getting from point clouds to models quickly?

is that from a scooter?

I would look for the simple underlying shapes where possible - the vanes are complex in shape due to the edges and a scan will show that but there is, I would say, a simple underlying shape for a vane that you can (maybe) model accurately with simple surfaces and then trim.


1 Like

DesignX plus training

Here is a link to a commercial case study on the reverse engineering of an impeller. Laser scanning proved inadequate and a CT scan was needed. If you haven’t already done so, I’d recommend a read.

3d Scanning Of Impeller By Laser And Computed Tomography (


This should be fairly simple to reverse engineer with a faro arm, but the last thing you want to do is create a pointcloud to accomplish this. I have modeled a lot of impellers and from your photo I can tell that the impeller shroud is supposed to be a perfect revolved surface as are the tops of the vanes and OD and ID of the impeller. All these revolved surfaces are supposed to have the same axis. The vanes are supposed to be a polar array where each is the same and all are concentric with the shroud center hub axis. The reason there are 5 holes drilled in the hub is because this part is not as perfectly concentric as it is supposed to be and it needed balancing. There is no point in trying to copy the existing errors - you will only magnify them.

The way I would sample the vanes is shown in the image below. I would markup one vane with a set of lines and then take a sample point at the end of each line (do not include the fillet at the base of the vane).
After you get the sample points into Rhino you can make lines and extend each line by about 1/4" and then create a uniform loft surface. You can use curvature graph to check for smoothness. If you use Rhino’s history to make the loft you can make small edits to the lines to eliminate waviness so that the flow lines of the surface follow smoothly the flow of the shroud.

Its been about 15 years since I used a faro arm but there was a feature called “inspect” which allowed you to see in real time how close any cad surface was to the real surface being inspected. Use that to determine if you need to refine the surface by point editing. I would think you should easily be able to get your surface to within .03" of the existing part. I doubt that the existing vanes are any closer than .03" to being the same.
You can then make a polar array of the surface and inspect the other vanes. The revolved surfaces you create for the shroud and top of the vanes can also be checked for accuracy.

Once you got the basic vane shape and the shape of the three revolved surfaces (four revolved surfaces if you count the center hole) constructing the whole model should be fairly straight forward. The revolved surfaces are marked with green Xs

The fillet between the shroud and vanes can be measured with a radius gauge and the center hole threads can be measured with a thread gauge.

1 Like