Reverse Engineering Techniques

I’ve been reminded recently about some of the advances in cheap or free 3D scan methods. The upshot was that using a cell phone camera and a list of free software packages, you can get a free scan to within 0.2 mm of the result yielded by using professional tools. While the value of the $X0,000 that you will spend to close that 0.2 mm gap is substantial, the result, even the improved professional result is not as good as it needs to be to move the product development process forward into drawing production and manufacturing.

I’ve done a few real projects working with scanned data. Most of the time, the customer wants an editable model at the end of the day. They don’t want an STL file, they don’t want rough-faced tessellated data, they want something that can be edited, and something that can be machined cleanly, or something you can make a mold from. STL data doesn’t give you CAD-perfect faces, it gives a lot of small triangulated facets. It’s great for 3D print, but if you need to add draft, it’s a bit of a sloppy mess.

For all the talk of precision from high-end scanners and software, the STL data isn’t usually CNC ready data. Depending on what you need the models for, you may be left with few real choices. If you want a model to use as a starting point for further product development, or you want to make drawings or changes to an old product you need to update, you really need BREP data with real planes, cylinders, cones, spheres and toroids. Something you can use either Synchronous of history-based modeling to edit. NX can use X Form or I Form to directly edit shaped models. Any data you get from a 3D scan requires cleanup to remove junk that doesn’t belong there, and then patching, to replace stuff that should be there. You also have to do smoothing to average out scan error. All in all, it takes a lot of time to get results that aren’t really usable, and are probably not what the customer wants anyway.

There are two ways that I’m familiar with to get machinable data. One of them is to use Geometric Raindrop or something in that league. That’s a lot of money, and to make it pay, you have to be doing that kind of work frequently, or have a big operation to absorb the cost. Even this doesn’t give you an editable model. You will probably get a good quality surface, but not something you can change. And at that point you’ve spent a lot of money on hardware, software and someone to use it expertly.

Subd methods would be great, but there is no way that I’m aware of to interchange between scan point data and subd. All of the model images shown in this blog post pre-date subd in Solid Edge.

The other method is to model using traditional CAD methods usually in combination with some of the other data types or methods. You may be able to model over a 3D scan to use as reference. You may be able to use photographs to get outlines. I have in the past put parts on paper copiers to get a shape. You will definitely have to do a lot of manual reference dimensions with calipers or other tools. The best results will come from using a combination of sources, but the most useful data in the end comes from a real CAD model, not just a straight scan.

Working from scan data in CAD can be a lot of hair-pulling work, and the progress on the software side is slow.

Sometimes I’ve been asked to do reverse engineering from old 2D drawings. Sometimes the product is a curvy model, and the 2D drawings really were never adequate to begin with. In cases like this, I generally work from a combination of the drawing and any existing parts to get a sense for what the real parts actually looked like.

All of the models shown on this page were created using some combination of processes, with the end result being a live CAD model that could be edited and changed in a way that straight scan data cannot.

Scan data itself is tough to edit because of the sheer number of points involved. There is often a lot of extraneous or erroneous data as well. And it’s not perfectly symmetrical the way a real CAD model would be.

If you are a service provider who frequently does this kind of work, then you may already have the tools and skills in place to achieve good models. But if you are a manufacturer who has a need for this kind of work infrequently, then it is probably most efficient for you to hire the work out. Just make sure that whoever you hire understands what you want to be able to do with the data they deliver. They may just do the part of it that they are familiar with, and let you figure out the rest. I’ve been hired often to bridge the gap when this kind of misunderstanding has happened.

The point I’m trying to make here is to make sure that you understand that good or even miraculous 3D scanning is not the end of the road for reverse engineering. Whether you want to replicate the physical properties of an existing product, need to recreate a damaged mold for which you have no CAD or CNC data, or you just want to use an existing product as a starting point for further development, a live CAD model will be what gets you there. Scan data can help you get to a live CAD model, but there is no magic button that makes that leap for you.