Sunday, August 22, 2010

3D-Measurement systems can be so cool

As a lover of modern technologies and an engineer myself, I always find new technologies (in all kind of places) to be interesting and fun to understand. Of course, I belong to a small subset of humanity that thinks those kind of things are cool (probably the majority is male).

So, if you are not into that kind of stuff, either bear with me or skip the post.

Modern measurement technology is getting better and better. When I look back only 15 years, then I don’t see any 3D-measurement stations that could virtually built a model of the work piece you sit in front of them. Then computer aided manufacturing introduced numerical control based computers to the production line and at the same time 3D-sensor units were introduced to control the end product (especially when it has a complex geometry).

As anyone knows who has worked in a workshop only once in his life (or played with wood cutting), you will never reach the perfect dimension you wrote down in your sketch. You will always miss it by several hundreds of Millimeter (or more). These imperfection of the real world production are included in the conception of the work piece by allowing for certain tolerances. However, you have to check your finished product to see whether it complies with those tolerances or not. To do this, we use different measurement tools. We can use rough measurment tools like rulers or gauges, which measure between 0.05 – 0.2 mm differences adequately.

But the more complex the piece gets the harder it is to manually measure those tolerances. So, we started using computers and fine metal pressure sensors to aid in the verification of tolerances.


As you can see this is already an improvement over manual measurement or even old-days two point measurement stations.

But while these methods work fine with metal, which has high stiffness and is harder compared to other materials. We often get the problem that plastics are not that stiff and thus bend when measured with a contact sensor tool.

The solution is optical measurement and this is really cool. You place your work piece on a well-lit white background and you have two cameras in a fixed angle to each other. You will then start to project a grid on the work piece to allow the cameras to set their coordinate system on the work piece.

Afterwards the cameras start to take pictures and because the computer knows the differential angle between the cameras, he can calculate a 3-D model of the work piece and compare it to a 3D-model of CAD-drawing.

The result is something like this:


The differences between theorectical and actual dimensions and tolerances of the car are seen in this picture. They are color-coded between blue (small differences) and red (high differences).

Yes, there are many problems with this kind of measurement (as with all measurements). The maximum measurable difference highl depends on the material and the exact set-up of the cameras and the work piece. For example, several metal pieces are a problem for optical measurement, because their reflectivity blinds the cameras thus they have to painted white and less shiney. However, this will add a level of paint to the material dimensions which can vary due to complex angles in the work piece (corners will have more paint in them that level surfaces f.e.).

But especially for plastics this is just a great way to measure them and will make verification of batches much easier and more precise. And what do those machiens cost? Well, a huge amount of money (around 500.000 €).

No comments: