Measure flatness - what's best?

[maxmarautodesk]

Dear community,

I am starting this thread, because I am facing the problem that I have to measure flatness on warped parts quite often. Therefore, the question came to my mind how to evaluate the flatness of a part in the best way.

Do you have any recommendations and what are the pros and cons of your methods?

So far, I can think of several methods which I would like to present and discuss here. Feedback on my methods is very much appreciated. You are welcome to add more methods and describe them in detail. I hope that we can pull together a comprehensive summary of how to measure flatness of simulated parts.

1. Use the script flatness.vbs in Moldflow

Detailed information can be found here:
http://autodesk.typepad.com/beyondmoldflowinsight/2012/11/api-examining-warpage-deflection-for-flatn...
In the new versions of Moldflow (I think 2014 and 2015), the script already comes automatically with the Moldflow installation.

Pros: It is very easy to use and no additional tools are necessary.

Cons: I have found that I get different values for the flatness when I do the evaluation manually. Moreover, I am not sure what is going on in the background of the script and how the flatness is calculated. In addition, it can be tricky or very time-consuming to select the right nodes for the flatness evaluation. I can easily happen that one selects the wrong nodes and therefore does not get the exact value of the flatness.

2. Manual evaluation of deflection
The manual evaluation is based on the deflection of the nodes. Based on the deflection values a new reference plane and the flatness can be calculated.

Pros: You know what you do and how the flatness is calculated. You can apply the method to any geometry.

Cons: It is very time-consuming and labor-intensive. It can be tricky to select the right nodes for the reference plane and flatness evaluation.

3. Use GOM Inspect for comparison of original and warped part

The original CAD part and the warped part (exported from Moldflow as STL file) can be compared using GOM Inspect. The software provides the possibility of an overlay of both with an evaluation of the deflection (in my opinion more convenient than in Moldflow). Reference planes for flatness evaluation can be easily created by best fit functions of the respective surfaces of interest (information: STL is a point cloud). I am able to do a point by point evaluation of the deflection. But I believe there is also the possibility to get the flatness directly from the fitting/point cloud data. Can somebody help here?
Pros: Very easy to handle, good visualization.

Cons: You need to learn how to use GOM Inspect and many of its functions.

I am looking forward to your contributions and answers! Thanks!

Best regards,
Max

[mtessier]

Hi Max,

I don't have any experience with methods 2 or 3.

In the past, I used anchor planes, but I found the result to be somewhat dependent on the anchor plane definition.

I have had really good success with the flatness.vbs macro.  The following is happening in the background:

1. Take all undeflected node positions, and use a best fit function (orthogonal regression) to define the undeflected plane.
2. Take all deflected node positions, and use a best fit function (orthogonal regression) to define the average deflected plane.
3. Determine maximum positive and negative distances from planes.
4. Report flatness as sum of maximum positive distance from plane and absolute value of maximum negative distance from plane.

A good check of whether you have selected nodes in error is to ensure that flatness before deflection is zero for planar surfaces.  If you have selected an interior node, this value will be non-zero.

I've found cut planes to be useful for selecting the right nodes, and it may be best to select the planar nodes after surface meshing, but before 3d meshing to avoid selecting interior nodes.

Hope this helps.

Matt

[harald_goetz]

Are you just eager to give the flatness a value, or do you want to optimise the flatness?

In the 2nd case I start to love the path plot. It also helps evaluating different approaches (with the same mesh) really fast.

Else the flatness script works fine.

Of course GOM does a good job, too.

And really - the measured difference should be the same in Moldflow and in GOM.

Harald

[maxmarautodesk]

Dear Matt, dear Harald,

thanks for your answers and your active contribution in the forum.

The comment of Matt about the anchor plane basically applies also to the method of manually calculating the flatness. It depends very much on how many nodes you take into consideration and which reference plane (= manual anchor plane) you end up with.

Regarding the comment of Harald, I can tell that I am looking for both, the flatness value and its optimization. The absolute value of the flatness is something that I can show to project partners and which is very vivid to them. However, the over-all objective is to optimize the flatness and push it into the accepted tolerance window.

My manual evaluation is very much based on the path plot. It comes in very handy, if you have the same mesh and can use the same nodes again. The same applies for the flatness script.

However, I see this approach also having some major drawbacks. The first option is that you select all nodes of the evaluated surface which is very time-consuming. It is difficult for complex parts to select just the nodes on the surface but all of them with the given tools in Moldflow. Moreover, using all nodes on a surface is hardly practicable, because after a certain number of nodes, Moldflow cuts off the node numbers in the selection box, so that it is not possible e.g. to copy the node list and reuse it in another study. You have to select all nodes again in the other studies...

The second option would be to reduce the number of nodes that you take into consideration for the flatness evaluation. Leaving out some nodes comes with the risk of not picking the right nodes though. This especially applies if your part starts perhaps warping in a different manner. The selection of nodes might be insufficient and not account for the largest deflections of the surface. This means you end up with a flatness value that is too small.

Also thanks for the insight to the flatness script. I think the reminder about zero flatness before deflection as an indicator whether one has chosen wrong nodes is of great value. The script works without problems, but I still get different values compared to GOM:

Plane number: Flatness.vbs // Flatness GOM

Plane 1: 4.2 // 2.9
Plane 2: 6.5 // 5.1
Plane 3: 3.7 // 2.4
Plane 4: 5.2 // 4.0
Plane 5: 3.4 // 2.5
Plane 6: 3.1 // 2.1

All values in µm. The values come from the same part, just different processing scenarios and the aim is to push the flatness below 5 µm. You could argue, the difference in absolute values is not very large, usually only about 1 µm, meaning that the flatness script and GOM give basically the same value. On the other hand, the relative difference is quite large, namely up to about 35 % which I consider being significant. Also, the difference of 1 µm is quite large compared to the acceptable value of 5 µm and this 1 µm difference can lead to the flatness being inside or outside the accepted range.

I am not sure where the difference comes from. Any comments or thoughts on that are very welcome. I can imagine three decisive factors:
1. The calculation of the fitting plane of the warped surface is different. According to Matt, the Moldflow flatness script uses an orthogonal regression. In GOM you can chose between a Gaussian or Chebyshev best fit. However, there is no difference in GOM between both fits, the flatness evaluation yields the same value. Still, the data base for the calculation can be different. In Moldflow, you select all nodes manually. In GOM, the data points are selected automatically by selecting the surface.
2. The export from Moldflow or the import into GOM. Something can happen during export or import. For instance, Moldflow exports the warped part as an STL. I have no information about the mesh of the STL file. Perhaps, the original mesh in Moldflow is not exactly portrayed by the STL mesh.
3. The operator. It can also simply be me that does something wrong in GOM, although I follow the flatness evaluation procedure as it is shown in tutorial video 10 of GOM Inspect V7 SR2. It should be noted that I use GOM Inspect V8 though. There are unfortunately not many tutorials available for GOM. If you have any recommendations, I would appreciate them very much!

Any further comments on my thoughts and the topic are appreciated. I believe there are more people having benefits of this discussion.

Best regards,
Max