Jump to content

Reporting total deviation between aligned actual and nominal data


---
 Share

Recommended Posts

---

How can i report a single value representing the total deviation between actual and nominal data after alignment?

As shown in the screenshots, i see a deviation value after pre-alignment, is this what i should report?

There is also that table that shows mean distance in the other photo. I am confused which value should be reported.

So if i have multiple comparison pairs in a study group and i want to prepare raw data for statistics  among groups , what value should be used?

Additionally, what is the calculation principle in GOM inspect? Is it Root mean square deviations? 

 

Appreciating your help!

 

deviation.jpg

Link to comment
Share on other sites

---

Please sign in to view this username.

 can you give some information about how you would use a singular value like this ?  It is difficult to describe the out of tolerance situation by these values.

However, if a percentage out of tolerance is what you want you can do this by applying tolerances to the various patches , applying tolerance by cad on this surface comparison, setting th legend to tolerance and viewing the statistics 

Link to comment
Share on other sites

---

Thank you for you reply. For example, i have three groups of data to be compared statistically. Within each group, there are a set of surface comparisons (Actual/nominal). So i want a single value for each surface comparison that can represent the total surface deviation. Based on this, i can prepare the dataset for each group and apply statistics by comparing the means of each group.   So from the screenshots  i shared earlier, i do not know what value should be reported for each surface comparison. Tolerance was standardized for all comparisons and i am interested in the whole surface rather than particular patches.

deviation.jpg

Link to comment
Share on other sites

---

Please sign in to view this username.

  im sorry i still dont quite understand what you are trying to achieve in the end .  I have many years of utilising 3D data and this feel a new one on me ! 3D surface analysis is complex, trying to boil down to a single number may really run the risk of not segregating your data in the way you want.

To be honest im really struggling to think that any value on the regular statistic label would help you.

This sis why i mentioned the percentage inside tolerance .  At least this gives you a descriptive value of some kind that says n percent of my part is beyond x mm which could be used to segregate data

Link to comment
Share on other sites

---

Hi

Please sign in to view this username.

 , i dont work for zeiss , but i used to work for GOM for many years .  I only add what i can here and i cant help via direct contact im afraid.

I know what you are trying to get at but i am still not convinced that a single value is going to help you completely.  The mean might not be a bad place to start but could have some problems .

For example, i can imagine this being a stage project , with your base scan as nominal and all test scans as various stages.

You could have an instance where a certain scanner performs very well in say 80 percent of part but not in the 20 percent.   In this case the mean might be lower but that higher deviation is an area you cannot compromise , say in a fine detail area.

Alongside the mean value id consider surface points in key areas , and perhaps some distances ( to evaluate scale) and , other evaluations you are interested in.  The discrete points and measures would then give values that are easier to assess for bias, where you need to decide an acceptable tolerance .

Hope this helps in some way

James

 

Link to comment
Share on other sites

---

Thanks so much for your kind response. I agree to your suggestion, However, since the model is flat surface without distinct landmarks that i can measure the deviations at, every time precisely, i opt for the mean to avoid bias due to random point selection.  But i totally got your point Thanks so much for co-operation .

Link to comment
Share on other sites

---

No problem

Please sign in to view this username.

.   So your object isnt the ear but a flat surface of some kind?

May i ask what your end application is?

There are options to calculate flatness via GDandT norms in the software and represent this in colours ., Or do surface deviation colour plots  to a plane .

This might be a good test for a prospective scanner .  I.e. Has the flat surface become warped or dished .

 

Link to comment
Share on other sites

---

Please sign in to view this username.

Sorry for the confusion , My object is the ear, I meant by flat that  it is a smooth surface without definite surface features or landmarks that i can always choose during deviation measurement . Therefore , i might not be able to choose the same exact measurement location for each surface comparison which can affect my results. That is why i thought of using the mean deviation over the entire surface but like you explained it might not be the best way to do it.
             
About the end application,  This is part of my PhD research because i am a maxillofacial prosthodontist and we use scanners for a digital workflow of prosthesis fabrication. The ear model is used as an example of digitizing a facial structure and i wanted to evaluate the accuracy of the scanner for such application.
 
Thank you very much for your last suggestion about flat surfaces.


Regards
 

Link to comment
Share on other sites

 Share

×
×
  • Create New...