MSCS Thesis Idea.

[RabiaSalman]

Image processing

 

Artifical Intelligent

[ecoli]

two words: machine learning

[RabiaSalman]

kindly tell me topic... not area.

[Enthalpy]

Mechanical engineering would need a software that converts automatically 2D CAD drawings into 3D models.

 

That is, 10 years ago, many technical drawings were made as 2D, which is essentially the same as paper design aided by a computer. These drawings still exist, are used, but get old-fashioned, with the associated drawbacks.

 

Computer-Aided-Design is presently done in 3D, which means that the software gets from the designer a 3D description of the parts; 2D views and cuts (for paper prints...) are extracted automatically from the 3D model.

 

Engineering companies would need to convert the old 2D files to 3D models but this means a big investment in designer time. An automatic conversion would be welcome. Asking a human to help the software is perfectly acceptable if not too often and in a bearable manner.

 

-----

 

A 2D technical drawing is supposed to contain all the information to represent the parts, but this has first to be understood by a software! Many rules are used to read a drawing. In addition, some information is implicitly re-created by the reader even if he isn't aware of it: one single diameter indication would qualify 10 identical holes without telling which ones, for instance.

 

This projects needs no machine learning, but the amount of expert knowledge needed by the software, plus some necessary interpretation, dictates to program it by artificial intelligence techniques.

 

To be accepted, the conversion software must necessarily use commercial file formats.

 

-----

 

The size, and maybe complexity, of such a software exceeds an MSc thesis to my opinion. But if one plans a following PhD, maybe a limited demonstrator can make an MSc thesis, for instance working only on parts with cylindrical symmetry, starting from sagittal cuts only. You can add features as the project advances, like reading the combination of a sagittal cut and a polar view.

 

I suggested this idea elsewhere several years ago but have heard of no implementation, so it must still be new. If you're interested, best wishes!

Marc Schaefer, aka Enthalpy