subscribe
 

Pattern recognition ensures quality OEM data exchanges

21st February 2013


Pattern matching, intelligent data mining and model validation have collided courses in recent years. The concept of using artificial intelligence technology that uses pattern recognition is not new. For years, it has been used to identify and match fingerprints in the FBI database; it has been used to locate photos on the internet by image recognition engines. By applying similar pattern recognition algorithms to the validation of CAD geometry, the technology proves to be a fast and robust method for insuring the quality of OEM master file data exchanges.

Geometry pattern matching principles can be used for comparison applications where one version of a part or assembly can be compared to another version. By using real-time feature recognition technology, users can see if two models are the same without the need for history or parametric constraints. Every major or subtle change in the geometry can be quickly identified using this technique.

For the past three years, Kubotek has been using 3D pattern recognition technology to validate 3D CAD models and assemblies for translation and revision changes. The majority of Kubotek users are not the source of the CAD data, they are not the designers. Instead, they are engineers required to work with 3D data as it moves through the supply chain. Therefore, it is important that they receive and send their CAD data in any format, and do so without propagated errors through the supply chain. Kubotek realises this by using native CAD files as well as industry standard formats like STEP and IGES

The foundation of Kubotek's technology is the direct-modeling approach, and uses open 3D environment and technology called Face Logic. Face Logic technology can use native CAD files as well as industry standard formats like STEP and IGES. Simply moving the cursor over the top of the master or copy model identifies features in real-time. Once changes are identified, the cursor can hover over the changed area, and the technology will allow portions of the model to be removed and edited directly.

By using pattern recognition, two models can be compared in any alignment, as long as their geometries match up. A model can be in any orientation, twisted and tilted, and the technology can still do comparisons. It also does not matter how many parts or assemblies are used to create the model, pattern recognition can still do comparisons.

Pattern recognition technology easily answers increasingly common, yet critical questions: how good is the file data, do differences exist, and where are those differences? One technique for comparing files is the more traditional 'point cloud' technique. This methodology places inspection points all over the models, but it is significantly slower and less accurate. It also relies on the proper amount and spread of inspection points within the model to result in accurate geometry comparison.

Comparing two models and their face points is more robust for validation. This comparison can be used in two different ways. First, when files are translated, it ensures data stays the same during the manufacturing of parts. For instance, validating using pattern recognition allows users to take the data that comes out of MasterCam and compares it to the original CATIA model. Secondly, comparing master and copy files ensures change orders do actually reflect the changes that have been requested. By validating the two models, differences between two revisions can be easily found.

Diverse audiences are using CAD tools differently within their respective departments. Experts CAD users tend to understand one or more CAD system and can navigate within those systems without problem. They can interrogate and verify parts at greater ease than people who do not know parts. People in the non-expert audience tend to be more involved in documentation or quality. They do not need to know how the model is designed, but they need to know if it is the same as the master copy. A simplistic interface to the pattern recognition technology proves to be the best approach.

Pattern recognition validation has been proven accurate by additional validation work. A digital gauge block was created. All the different surface types were mathematically calculated by hand. Points were added to all the mathematical surfaces, and were then looped through different CAD systems such as Pro-E, CATIA and UG. They were then re-validated to make sure all data remained intact. The generated digital gauge block and its mathematically defined parts for each system were then transferred to STEP and IGIS to do comparisons. After all of this, validation comparisons were repeated. The results were all green lights, proving that all translatrs and mechanics worked perfectly.

Additionally, a blind field test was conducted. A manufacturer received drawings in one file type from the OEM. They rebuilt the part from scratch so they could manufacture it. At the time, three known changes existed in the CATIA model versus the original master model. Kubotek was able to detect a total of seven changes, four of which were not documented on the change order. Further investigation determined that the undocumented changes were made intentionally by a manufacturing engineer, but they were not documented properly.

This technology has the potential of providing a solid foundation for part search and change control applications for PDM. This could enable the development of advanced PDM tools built not just on searching databases for text attributes, but on discovering the changes made to the CAD model geometry itself. As design data moves from the OEM through the supply chain, into and out of disparate supplier CAD/CAM/CAE software, subtle and sometimes undetected changes can occur altering the design geometry. These changes can be undesirable, affecting quality of the manufactured product.

In conclusion, Pattern Recognition is a viable method for comparing CAD data sets. It provides translation reporting capabilities for quality assurance documentation. The technology also provides revision reporting capabilities for engineering and manufacturing groups to manage change control. Validation tools, particularly ones that use pattern recognition and direct model editing, can be a critical component of successful OEM master data exchange with supply chain partners.

Enter √ at www.engineerlive.com/ede

Luca Cariglia is CEO, Kubotek Europe srl, Costabissara, VI, Italy. www.KubotekEurope.com







Subscribe

Previous Issues

 

 

 

 

 

 

 

Subscribe



Newsbrief

twitter facebook linkedin © Setform Limited