December 2002
In the Summer we had the opportunity to meet with folks from EDS PLM Solutions to discuss and demonstrate their knowledge based technology incorporated within the latest versions of Unigraphics NX, called Knowledge Fusion (KF). An EDS PLM team came to our offices to display their latest technology, more tightly integrated within Unigraphics NX. We met with the product manager, and an application engineer, accompanied by the ever present head of UG marketing, there to extol the beauties of the software.
In 2001 we wrote a series of articles, published in this newsletter, on these rapidly emerging technologies. We reviewed Catia's Knowledgeware, Pro/ENGINEER's behavioral modeling, and ICAD in a series of articles. Since then, this technology has continued to expand, even becoming a key differentiator between large scale integrated systems like Unigraphics, Catia, and Pro/E as compared to mid range systems such as SolidWorks, Solid Edge, think3, and Autodesk Inventor.
We applaud this growth and look forward to a significant expansion of knowledge based engineering. Why is this the case? Because capturing such knowledge AND having the ability to easily reuse it, should result in better and faster designs. The more repetitive such knowledge is (the more often certain actions are repeated), the better chance for savings and reduced errors.
To understand how it all works, readers must realize that there are several aspects of knowledge based engineering.
First, the system must have the capability to capture relatively low level engineering processes by non programmatic means, in a relatively short time. Second, it must have the capability to remove simplistic redesign drudgery by easily executing these captured engineering processes. Third, users must be able to easily search fro and find these captured processes and be able to apply them to their designs. Fourth, these processes must be associative, recalculating to the changing underlying geometric elements.
Today's state of the art in knowledge based engineering (KBE) resides within two CAD programs -- CATIA and Unigraphics NX. These systems tightly integrate capturing the knowledge as well as deploying the ability to use this captured knowledge to normal users.
Most of what we saw during this visit by EDS involved demonstrating how such knowledge capture is deployed to users, rather than how new rules are developed.
We have more detailed images and explanations available on our CAD-portal.com web site.
Please review beam1.jpg through beam3.jpg here. We saw a quick demo concentrating on an application of aircraft tubes routed through beams. The idea here was to use an already developed knowledge fusion feature. This predefined knowledge feature enabled us to generate a beam at various locations in relation to the pipes in an aircraft design. This is called the "fuselage beam feature."
The capability of the knowledge-based integration within Unigraphics allowed the system to perform an operation that not only generated the beam, but also allowed the application of additional features, which would survive in the recalculation of the knowledge application. This is important because treating knowledge as features allows such features to be recorded as part of the modeling history, and regenerated, as necessary.
Quick/Check, part of KF, allows the entry of expressions to determine whether defined measurements are violated. In this example rules were built in using an expression and a distance location. Entering a new beam position resulted in a message telling us we had a problem since a predefined distance check was violated. Very interesting was the fact that this distance check and the expression check, which we created later, became part of the feature tree.
Please review UDF_crankshaft.jpg through UDF_crankshaft_3.jpg here. In this example, we had a crankshaft with an involute spline about to be placed so it can mate with the transmission. Included with the involute spline was also a set of knowledge rules. When we applied the feature to the crankshaft model it executed not only the definition of the feature, but also executed it's embedded knowledge rules. Engineering rules typically consist of pre-established formulae and properties accessible in corporate databases. In our case, we violated one of the embedded rules during this application and the rules suggested a way to resolve the violation.
On the original crankshaft, before we placed the user-defined part, we performed a direct face modification, which included a parameter as to the location of the shaft face and another face on the part. When we executed that later, it automatically changed the UDF calculation of the involute spline to increase its length.
Please review airfoil_1.jpg through airfoil_3.jpg here. We looked at an airflow-banding example which lent itself perfectly to KF because of the difficulty in performing a complex engineering process. By developing KF rules, the process became easily repeatable. The complete set of engineering rules required to generate the banding envelope is encapsulated within the Airfoil Banding Feature.
When applied, the Airfoil Banding Feature presents a simple user interface that guides the designer through selecting the required geometry and entering the numeric parameters for design of the banding envelope.
Initially, we had defined an airfoil generated by lofting a solid through a number of cross sectional curves. We wanted to develop offset surfaces from the airfoil, or to wrap a ÒbandÓ around the airfoil solid.
We used a feature called adopt geometry. The system adopted the geometry of the airfoil. The next step was to add a child rule. The child rule we chose was called airfoil banding. Adopting this child rule into the part, automatically built a dialogue box shown on airfoil_3.jpg, wherein we could alter the parameters of the resulting banding. Click on the thumbnail image for a better view of the allowable parameters.
We reviewed some of the parameters involved in the airfoil
band. We made a number of changes to the underlying solid body; when the solid
was recomputed, it automatically recomputed the proper banding. The result of
the banding is more than just offset surfaces -- in different portions of the
underlying solid the banding has a different offset from the initial solid. We
have a number of images of the calculation of the airfoil band. We were
interested in the steps needed to create the airfoil banding child rule. The
EDS folks claimed that it is done without resorting to programming, but did not
demonstrate this.
The advantage of this airfoil banding demo using a KF application is that it operates on whatever shape is input. The KF application generates the shape of the geometry from the airfoil which was the input. Because it is a feature, it is associative to the solid and any tool paths or FEA mesh associated to the airfoil band geometry. When the banding shape changes, they all update. In this example, when the basic curve or applied scale of the solid changed, the airfoil band updated and the tool paths updated.
See Catia_interop1.jpg through Catia_interop6.jpg here. We next saw a demonstration of the capability for knowledge fusion to bring in a CATIA part and an associative way to link it into Unigraphics. This particular application is not available to users, but is a demonstration of some advanced capabilities.
The idea here was to import a Catia developed bumper-mounting bracket to connect the body-in-white structure of a rear cargo door panel and a bumper, both of which had already been designed in Unigraphics. UG opened a CATIA part using an associative parent link. An add child rule asked us to specify this Catia part, which was brought into the Unigraphics model after automatically being converted to Parasolid geometry using a UG translator. We then built additional geometry on the CATIA part -- drilling a hole in one of the faces and adding fillets to some of the supporting ribs.
Next, this was placed into an assembly (see Catia_interop3.jpg), creating several mate constraints to bring the bracket between the support structure and the bumper. Assuming the Catia designed part changed we re-executed the import KF application to get the latest version of the imported part. Even though the part had changed, all of the mating, filleting and hole creation were successfully executed. The associativity works by placing succeeding features onto the same face and element ID's used previously.
This works currently only for CATIA 4 and all Pro/E parts. This is another example of a potential use for a knowledge fusion feature.
We had a tele-conversation with one of the Unigraphics AE's to try to understand what was involved in adding a knowledge fusion rule. After several teleconferences we concluded that developing these rules involves a high degree of training. Skill is involved in UG geometry internals as well as an understanding of the object oriented structure of KF techniques. The construction of these rules captures parameters from UG, but requires a strict regimen to develop the rule. Although specialized skills are needed to develop the rules, once the rules are built they are relatively easy to apply to geometry during a typical design.