Rules of engagement

The objectives of collaborative engineering are often foiled by problems with data sharing. Charles Clarke and Ken Steel offer some timely advice.

The objectives of collaborative engineering are often foiled by problems with data sharing.

Indeed, the only 100 per cent reliable data translation method is when collaborating teams use the same version of the same CAD system running on the same hardware or hardware platform.

For instance, UNIX and Windows models produced by the same software are not necessarily exactly the same, and although interoperability is improving, data translation is still often much more successful in UNIX.

If large databases are to be shared between Windows and UNIX installations, common tape archive formats are often a problem. And, if a dedicated line isn’t available, transferring a large evolving database between remote sites often involves creating and mailing huge piles of CDs.

The familiarity of theWindows interface lulls new users into a false sense of security – just because it looks like other Windows applications doesn’t mean it’s easy to use. And understanding the modelling philosophy of a new system is not necessarily helped by the familiarity of the interface.

Knowing your way around a mid-range solid modeller neither makes you a good designer nor means that your models will be useful or usable by someone else. So good modelling practice is crucial for successful collaboration.

However, the success of ‘sketching in 3D’ is subjective and depends on the designer’s experience and skill sets. Often, sketching final solutions directly on to CAD is a disaster.

Good design practice continues to include simplifying 3D issues, by cutting 2D sections and trying out ‘what-if’ solutions on full-size plots, before committing to modelling in 3D.

If you’re not clear what you are trying to achieve, repetitive actions of similar modelling tools can clog up model files, make them too big or unusable.Model structures and history trees need to be simple andobvious if your models are going to be useful to someone else and be easy to navigate around.

To avoid this kind of thing, winning modelling strategies are:

1. Create a parallel not a serial tree where possible, leave Boolean unit features until the end of the tree.

2. On very complex models, assemble clearly-built and signposted parameterised milestone models. If all else fails you can unite these to form un-parameterised building blocks, enabling future editing.

The extent of how easy it is to edit models depends on the level of design authority between collaborating partners – some designers will only collaborate if they have authority for their parts and they share them in a frozen condition so that they can be used in other assemblies.

Essential rules of engagement should include clear strategies to indicate tolerances – modelling in reality for manufacturing tolerances, glue gaps, clearances, limits and fits, and sensible cutter sizes and radii.

In all ‘mature’ modelling systems, there are many different ways to create even a simple geometric entity or feature. Which is the best is entirely dependent on the designer’s preference.

As the model increases in complexity, constraints mean the system will only allow one method at that point. An experienced user can develop advanced modelling techniques based on familiarity with the system, and it is not uncommon to come across models which are unable to be edited by the average user because of advanced or creative modelling practice.

CAD data translation is not the painless experience that most CAD vendors would have us believe. Even modern native file translators only work, at best, 90 per cent of the time.

Poor data translation will often mean sacrificing the history tree. Many mid-priced packages are unable to import files from their more expensive stablemates without losing editability.

There are a number of ‘gotchas’ that can cause pain when data translation fails (or appears to work). The radius/diameter that has been converted into a spline or NURB, is very common, rendering it difficult or impossible to measure (because it’s no longer circular). Also common is losing the ability to distinguish adjacent tangency or curvature control on poorly translated surfaces. In the end, recovering from poor translations relies on getting an experienced CAD user to extract what he can and re-build the model.

Healing software is also not the panacea. It often doesn’t make ‘pretty’ decisions on visible automotive Class A surfaces where the slightest ripple will bounce light all over the place. It can also be dangerous for tolerance-critical applications because it relies on loosening the tolerances of the original surfaces in order to bridge gaps and fill holes.

Collaborative engineering is now an established reality in many industries – but communication is the key and sorting out the housekeeping can go a long way to making it successful and painless for all concerned.