Beating the CAD interoperability nightmare

CAD interoperability – converting and running engineering models across different software vendors’ systems – has been a nightmare ever since the formative days of CAD/CAM, but of late it’s reared its ugly head far more. Difficulties used to be confined to the machine shop and analysts’ office; data translation only became an issue for designers when one CAD system was dumped in favour of a new one, leaving serious legacy data behind. However, with globalisation, collaborative working and the gradual move in some industries towards PLM (product lifecycle management), CAD interoperability has become a real challenge for many more designers and managers. It’s getting easier, but only slowly. As Peter Thorne of engineering IT analyst Cambashi says: “It is amazing the lengths that people go to, to avoid having to translate data. But the good news seems to be that geometry translation technologies are gradually getting the upper hand on a problem that always seems to be so much more difficult than it at first appears.” Old problems first. According to John Meaney, CAD translation software vendor TranscenData’s sales director, “Our own research suggests that 20% of a typical NC programming job is spent preparing data for processing, a figure that rises to 70% for FEA (finite element analysis) or CFD (computational fluid dynamics). This is crazy – highly qualified engineers being paid to answer complex questions, but spending two-thirds of their time second-guessing design intent and patching up messy CAD models.” Over the years there’s been a multitude of ‘solutions’. One of the first was neutral file formats – IGES was (and for some still is) popular. But success depended on compatibility of the translators in the native and target systems. If you were lucky it worked 95% of the time, but you could bet your job it would fail when under pressure. STEP is the ‘modern’ neutral format for 3D data, but it’s taken an age to come on stream and its effectiveness ‘stand alone’ is still patchy. Hence Autodesk’s DXF popularity, and celebrated CAD analyst Dave Burdick’s assertion that the only 100% guarantee of accurate translation is ‘like for like’, which still holds true. But ‘like for like’ is not that simple. Attempts to mimic it using the kernel approach relied on systems using the ACIS or Parasolid geometry kernels outputting SAT or XT files that could be read by other systems based on those kernels. These were better but still not 100%: developers develop in isolation, albeit using a common set of ground rules, so there’s still scope for interpretation and incompatibilities. Tolerances also vary from system to system, resulting in surfaces moving too far from the curves that define their boundaries, or neighbouring surface edges being close enough to ‘fuse’ in one system, while another stubbornly refuses to do so. The tiny gaps that result – so small you struggle to spot them on screen – play havoc with an FEA program that has to build a mesh around them. Briefly in the late 1990s Microsoft OLE (Object Linking and Embedding) seemed to promise a brave new world, with ‘drag and drop’ transfer of CAD data as objects between dissimilar applications. But it never took off because so few vendors implemented it. UK engineering leads So here we are, still waiting. A surprise development recently was that of e-drawings from SolidWorks – a technology that represents views on a drawing as ‘virtual’ solids that can be viewed from any angle and interrogated. The data format is compact for e-mail and distribution on floppy disks or whatever. So good are they that e-drawings are now set to take over from DXF as the people’s choice for solids. Incidentally, to give a clue for the scale of this, at the recent SolidWorks 2003 launch, the firm listed 30 data exchange utilities for the software, ranging from IGES to native model converters, and 11 specialist data exchange companies. But today the issue of data translation has gone further. On the one hand, we need feature recognition so that model intelligence can be transferred, and on the other, we want repair and healing for data that’s damaged or incomplete. And this is where the specialist data exchange companies come in, with the UK leading the world through companies like Theorem Solutions and TranscenData. Theorem is now the partner of choice for virtually all the major CAD vendors for bespoke data translation and direct converters. The firm also provides web-based translation facilities using CADVerter.com, a dedicated conversion portal. “Cadverter.com is the world’s first comprehensive e-engineering data exchange bureau service,” says Stuart Thurlby, managing director. “It is quick and easy to use, and helps hard-pressed design engineers free up valuable time to devote to the core aspects of their work.” He would wouldn’t he, but so does legendary car maker Aston Martin, which used the software on its new Bond car, the V12 Vanquish. The car is the result of collaboration between Aston Martin, Cosworth, Ford Research & Vehicle Technology and Lotus Engineering on the design side, plus a multitude of suppliers and subcontractors for manufacturing. “In the few instances where translations were partially incomplete, Theorem’s CADhealer restored these without the need for re-work at a CAD workstation,” says Aston Martin CAD manager Jordan Andress. “It helped us achieve significant savings in effort, time and cost.” There’s yet another problem: poor model quality. As Meaney says: “Designers often don’t fully understand how other applications work. They will happily sculpt the shape they want with repeated geometric operations, little realising that tiny residual surfaces they leave behind can represent massive overheads when it comes to meshing or toolpath generation.” Harald Hofmann, manager of the CAD system management department at Loewe AG, one of Germany’s most renowned TV and loudspeaker manufacturers, explains. “We had a particularly tricky problem with the data model for the rear wall of a TV set. Transfer errors had caused such severe difficulties for our designers that surfaces had been superimposed twice – even three times – making the model almost impossible to work with. CADfix was able to filter out the duplicated surfaces and the model was quickly ready to be imported into [SDRCs] Ideas [CAD].” Similarly at Peugeot-Citroën. “The CAD models produced had been problematic; the format changes they had undergone made them almost impossible to utilise directly,” says Nicolas Tournier, an engineer at PSA. “CADfix allows us not only to repair the CAD models quickly, but to generate a much higher standard of model which can be immediately worked with in VR (virtual reality).” And again at Cosworth Technology (the VW/Audi company), which accepts geometry from multiple systems but still needs help with data exchange. “We recognise that different modellers are particularly well suited for different engineering tasks,” says Mike Robson, principal CAE systems engineer at Cosworth. “We might use Pro/E for a cylinder head design, or Catia for a complete engine assembly.” So switching between modelling environments is key at this level, but also when it comes to FEA format conversions. Robson cites an early example of converting a cylinder head from Pro/E, which had caused several hours of problems until CADfix was used and resolved the geometry into a coherent solid, generating a fully surfaced model. The fact that it did the trick has meant use and popularity of CADfix spreading at Cosworth. Meaney sums the situation up nicely. “We are still a long way away from true engineering data interoperability, but the situation is improving all the time. The technology is becoming more sophisticated – and in some cases more tolerant – and better working practices and automation will help further. It may be that the holy grail of universal interoperability is not such an impossible dream.”