The posting below came to me via an excellent CAD-focused ezine. It is reprinted with permission from upFront.eZine Publishing, Ltd.. Ralph’s focus in this post is normal manufacturing. Commercial construction is manufacturing, but we make large, complex prototypes (virtually no two buildings or other structures are identical in the way that more commonly manufactured items may be identical). The article was particularly striking because we are going through a large wave of start-ups, very often led by tech people who have only a very limited understanding how structures are built. A core issue in this post is the (large and significant) difference between “designing” a thing and actually “making” it. The issue exists on steroids in construction. If these issues are not obvious to you, reading this may help you understand.
|CAD ought to love CAM, but it doesn’t. The whole point to us designing things is so that things can be made. And so it’s kind of weird how much of CAD is hostile to CAM [computer-aided manufacturing]. While today’s MCAD software usually recognizes that the output of its designs is for manufacturing, drawings produced for BIM and civil projects do not.
Prepping for manufacturing isn’t particularly innate to the CAD software we use. We can draw the entire solar system at 1:1 scale, right down to the lettering on a lunar lander’s plaque (as an early AutoCAD sample drawing showed), but most CAD programs can’t prepare the drawing to manufacture the plaque. When we send the drawing to a plotter or a PDF file, we drafters and engineers see our job as done.
The gulf between CAD and CAM exists because CAD works with mathematical certainty, while CAM works with tolerances, a technical word for “uncertainty.” We designers take pride in our consistent use of tools that are precise, as CAD vendors reassuringly remind us, placing geometry using object snaps and parametric constraints while working to 14 decimals places of accuracy.
CAD is so precise that CAM operators cannot use the drawings we produce. CAM understands the real-world problem of tolerance, the plus-or-minus dimension in the order of 0.01mm or more that comes into play from the imprecision inherent in lathes running in machine shops, or the gears and hydraulics of graders guided by GPS while smoothing freeway expansion projects.
So CAM operators need to tweak CAD drawings. This is one of the marketing pushes behind companies like SpaceClaim and BricsCAD. The direct editing found in these programs allows machine shop operators to import models from “any” CAD system, and then adjust the drawings to suit CNC [computer numeric controlled] output, such as removing unneeded details (defeaturing) and extracting geometry for fixtures and tooling.
Most of us CAD users were never trained in manufacturability, and we so have no idea about it. Manufacturability is the extra information that ought to be included in CAD drawings so that parts can be manufactured without CAM operators needing to massage our output. Steel molds must include mild draft angles so that plastic parts pop out easily; sometimes the molds need pipes to deliver additional molten plastic to fill the mold properly. When the flanges of sheet metal are bent, allowance must made for the bend radius, known as the k-factor (or y-factor in PTC software).
There is little manufacturability in regular CAD programs. AutoCAD, Microstation, and nearly all other CAD programs are ignorant that wood cut for kitchen cupboards needs to account for the width of the cutting blade or how to nest the parts to minimize waste. Nesting places parts optimally on plywood or sheet metal to minimize waste. Regular CAD doesn’t do this.
The most basic manufacturability tools are GD&T [geometric dimensioning and tolerancing] and PMI [product manufacturing information] to tell manufacturers how much inaccuracy can be tolerated and the material to use. Only in the past coupleof years have translation packages boasted about being able to translate PMI. Regular CAD packages now offer STL[stereolithography], the universal format for 3D printers .These are baby steps, given that CAD and CAM have both been around since the late 1960s.
So: what’s changing? CAD is a mature market, as the CEO of PTC is fond of pointing out. There are no new customers, only customers to be enticed from competitors with functions competitors don’t offer — such as better output to manufacturing devices. Once one MCAD system does it, competition ensures all the others quite suddenly find themselves quite congenial to the idea. We’ve seen the same rush in other areas, such as direct modeling, generative design, and lattice generation.
The disconnect between CAD and CAM didn’t always exist. The first CAD systems were developed by McDonnell Douglas (Unigraphics, now NX from Siemens) and Dassault Aviation (now from Dassault Systemes) purposely to output to CAM.
But most systems developed differently. CAD software is made by PhDs comfortable with multi-dimensional matrix transformations. CAM software is developed by practical users, whose software typically reads in a plain DXF file output from CAD, then adds tooling and path information, and outputs it as g-code (the DXF of CAM). Their concern is for roughing strategies, 2- through 6-axis milling, and turning-drilling — not in finding another way to draw a line or extrude a surface.
G-code is not universal, unhappily. Pretty much every CNC machine offers a unique set of capabilities, and so the g-code needs to be tweaked for each one with post-processors — an annoying task for a CAD vendor that doesn’t particularly want to be in that end of the business in the first place. On the other hand, CAM software vendors like post-processors, as they make significant revenues selling machine-specific code.
The split manifests itself in other ways, too. Nearly all MCAD software is owned by just a few corporations, but there are forty hundred companies offering CAM software. They have names that we designers may never have heard of, like Alphacam, BobCad-Cam, CamWorks, Dolphin CADCAM, Esprit, FeatureCAM, GibbsCAM, Hypermill, Mastercam OneCNC, PowerMILL, SolidCAM, TurboCAD/CAM, Visual Mill, and ZW3D CAM. While many are home-grown, others are repackaged from OEM providers, such as those made by MachineWorks of England.
Here’s another split from CAD: CAM has its own language, with jargon like swarfing, flow cuts, toolpaths, cutters, multi-axes, point control, grooving, and inside roughing. Mold design is sufficiently complex that it is sold separately from CAM software.
Nobody know why the CAM software industry never coalesced. Some attempts were made, the biggest by Vero Software. They raised over $10 million to own a dozen CAM packages; but then in 2014 Vero was taken over by Hexagon. If MCAD vendors haven’t already developed their own CAM software, then most of them purchased one or more firms, such as Autodesk acquiring HSM, and 3D Systems buying GibbsCAM, among others. If they don’t own one, then they embed CAM software from third parties, as Solidworks and OnShape do.
There two ways to make a part: remove material or add material. Originally the output from drawings was subtractive manufacturing (machining). The more difficult problem is adding material, but even additive manufacturing (3D printing) is more than twenty years old now.
Awareness of 3D printing exploded a decade ago with those cheap 3D printers that were marketed for homes. Well, that ended as a bust. I recall one vendor stating they would place their 3D printer in every child’s bedroom, but then wouldn’t answer questions from us skeptical media about how children would be protected from fumes of melting plastic, the hot surfaces, and the post-print cleanup, which sometimes involve noxious chemicals. The samples that I collected from them still stank a year later.
The collapse of home 3D printing should not have surprised, as it had everything going against it: small parts printed at poor resolution that might well collapse on themselves, and that printed slowly. Children watching in awe lasted a couple of prints. The cost of the plastic filament refill was where these companies made their profit — expensive, in other words, like printer ink.
In contrast, the industrial version of 3D printing is booming, although here we are talking about machines costing $50,000 and up, that take up a lot of floor space, and requiring proper venting. I am fascinated by the new variations on materials announced every few weeks to output parts made of multi-colored plastics or sintered metals.
And so we have an industry balancing between promise and frustration. The promises of AM are wondrous:
· Outputting intricate parts that are impossible to make with subtractive manufacturing
· Producing spare parts on demand to eliminate dedicated warehouses
· Generating prototypes more quickly and avoid over-nighting to outside manufacturers
· Minimizing material cost and weight, especially in surgical and aviation applications
Sure, mistakes are made with subtractive manufacturing, but mistakes seem more precarious in additive manufacturing. AM works from STL files exported from CAD programs. STL is a simplistic format that consists purely of 3D coordinates for the myriad of triangles defining the surfaces of models; the format doesn’t even include units. It’s these surfaces and triangles that create problems for 3D printers:
· Too thin walls lead to collapsed 3D prints
· Non-watertight designs with one or more holes in surfaces, and edges that don’t match
· Too many triangles used to define surfaces or ones that overlap unnecessarily
· Triangles with inverted normals or ones that cut into each other
Normals are vectors that indicate the inside or outside of each face of a 3D model. When a normal points the wrong way, the 3D printer thinks the inside of the model is the outside of it.
Software firm Materialize reports that one in five parts fail during 3D printing, because walls were too thin. As a result, an entire sub-industry has sprung up to write software that fixes these problems and optimizes the 3D printing process, such as by shrink wrapping the CAD model to make a shell copy of the 3D solid.
We see market inefficiencies in the disconnect between CAD and CAM that have gone on for 50 years. Small firms are successful in targeting niches ignored by bigCAD, especially for kitchen cabinets.
The linkage between CAD and final product is pretty tight for MCAD, but completely lacking in other major disciplines, such as architectural and civil engineering. The complete lack of a CAM connection to BIM is a primary reason the construction industry is less digitized than most.
[This article first appeared in Design Engineering, and is reprinted with permission.]