CAD is straight and square, generative design more a free spirit
Parts made with standard stock material are confined to having a standard look. For artists and other creative types, this is too much to bear. Architects are confined with residential buildings that are measured in multiples of 4-foot-by-8-foot sheets, the standard size of plywood in the U.S., and just as square and flat. Product design is also confined to few methods of manufacture, as stock material is round tubes, bars, channels, I-beams and the like.
To a lesser extent, parts made with CAD also suffer from limitations of stock shapes. One CAD shape may not look like another, but with the limitations in starting shapes, CAD shapes are not exactly as unique as fingerprints. Step back and you will see that the shapes generated with CAD programs look more alike than different.
CAD programs make parts from a limited library of primitive shapes. Solid models start as spheres, rectangular solids and a few other simple shapes, and are then combined to make a part. It’s a building block approach that yields—with enough time and effort—the shape of a part. A bit of variety can be added by extruding and revolving a 2D sketch to make a solid. But the idea of making almost any natural, organic, curvy shape is enough to scare off the savviest solid and surface modeling wizards.
It’s no wonder that CAD parts are straight, flat, round and symmetric. They have sharp corners and other characteristics that don’t exist in the real world. What we have succeeded in creating are unnatural parts—parts that exist because they are the only shapes our primitive modeling tools can create.
Parts are shaped by the tools available to make them, whether they are shaped by hand tools or 3D CAD software.
Asked about sculpture of an organic form of Y, an artist X is supposed to have said, “It’s easy to make Y, you have to chip away anything that doesn’t look like Y.”[i]
Why? Because when it gets down to it, the objects that nature makes are nearly impossible to model with CAD. Examine anything—a leaf, a tree, an insect, its digestive system.… Is there anything at all that occurs in nature that can be modeled with CAD? A few crystalline forms, so rare, so admired, if only because they are the freaks of nature—more the exception than the rule.
It may be obvious that CAD can’t possibly model Michelangelo’s David, but let’s admit that it can’t really model any part of him. Not his hair, his ears, his oversize hands, his undersized.… Well, you get the point. If not, the point is this: almost everything naturally made is impossible to model in CAD.
The Euclidean perfection created by CAD was welcome at one time. That was when perfection was rare—a novelty in an age of hand drawn designs when parts were created with hand tools, when no two parts were the same, when wood and metal were hewn and hammered. In our modern industrial age, perfection is plentiful. Perfection has become cheap and common. So common for so long that a slight imperfection is now a feature, with a retro appeal that’s come into vogue.
But, while imperfections may be pleasing aesthetically now, they are not by design optimized for any particular environment.
It is when a design can be any shape at all and when each shape can satisfy the conditions of its intended environment that generative design gains a creative edge. Reminder, we are talking about functional optimization, not aesthetics. Unfettered by bias, tradition, culture, and so on, generative design is a real free spirit, which is able to create designs beyond the imagination of a conservative or tradition-bound engineer. Generative design programs have no consideration, knowledge or bias toward what has existed before. They are not bound by history.
Multiple Load Cases
It is true that some optimized designs may appear outlandish or impossible to manufacture. Their appearance could make an engineer realize that the part could fail in a load case not entered into the generative design program. But this is a fault of omission by the engineer, not the generative design program. The only generative design programs worth using are those that are able to accept multiple load cases and design parts that can survive all of them.
Multiple load case capability in a generative design program will create parts that can withstand tensile loads and torsional loads, for example. A tension-only load case will create shapes with stringy shapes because what is better to dangle a load from than a string? A torsion-only load case will have load paths in the shape of a cylinder because what shape is better to resist torsion than a round tube?
Who knows what shape the combined load case will produce? But don’t worry. Generative design will figure it out, building up a part load case by load case.
This is theoretically possible only if the next load case adds to the shape that withstood the previous load cases—without removing anything from the shape. It has to be an additive process, with each load case allowed to add volume—not subtract it.
This is not how single load case optimizers work. A single load case that starts with the shape generated from another load case will undoubtedly undo the optimizations of a previous load case, rendering it null and void.
The ability to preserve geometry, no matter how complicated it is, is critical for real-life design.
Where X = {Michelangelo, John Ruskin, George F. Pentecost, Boys’ Life Magazine, Orison Swett Marden, The Methodist Quarterly Review, Anonymous} and Y = {David, Venus, angel, elephant, masterpiece}. From QuoteInvestigator.com.
