The End of CAD


Design is mystical. Something does not exist...then it does. I think it is a special way of touching the mind of God, validating man having been created "in the image of God."

Design is not deterministic. Given a set of requirements, there is no formula into which they can be pumped to yield an optimum design - or even a good one. The process of design is invariably one of trial and error.

It´s fascinating and counterintuitive that the process offers no promise of convergence - nothing guarantees that more trials will get you closer to an optimum design. That makes great designs all the more impressive.

The only improvements in the design process for centuries were in documentation - stable media, better pencils and pens, and mechanical drafting aids. In his 1950 science fiction story, "The Door Into Summer," Robert Heinlein described Drafting Dan, a keyboard-controlled drawing robot whose essentials began to be realized in the early CAD systems of the ´60´s.

Then came 3D, photorealism, kinetic motion - and design documentation acquired more verisimilitude at the pace of Moore´s Law, by which computing power increases ten-fold every three to four years.

Design can be characterized as a dialectic, a term used by the philosopher Hegel to explain the spasmodic rhythms of history´s unfolding. While it may not work well in history, it does work for design: In the design dialectic, the "thesis" is the concept in the designer´s mind. The "antithesis" is the external manifestation of the design-drawing, model, or prototype. A designer brings the thesis to bear on the antithesis to yield the third stage of the dialectic-synthesis. This is usually a refined model or prototype, which becomes the "thesis" for the next cycle.

Although there is no guarantee of convergence, designers need to iterate-to go through multiple cycles of the dialectic-to refine the design. This makes the computer a great design tool. It can build detailed and accurate models quickly and inexpensively, and test them to reveal inadequacies, thus providing the basis for another iteration.

But for the design tool to be truly wonderful, it must create as complete and precise a model as can be achieved-and as complete an environment as possible in which the model can be embedded. For example, product behavior is best modeled in a "world" in which "gravity" and "friction" are provided by the computer.

Simulation models can be produced either by analysis or by synthesis. Analysis is the hard way because it calls for finding mathematical equations to represent the functions of all parts. But many behaviors are difficult to represent in this way, so building models can be impossible.

A synthetic approach makes no assumptions about underlying mechanisms. It simply seeks to produce a "converter", an equation or mechanism that imitates reality. Given the same inputs, converters produce outputs similar to those of the systems being modeled.

This behavioristic approach is at the root of a powerful therapy called neurolinguistic programming, or NLP. Ideally, practitioners are taught to observe behaviors, and not to attempt explanations. (Sadly, the ideal is not often achieved, but that´s another article.) Making behavioral choices does not "mean" anything, except perhaps that they need to know about other choices. No assumptions are made about what people think or feel.

Similarly, we can create simulation models that behave like the things we are modeling without knowing or caring whether the mechanisms that convert inputs to outputs are in any way the same. For example, a flocking algorithm by Craig W. Reynolds models the behavior of flocks of birds, schools of fish, and other groups in motion by applying simple rules to the behavior of each member of the simulated group. The visual effect is astonishingly realistic. The point is it does not matter whether or not birds or fish use a similar set of rules to govern their behavior when they move in groups.

"End" has multiple meanings. It can mean the terminus, the final point; it can also mean the goal. In this essay´s title, I used it in both senses. The terminus of all CAD will ultimately be the creation of simulation models. And the goal of CAD is, in fact, simulation.

Simulation, although a specialty, ought not to be. Anyone who thinks about products should see them work on screen where mistakes are easy and inexpensive to find and fix. We are not there yet. But if CAD vendors listen to me, we will be in a few years.

The design processes must also be designed. They can be workflows for engineering and manufacturing, for example, or even information flow within an enterprise. The complexity of such systems makes them even better candidates for simulation than product designs.

To help your company realize the full potential of engineering, think about modeling the business and the role of engineering. A tool that does so simply is ithink at www.hps-inc.com.

But however you do it, do it, and help your organization consider the full context of its business.