Mechanical Design Atomation - Hype or Hope

What's in a name? Quite a lot, actually. Words have perceived meanings and when the reality doesn't match the perception, confidence is lost. The engineering IT business has historically been rather full of examples of this syndrome and some would say that Mechanical Design Automation (MDA) is a case in point. To a purist, 'automation' might imply taking the human out of the control loop. So how can it be possible to fully automate such a cerebral process as mechanical engineering design, at least until every one of a myriad decision rules are written down and codified?

The answer is the classic one, "With difficulty." An electronic engineer would describe mechanical engineering as "analogue" in the sense that there are no obviously right or wrong solutions to an engineering requirement, only better or worse design optimisations. Fully automating the design process may not be impossible but it's certainly taking a long time. It is easier in some areas than in others, depending on the level of complexity and the extent to which engineering standardisation has occurred.

Right now it looks as though the constraints on design automation development are financial, cultural and business rather than technological ones. Much of the necessary technology exists, but the costs of bringing it to a wide market across most of industry are not yet justified by the demand. The cost of the necessary computer power is a key issue and may well be the controlling factor for some time. Let's hope that Moore's Law continues to hold.

We also know that in really complex situations the best solution is usually to design the human into the control loop, not out of it. At present we have no choice. We are evolving from 'computer aided' towards 'computer automated' design, with the human still firmly in the loop though using a growing number of design and engineering tools within an interactive operating environment. These tools contain increasing amounts of automation at low level, but at higher level they still present 'islands of automation'. Joining up the islands is the current focus of activity in our industry. The emerging challenge will be to roll up all this integrated technology behind a user interface dialogue that talks in terms of product specification performance and market requirements. That's harder and we must, as always, distinguish between the vision and the marketing hype.

There are two kinds of acronyms in the engineering IT business - those that clarify and those that obscure the facts. The term MDA, when first introduced, fell pretty squarely into the second category, the purpose being to imply that the product would deliver more than it ever could. I well remember the snorts of derision among engineers when some of the more imaginative marketing folk presented their 1970s era computer aided drafting products as MDA. Engineers, who tend to have a liking for precision and accuracy, observed that these systems automated little except the movement of a pen, and certainly did no designing. They didn't even make the drawing process significantly faster though they did make draftsmen more productive because drawing changes and modifications could be done so much more easily. That impacted the design process, because design is largely about iteration and optimisation. A very small first step towards MDA had been made.

A few nuggets of genuine automation of the drafting process followed. Ashlar notably developed Intelligent Drafting Assistant technology, which allowed the system to anticipate drawing intent and automatically generate the draftsman's likely next move. This seemed pretty smart at the time, but its prize-winning status indicated how far from anything approaching design automation we were. Parameterisation and high level programming languages like GRIP from Unigraphics were more significant inventions from the productivity point of view and here we see some early pointers to the evolutionary path for MDA. Someone willing to think out, codify and programme the drawing structure for a component or sub-system could save a lot of time. It was hardly design automation but it introduced the key concept of parametrics that was to become one of the main foundation stones for all future progress towards MDA.

The advent of parametric solid modelling in the 90's marked such an important revolution that our friends at PTC may perhaps be forgiven for persisting in their rather imaginative use of the MDA label when referring to their MCAD products! The real significance of this technology was the paradigm shift in the kind of 'thought processing' environment presented to the user. Designers and engineers suddenly wanted to get hands-on to the new generation CAD systems, not simply to produce better kinds of design documentation but in order to help them think about the product, to experiment, iterate and optimise. Here at last was a system for interactive digital product modelling as distinct from drafting, the drawing output being more or less automatic. At the same time the fundamental structure of the computer aided designing process became far more rational as object orientation became not only a key advance in programming methodology but a feature of the design process that was very familiar to any engineer.

Having got the operational interface right, MCAD system designers could concentrate on automating as many as possible of the low level operations that make up so much of the design process. In parallel they introduced somewhat higher level automation for commonly encountered tasks such as weldments, sheet metal design, plastic components and mould design. Characteristically, these are areas where the manufacturing technology that underlies design has become sufficiently standardised to make this level of automation worth developing. In economic terms, however, the increase in engineering productivity gained from these levels of automation is less significant than the opportunities that the new MCAD technology created, on the one hand for integrating CAD and CAE and, on the other, for managing the whole of the product development business process. As MCAD systems have become powerful enough to model and simulate complete product assemblies, with all their defined relationships and attributes, the focus of technology development has shifted towards tools for managing the whole collaborative product design and development process, with its complex work flow, information communication and decision making requirements. Hence the need for a new acronym - Product Lifecycle Management (PLM). This has been criticised for meaning whatever we want it to mean though it is, in fact, an important indication that the focus and style of our industry has undergone a major shift. By its nature PLM is rather reminiscent of traditional information processing. It's not surprising therefore that the consultancy component of the so-called design automation business is on the increase.

So is MDA a dying topic or the next revolution? Mike Evans, discussing the analogy between MDA and EDA in this column (July 2003), has pointed out some of the fundamental difficulties facing attempts to emulate the integrated circuit industry's approach to design. In that industry the design of many products nowadays proceeds by describing them at the level of the performance specification - all logic design, circuit design, physical layout and manufacturing masks being automatically generated by the system and validated by simulation. Mechanical engineering presents a far more complex challenge than an integrated digital circuit but there is nothing wrong in principle with the concept of top down automatic design, starting with a statement of the performance or functionality the product is meant to deliver. Some of the barriers to adopting such an approach are cultural, professional and economic. And, to some extent, the existing MCAD technology, and the way we have learned to apply it, tend to get in the way of the new 'numerical engineering' mind set that will be required.

I am old enough to remember how aircraft gas turbines were designed numerically long before CAD was invented; using computers hardly more powerful than today's pocket calculators. Design started by specifying the engine at 'a high level of abstraction' - the performance requirement - and designing top down to a well defined set of rules and analytical procedures. The keys to success were the systems engineering approach; a high standard of engineering knowledge that was religiously documented; and a culture of numerical engineering. In replicating such an approach we will see increasing overlap between so-called MDA and Knowledge Based Engineering. As always, the aircraft industry is taking the lead in implementing rule-based design technology such as ICAD, and it is not surprising that Dassault Systemes has bought up the technology. The challenge now for our industry is to make this kind of technology more affordable to the broad sweep of product design companies.

Given the state of the art in aircraft and automotive design we joke that, "Their products are all starting to look the same". This is almost inevitable as those industries approach the ability to optimise designs almost perfectly against a given set of performance requirements, design rules and common manufacturing constraints. Other industries will follow suit as the technology becomes more affordable: it's only a matter of time.