Physical Modeling Consortium serves a wake-up call on the automotive industry: Part 1 - A looming problem

Over the past two decades, powerful engineering modeling and simulation systems have evolved to tackle the fundamental technical challenges of MBD, and as a result, the names of the more popular software have become household words. So far, so goodor is it?

Recently at a meeting of the new Physical Modeling Consortium (PMC) held at consulting company IAV (a spin-off from the Volkswagen group) headquarters in Berlin, Germany, Alex Ohata, a senior engineering manager with Toyota in Japan, presented, for many of us, a shocking statement. He projected a critical lack of analytical and modeling tools as the complexity of design tasks increases each year. Furthermore, he warned of the consequences due to either complacency or ignorance among engineers.

"Taking the countermeasure may be delayed due to the hesitation for the investment and the subjective impression that we have done well. However, the delay will cause more difficult situations in the future," stated Ohata. [Ref: Alex Ohata, Toyota Motor Corporation, "Vision of Future Plant Modeling; Model-Based Development (Plant Modeling Environment)", Second PMC Meeting, Berlin, February 21st, 2008, Alex Ohata]

The PMC is a new organization, founded in 2007 by the engineering software company Maplesoft and Toyota to raise awareness of alternate mathematical modeling frameworks for physical systems. The organization's establishment came at the heels of a major announcement by the software company of a multiyear agreement with the automaker to collaborate on the development of a new generation of software tools for modeling physical systems.

The term "physical modeling" is becoming more common and refers to modeling techniques that are most concerned about explicitly working with the physics that underline a modeling context. "Physics-based modeling" is sometimes used for the same context. In the world of control, physical modeling is usually synonymous with "plant modeling," with "plant" referring to the system being controlled. The emphasis on physical modeling contrasts the traditional paradigm in engineering modeling and simulation that stresses time-based signal flows (such as in Simulink from The MathWorks).

To make things even more complicated, some use the term "acausal modeling" to stress the fact that the model layout is not based on temporal signal flows but on basic physical component connectivity: If a spring is connected to a mass, then the modeling environment should have a spring connected to a mass. The traditional approach is called "causal modeling" to stress the time-based signal flow.

This confusing vocabulary is a clear indicator of the emerging problem in industry. Many people know there is a problem and are starting to talk about it openly, but the news is still fresh and we have yet to establish a cohesive framework for communication or collaboration. This is exactly why the PMC was created.

One of the most harmful aspects of inadequate tool support for the physical or acausal view of modeling is that the only option remaining is manual calculations, using paper and pencil. The reality is that causal systems, which are great in real-time application and control strategy development, still require a clear sense of what the underlying model equations are in some detail. This means that someone needs to sort out the physics and assumptions and deduce the final equations step-by-step. Reams of paper and piles of chewed up 2H pencils later, a devoted engineer will get some equation—he or she will also pray that no minus sign was dropped a few pages ago; otherwise, it's back to the drawing board.

Historically, automation of this derivation process was considered out of reach for most engineering software—and as Ohata's comments seemed to imply, people seemed happy enough that some part of the design cycle was well automated! Over the past decade, however, developments in specialized mathematical algorithms (in particular, symbolic algorithms) coincided with this emerging need for model development automation, and, in a scant few years, the pieces began fitting together.