Letters to the Editor: On Engineering and Educational Software, the Value of Conferences
November 15, 2011
To the Editor:
Many industries use computational software to model and simulate multibody dynamics (DAEs, ODEs, and eigenvalue problems). I support Alex Barnett's (Letters to the Editor, June 2011) concern with current software quality. I spend too much time separating a system's physical response from the errors. Spurious solutions and numerical chaos can dominate critical simulations, magnified by a managing engineer's drive to get a process done immediately, right or wrong.
This challenge is accelerated by co-simulation software, whereby one system problem is divided into a few independent solvers connected by gluing algorithms, e.g., an MBS code linked to a tire model. State-dependent terms are changed into applied forces, and mild delay differential equations are created. Another problem is that common ODE and DAE solvers correct solutions for error tolerance and not the total system's energy balance and integral invariants. More engineering software needs both to verify solution accuracy over a wide array of physical problems and to warn of limitations.
Engineers model mathematical problems, and applied mathematicians create solutions. The two groups need better collaboration, which can begin with SIAM News Letters to the Editor.
Allen P. Kovacs
FEV, Inc.
[email protected]
***
To the Editor:
I would like to share my positive experiences with providing short compact educational software to students, collaborators, and the scientific community. Somehow, it appears that a code that is easy to overview and actually works will catch people's attention and willingness to learn much more than equations, pictures, or even pseudo-code will. I have found that if a code is short enough, people tend to learn what every line does and make the modifications they need on their own, while longer codes are often used as black boxes that are simply "expected to work."
This creates a new challenge for us as educators and programmers---how to write codes that are as short as possible but still useful and general. To some extent, this is in conflict with common good programming practices, such as input validation, handling of rare cases, a single statement per line, short sub-routines, good commenting. . . . One could even argue that we promote bad programming habits with our beautiful but compact code examples. I believe that educational software is a category by itself, with a different purpose and therefore different requirements. Besides, since the users understand the entire code, they can easily make these improvements if needed.
Per-Olof Persson
University of California, Berkeley
[email protected]
***
To the Editor:
What is the value of conferences? With budgets tight and videoconferencing available and the ever-present concern about global warming, more and more people are asking this question.
But I think that people who ask the question often miss the main point of a conference. Talks, while important, serve mainly as an introduction to encourage participants with similar interests to develop fruitful collaborations. While these collaborations could be developed long distance, current technology does not seem to be sufficient to foster that, and it is not clear that it ever will be.
One thing that I think our community should do is to talk more about what it is that is really important about conferences. Faculty members should discuss with their graduate students strategies for getting the most out of a conference. It is easy for outsiders to conclude that if participants do not attend most of the talks then they are goofing off, and, well, sometimes they are. But sometimes they are accomplishing far more than could be accomplished by simply sitting through talks and more than would likely be accomplished through videoconferencing.
Anne Greenbaum
Professor of Applied Mathematics
University of Washington
