MODELING  PHILOSOPHY 

ON MULTIDISCIPLINARY DESIGN AND OPTIMIZATION


The recent advent of modeling tools has masked somewhat the importance of the actual creative ideas in engineering. 
To be sure, the roles of the tools and the creative ideas they help materialize are strongly coupled.  It is very rare, if at all possible to find seminal engineering accomplishments without concomitant advances in tools.  In modern engineering, modeling has been used with a number of benefits.  These include better understanding, at least in teaching the material to new engineers, and the potential to reduce the design time and optimize the performance of products.

In electromagnetic tools for communications engineering, the fundamentals have been well established by Maxwell.  So, one can find several computational electromagnetics tools in the market solving Maxwell's equations with various degrees of assumptions and applicability.  There are ingenious methods published and even more ingenious details in making some of these tools work.  When it comes to electromagnetic design, however, the creativity that comes with it, is to a large extent decoupled from the modeling tool(s) used to simulate it.  Our view of modeling is that the judicial combination of modeling tools and design creativity has the potential for tremendous progress.  Even if one has good intuition about individual mechanisms, it is difficult to quantify the net results of all these competing mechanisms without modeling.  This is because complex designs typically involve several different mechanisms, each one pushing performance along a different direction.  Thus, modeling can help answer a lot of "what if" questions without necessarily having to create physical prototypes.  When it comes to the accuracy in predicting the performance of a complex model, the jury is still out.  We have found that the  amount of effort required to get good, absolute agreement between theory and experiment often is not realistic.  Product design cycles move too fast and do not allow time for "academic" studies.  However, this is not very serious.  As long as the results are close, modeling affords you to isolate the effect of certain input parameters and perform sensitivity studies on anyone of them.  Indeed, over the years, my team and I have been convinced that for complex problems, the low fruit in modeling comes not from predicting performance in an absolute sense.  Rather, the value of modeling comes from the ability to Design Experiments (DOE), perform parameter studies and examine the trends and the sensitivity of the important inputs.

Often, as the designs become more complex and characterized by the need to "stand out", they defy intuitive solutions from within a single discipline.  Multiple engineering disciplines, traditionally decoupled in our education system, need to be employed for the optimum design.  Scientists and engineers started by being very well rounded learners, basically knowing a lot of the body of knowledge that was out there.  Due to the exponential increase of knowledge, however, and the need for continual expansion, specialization occurred.  Engineers and scientists immersed themselves deeper and deeper into their areas of specialization.  The educational system, and subsequently, the organizational structures
in large companies followed.  The Mechanical Engineering department is in a different building from the Electrical Engineering on many university campuses. 

For some reason, Nature decided not to follow us :-)  Natural problems seem to be very multidisciplinary. 
Designing a modern desktop computer, one would think, is all about the Central Processing Unit (CPU).  However, no CPU could last long enough without a good heat sink and a good fan.  And, there would be no quiet desktop without a good solution of all the vibration problems of the computer's case caused by the CPU and its fan.




So, here are a few of the lessons we have learned after years of engineering modeling and simulation activities:

 Lesson 1: Satisfactory modeling tools may not exist for the state of the art.
By the time tools have been validated in the realm of applications of the state of the art, the state of the art, typically, has moved on.      




 Lesson 2: Trends and sensitivity analysis are often more important than absolute   accuracy.    Remember, a broken clock shows the absolute time twice a day, but it is useless as a time piece. 




 Lesson 3: For ultimate products, we need to perform Multidisciplinary Design and Optimization.




In what follows, we attempt to describe our view of engineering modeling and its evolution using examples from architecture.



THE pre-CAD ERA


                      
 


The Parthenon at the Athens Acropolis.  An impressive temple by Iktinos and Kalikratis built in the 5th century BC.  Some of its arcitectural characteristics shed light in the sophistication of the design in that, pre-CAD era.  For example, there is an optical illusion when one stares at a large building like this.  If its columns were vertical they would appear tilted.  Therefore, the columns of the Parthenon are "pre-distorted".  They are actually tilted with the end result of appearing imposingly vertical to the observer standing in front of it.  There are several other characteristics that this design has and can be explored here.



 
 



A Roman Aqueduct in Segovia, Spain.  This structure has been standing for 2000 years.  There is no mortar used to hold the plinths together.  The entire structure is kept together by gravity and virtue of its shape.  One critical  and creative idea that can be seen here is the form of the "arch".  This idea is an engineering "masterpiece".  The arch allows the spanning of distance with less materials than the straight pillars and trusses could achieve.  Again, one sees creative ideas at work.  No doubt there must have been a lot of experimentation prior to this achievement.  Many preliminary prototypes, several failed projects and many-many slave workers.



 





The Westminster Abbey in England.  This is an example of the Gothic Architectural style.  Its distinguishing feature is the height achieved.  It combines the straight pillars and the arch.  It uses large radius arches but not full in their angular span.  So, it brings the pillars closer together than a full 180 degrees arch would do.  This innovation achieves a larger vertical height that the straight pillars or the arches alone could ever do.  Again, there was a lot of experimentation and failed attempts trying to reach unprecedented heights by the church before the art of the Gothic style was mastered.






THE early-CAD ERA
 
 




The skyscraper is a marvel of engineering.  The beauty of the design is reflected by the goal achieved, namely, the maximization of volume and square footage for a given footprint.  This design represents the early CAD era.  Systematic solution methodologies were first invented to assist engineers, then, with the advent of computers, CAD tools were developed implementing these systematic analysis and solution methodologies.  Finally, when these techniques were applied and practiced for many years, it was not uncommon to develop hard-copy tables where several of the key characteristics of the design where tabulated for immediate access.






THE late-CAD ERA

 




The turning torso by Santiago Calatrava in Moelmo, Sweden.  This is an example of the late CAD era.  Although the physics of the problem here is well understood, tabulated results are not adequate to complete the structure.  Imagine, for example, the calculations needed to determine the wind loading on this structure.  Twisting and torsion must be a key element in the analysis of this structure and has ramifications to the strength of the materials chosen for its construction.  This design represents the trend of our era, "the desire to be noticed, to stand out".  Additionally, the economic realities of our times cannot afford us many prototypes.  Labor costs are expensive and democracies cannot tolerate many failed attempts or capricious waste of public monies.  Our new buildings, our new products have to be built fast, stand out and "work" the first time.





The road to Multidisciplinary Design and Optimization
 







The Walt Disney Center for the Performing Arts in Los Angeles, California, by Frank Gehry.  Like Calatrava's turning torso, this is an example of a late CAD era building.  The trend of being "noticed" is evident, perhaps at the expense of functionality, maintenance and/or cost.  One true story about this building points to the Multidisciplinary vision on Design and Modeling.  Apparently, after its construction, passers by and residents in nearby apartment buildings complained of high temperatures and hot spots (*).  The metal panels of the building were too shiny and the curved surfaces inadvertently formed large and efficient light and heat reflectors.  At a serious and non-budgeted expense, the metal panels had to be matted down so that they would not reflect as much (see the worker hanging from a rope while sanding down the panels in the second picture).  While Gehry started with an architectural design, normally an architectural and civil engineering effort, the building's characteristics had transcended it into a thermal problem for the space around it.  Multi-disciplinary design is needed to fully address complex problems in our era.


(*) When I saw a program on the PBS channel one night about this, I called LA real estate firms tring to confirm the story.  At the beginning, the agent thought I was interested in buying an apartment at a nearby building.  So, when I mentioned the increased temperature issue, I was told that they have fixed the problem, "... now the temperature is only a couple of degrees higher". 



Multidisciplinary Design and Optimization

Holistic design is needed to optimize complex problems.  For example, one cannot design the electromagnetic aspects of a cellular phone without taking into account thermal, acoustic and mechanical problems at the same time.  Although more difficult to train for and achieve, multidisciplinary design promises to render products with truly optimum performance.  Design tools compatible for, and capable of coupling together, different disciplines are needed and are being pursued by  several groups around the world. 

We are still at the very beginning of this effort and a lot more needs to be done.  New ways of looking at geometric representations,
so that they can be consistent for the physics-based analyses methods of various engineering disciplines are needed.