What exactly is Engineering Optimization

[Aero51]

What exactly is "Engineering Optimization"

I have the opportunity to take a grad class entitled "Engineering Design Optimization" this fall and I am curious about the subject. I have never taken an optimization class before and to be honest I don't quite understand why optimization is sub-discipline of engineering. In every engineering problem there is always the issue of design. More specifically, what is the best design? And upon further inquiry: what is 'best'?

I would guess then the "optimization" discipline is the study of ways to achieve certain engineering objectives.

What, then, do professors and those in industry do whom study optimization? What is to be expected out of the standard optimization class?

 

[SteamKing]

In general, the object of engineering design is to, among other things, make it lighter, faster, stronger, and most importantly, cheaper. It also must meet applicable regulations and not fail in service.

 

[timthereaper]

Generally, for any given product, you have some sort of design metrics you're trying to achieve and others you really don't care about. In aircraft design, for example, you choose what you want the plane to do and design accordingly. Fighter jets have different metrics than commercial planes, so optimization is really just trying to adjust your parameters to achieve your goals. Maybe you favor speed over fuel efficiency or endurance or whatever. Taking it down to a structures example, maybe you care about strength to weight ratio over cost of materials, etc. Optimization is really the art of mastering the trade-offs in design and to some degree, it really doesn't matter what you're optimizing because the techniques tend to be really similar across the board.

The class you're asking about is probably going to teach those general optimization techniques. From talking to people who've taken a class like that, they said the class covers Pareto efficiency and determining the Pareto frontier for product designs. If you're interested in focusing on product design, I think it would be a great class to take.

 

[skeptic2]

This might give you a feel for what you're in for.

http://books.google.com/books?hl=en...ring optimization theory and practice&f=false

 

[berkeman]

I have the opportunity to take a grad class entitled "Engineering Design Optimization" this fall and I am curious about the subject. I have never taken an optimization class before and to be honest I don't quite understand why optimization is sub-discipline of engineering. In every engineering problem there is always the issue of design. More specifically, what is the best design? And upon further inquiry: what is 'best'?

I would guess then the "optimization" discipline is the study of ways to achieve certain engineering objectives.

What, then, do professors and those in industry do whom study optimization? What is to be expected out of the standard optimization class?

This might give you a feel for what you're in for.

http://books.google.com/books?hl=en...ring optimization theory and practice&f=false

It sounds like it could be an interesting and useful class, depending on what is taught. skeptic's link looks mostly mathematical, but I didn't read very far in it yet. If it's just a dry math class, it could be pretty boring. But if they bring in lots of real-world examples, and talk about different kinds of optimization, that could be a good class.

Can you post a syllabus or class description? What textbook do they use?

More advanced topics in such a class would be things like:

• In mixed-signal (analog/digital) ASIC design, how can you best decide which functions to do in analog versus digital, when you have a choice (like filtering, etc.)?

• In intellectual property considerations on new product designs, what should you patent, and what should hide inside your product (inside the chip or the software program)?

• When scheduling and managing engineering projects, how do you trade off working on certain opportunities at the expense of other opportunities ("opportunity cost")?

• Etc...

 

[Aero51]

Course Description:

MAE 531 Engineering Design Optimization UNITS: 3 - Offered in Fall Only
Prerequisite: Graduate standing in Engineering is recommended.
Nonlinear optimization techniques with applications in various aspects of engineering design. Terminology, problem formulation, single and multiple design variables, constraints, classical and heuristic approaches, single and multiobjective problems, response surface modeling, and tradeoffs in complex engineering systems. Numerical optimization algorithms and implementation of these optimization techniques. Graduate standing in engineering recommended.

Book

Engineering Optimization: Theory And Practice Ed: 4 Th Yr: 2009

It was between this class and airfoil theory and quite frankly airfoil theory is pretty easy to understand and I already did a lot of self study in the field.

 

[berkeman]

Bwooha! Looks pretty dry from the description. Do they have the book at your library or bookstore, so you can look through it? Can you talk with the professor, to see what kind of real-world examples they explore in class?

 

[Aero51]

I can just ask my friend, he took the class last semester. I texted him about it but he never got back to me. Either way I'll see him in a few days and will find out. But also remember sometimes professors onethirdass things and don't cover what the description says at all.

 

[berkeman]

Cool. Let us know.

 

[Aero51]

I talked to my buddy about the class. He basically said it was very computational and theoretical. He didnt like it because he isn't the best at programming. I think I'll take the class over airfoil theory as it has a much broader spectrum of applications and it will be good programming practice.

 

[carlgrace]

More advanced topics in such a class would be things like:

• In mixed-signal (analog/digital) ASIC design, how can you best decide which functions to do in analog versus digital, when you have a choice (like filtering, etc.)?

• In intellectual property considerations on new product designs, what should you patent, and what should hide inside your product (inside the chip or the software program)?

• When scheduling and managing engineering projects, how do you trade off working on certain opportunities at the expense of other opportunities ("opportunity cost")?

• Etc...

While what you listed sounds like a very interesting and truly useful class, I don't think that is what the course in question is going to be about. An optimization class in engineering (I took one) is about casting (or you might say contorting) engineering problems into specific forms with constraints that allow a computer program to maximize or minimize some set of desired parameters.

Typically these techniques boil down to "convex optimization" schemes which is any system of equations with which you can find a global minimum using a gradient descent algorithm. So, in a sense, optimization is very similar to adaptive filtering from a math standpoint.

From an IC design standpoint, these tools are not helpful yet in partitioning SOCs between analog and digital blocks. They have typically been used on rather stereotyped systems for choosing parameters. For example, a pretty well known optimization tool allows you to enter jitter, bandwidth, and spur levels (along with some other stuff) and then it optimizes a PLL using a set architecture, or it tells you it's impossible. These tools can't yet invent new structures.

Interestingly, a few companies have been founded purporting to have finally solved the Analog Synthesis problem (one of holy grails of integrated circuit design automation). They have invariably over-sold their technology and then crashed-and-burned ... *cough* Barcelona Design *cough*

 

[timthereaper]

Too bad the class isn't what it sounded like. Good luck in airfoil theory.

 

[dabulper]

Other optimization techniques

While what you listed sounds like a very interesting and truly useful class, I don't think that is what the course in question is going to be about. An optimization class in engineering (I took one) is about casting (or you might say contorting) engineering problems into specific forms with constraints that allow a computer program to maximize or minimize some set of desired parameters.

Typically these techniques boil down to "convex optimization" schemes which is any system of equations with which you can find a global minimum using a gradient descent algorithm. So, in a sense, optimization is very similar to adaptive filtering from a math standpoint.

From an IC design standpoint, these tools are not helpful yet in partitioning SOCs between analog and digital blocks. They have typically been used on rather stereotyped systems for choosing parameters. For example, a pretty well known optimization tool allows you to enter jitter, bandwidth, and spur levels (along with some other stuff) and then it optimizes a PLL using a set architecture, or it tells you it's impossible. These tools can't yet invent new structures.

Interestingly, a few companies have been founded purporting to have finally solved the Analog Synthesis problem (one of holy grails of integrated circuit design automation). They have invariably over-sold their technology and then crashed-and-burned ... *cough* Barcelona Design *cough*

There exist other types of optimization techniques that mimic nature as: genetic algorithms, swarn particle optimization, simulated annealing, ant intelligence... These techniques are not so constrained as the classical ones.