Why Simcenter HEEDS ?

Don’t Just Simulate, Innovate!

HEEDS Software

Why the Need for Simcenter HEEDS Software?

• It is unquestionable that these types of design questions arise every day and in EVERY industry (not just a special or occasional issue)
• With Design optimization Software like HEEDS, it should be possible for EVERY simulation engineers to answer these questions quickly and easily (no special expertise necessary)

These type of design questions aren’t infrequent.  They arise every day in every industry.
 

Automotive

For instance, in automotive applications, we need to know the number, location, and size of holes to place in a cooling jacket gasket to arrive at relatively uniform temperatures between the engine head and the block.
 

Aerospace

Or, in aerospace, engineers need to decide on the taper, twist, and cant angles of winglets to minimize drag and maximize lift.
 

BioMechanics

Or, a biomedical engineer may be tasked with coming up with a new syringe design shape that easily draws blood without rupturing red blood cells, but can’t easily be used multiple times so as to limit the spread of infectious disease.
 
These design questions ought to be easily answered by any simulation engineer without undue difficulty.  So, you might ask yourself, how easy is it for you to use simulation to answer these types of questions today? Would it help if the process were simpler and faster?

Featured HEEDS Success Case Study: Becker Marine Systems

Mewis Duct design exploration delivers $500k saving per ship in 6 weeks

• Deliver 4.5% fuel savings in 6 weeks

• Saving $500k per ship/year

• 1000+ ducts installed

There is tremendous value in using design exploration to find better designs (cost savings, competitive advantage, brand strength, etc.)
The Automated design exploration methods we will show you allow you to consider many more design variable changes (40 considered here) than previously possible
The stakes for performing design exploration and driving innovation are quite high.  Let’s consider the recent experience of Becker Marine Systems.
This company creates energy savings devices for marine vessels. One type of device is shown here: it’s called a Mewis Duct and it is basically a flow-directing device positioned near the propeller. 

“The success of the Becker Mewis Duct® depends almost entirely on the CFD process that we use to define it. Without accurate CFD simulations, we wouldn’t be able to tune each duct to the specific flow conditions generated around each hull.”
– Steve Leonard, Head of R&D

The key to the success of Becker Marine’s product is that the particular shape of the duct that provides the most energy savings varies depending on the hull geometry and propeller. So, what they’ve done is created a process where they calculate the flow field around the hull and propeller, then they automatically explore the design space varying 40 design parameters to define the duct geometry. And what they are able to achieve is a 4.5% improvement in fuel savings, delivered in 6 weeks.

What does that 4.5% fuel savings mean? Well, it’s approximately $500k per year for each ship. Thus far, they’ve installed over 1,000 ducts, for a fuel savings of over $500M per year.
This is the type of value that is possible to achieve using design exploration!

Driving Innovation Through Design Exploration

• These days, every company wants greater innovation, but they’re unsure how to achieve it.
• Here at Simcenter, we’ll show how to get more value out of their modeling & simulation investment and drive innovation at the same time!

 
Driving greater innovation is a mantra at most global companies these days. Boston Consulting Group conducted a study that showed that over 83% of Fortune 500 CEOs list ’Driving Greater Innovation’ as one of their top three priorities, but very few have any well established corporate strategy for innovation.  In other words, they all want greater innovation, but they’re unsure how to get it.
 
At Siemens, we’ve set our sights on helping engineers make better use of their modeling & simulation efforts to drive real product innovation.  I’ll share the details with you today.

The New Paradigm for Design Exploration Using HEEDS Software

New HEEDS process for design exploration

In the latest release of HEEDS Software we’ve removed the previous bottlenecks and barriers:

• No need to simplify model (can efficiently find good designs with a large number of variables)
• No need to select algorithms (HEEDS automates this)
• No need to use just one search strategy (HEEDS leverages all strategies simultaneously)
• No need to tune algorithms (HEEDS automatically tunes throughout design study)
• No need for iteration (HEEDS finds the best way to traverse the design space each time)
• No errors introduced (HEEDS doesn’t rely on any response fitting)
• No need for optimization expertise (it’s built into HEEDS)

Right from the outset, we developed HEEDS with the objective to overcome these issues with the traditional approach. We started by focusing on the complaint that the traditional approach was “too difficult.” So, we built a search framework that would analyze the problem, choose the best strategy, and automatically tune the strategy.  This made it much simpler for the user.

Then, we confronted the complaint that the search was “too costly.”  So, we modified our framework so that it would combine different search strategies simultaneously throughout the design study to quickly find better-performing regions in a fraction of the time required by traditional single algorithms and approaches.

Next, we focused on taking maximum advantage of parallelization and efficient licensing schemes so that we could search the original CAE model directly and more efficiently than a DOE/RSM approach.  So we were able to completely eliminate the error-inducing issues of the traditional approach.  Lastly, we automated the whole process in a product named HEEDS that helps engineers to discover better designs, faster.   

Traditional Process For Design Exploration

 

 
The diagram above Illustrates the traditional process for design exploration. The traditional process leads to limited usage (occasional use, only by experts, only on toughest problems)
 
At this point, you’re probably thinking that “yes, that looks great, but I’ve seen design exploration tools before, what makes this so different and better?”
You’re certainly right that there have been previous attempts to provide design exploration software.  Most of these originated more than 20 years ago and fell far short of the goal, for a variety of reasons.
Although they leveraged the best technology available at the time, they suffered from three fundamental limiting factors:
1. CAE simulation models weren’t very robust and often didn’t rebuild reliably around the baseline
2. Compute resources were fairly limited, and
3. Design search strategies were too inefficient and required more evaluations than was practical

Because of these limitations, the traditional design exploration illustrated here was established.
Because search techniques weren’t robust enough to effectively search a full CAE model, the user was asked to simplify the previously validated CAE model either by screening variables to see which ones were most influential around the baseline design, or by sampling the design space with a Design-of-Experiments (DOE) approach and then fitting the sampled points with a smooth mathematical function, called a response surface (this is the ‘so-called’ response surface method or RSM).

After that simplification, the user was confronted with a dizzying array of search algorithms (e.g. gradient search, genetic algorithms, particle swarm, etc.) and asked to choose one that was most appropriate to the problem at hand. Next, once an algorithm was chosen, the user was asked to provide tuning parameters specific to the algorithm (e.g., things like mutation rates and cross-over points for genetic algorithms. These terms are unfamiliar to most simulation engineers.).
If a user could get through those significant hurdles, they attempted to search the simplified design space with their chosen algorithm to find a better design.
But since the simulation model had been simplified, the user had to check the solution on the original model and then refine various steps of the process.
The problems encountered with this approach are that it introduced too much error, required too much iteration, and demanded far too much expertise for most simulation engineers.
The reaction was often that this approach was too difficult, too costly, and too error-prone.
The result was that this approach became relegated to use only by experts on the toughest problems and, even then, the expected gains were usually very limited.