Accelerating the search through millions of “what if” scenarios

In the climactic moments of a movie I was watching the other day, the Sorcerer Supreme explored more than 14 million possible futures in his quest to save earth. His goal: to find the single timeline where the superheroes emerged victorious against that mad titan.

This monumental task, a “what-if” scenario exploration, is a big challenge that is also faced daily by engineers designing complex electronic systems. When designing a space, there are plenty of possible options of what a design could be, with every parameter option accounted for. This means all the possible combinations one may run to explore a trace’s width or length, via spacing, pad sizes, etc. which would potentially lead to millions of combinations that we simply don’t have the luxury to explore. The methods of magic and sorcery they used in the movie, while powerful, were essentially a sequential, manual observation of possible futures. What if that sorcerer had access to the intelligent optimization and parallel processing capabilities of HyperLynx Design Space Exploration?

What is Design Space Exploration (DSE)

It’s the process of finding the optimal combination of variable values that best meet your design goals. To do so, the user defines:

• The range of design variables to be explored (design space)
• The output metrics used to compare design behavior (responses)
• the values for those responses that define design success or failure (design targets)
• the goal for each response, whether to maximize/minimize its value (design goals)

HyperLynx DSE enters the chat:

HyperLynx Design Space Exploration (DSE) is designed to automate the process of exploring vast design parameter spaces and identifying optimal solutions. It accelerates design optimization by learning from past simulation results and using that knowledge to guide future exploration through HEEDS optimization software. HEEDS is an exploration engine, which allows designers to automatically and concurrently explore hundreds of design parameters and their relationships in a surrogate-model-based optimization and machine learning principles.

Rather than treating each simulation as an isolated evaluation, HEEDS optimization software builds a mathematical model (surrogate) of how the design responds across the parameter space.  This mathematical approximation is called the Response Surface Model (Fig 1) that accelerates the process in the following way:

• Tries a few combinations first
• Learns the relationship between design variables and electrical behavior and builds a multidimensional model based on that.
• Adjusts the model based on the results from the previous study
• Predict what the model looks like and runs the next study.
• Repeat until it converges on the best results.

The SHERPA algorithm intelligently guide this search. It explores the full design space to avoid getting trapped in a “local minimum or local maximum” while creating that model, so the whole space is observed. It doesn’t necessarily run every single permutation because it can learn from initial runs and prune unpromising paths. This is like learning that “letting the mad titan kill one of the superheroes” is always a bad future and thus deprioritizing similar scenarios.

Even engineers with very little design optimization experience can use HyperLynx DSE to discover optimal designs in a fraction of the time it would take to perform a handful of manual iterations. This efficiency can save days or even weeks of CPU time during common optimization studies.

Finding that one reality

While the Sorcerer Supreme in the movie looked at these possible realities, he still had to filter them through sequential observation.

Instead of simulating 14 million scenarios, HyperLynx DSE can intelligently zero in on regions of the design space that are most likely to yield optimal results, drastically reducing the total simulation time and computational resources. As more it learns how the variables affect the results, better points are chosen to converge on a solution faster.

With AI/ML Integration, engineers can enjoy similar guidance and wisdom like what we see in movies, learning from past “futures” (simulations) to predict promising design paths and further accelerate the exploration, making the process even more intuitive and efficient.

In Summary, HyperLynx DSE can help you optimize your design’s performance by exploring a very large design space in a fraction of the time required with traditional methods. Not only does this save time and effort; it provides a substantially more optimized design than is possible without these advanced methods.

Interested in learning more about Hyperlynx? HyperLynx workshops get you up and running quickly by documenting complete workflows with detailed step by step instructions and videos. HyperLynx workshops guide you through the process with working examples so that you can hit the ground running on your own design simulations.