Intuition plays a critical role in all stages of a design exploration study, from defining the problem statement to building the simulation model to interpreting the results. But what about the search process itself? Should we make design improvements based on intuition, or should we allow a mathematical search engine to explore the design space for better designs? The answer is both. We call this shared process collaborative design exploration.
The SHERPA search strategy allows you to inject your design ideas before and during an exploration study. Before you start a study, you can seed it with multiple ideas (in the form of actual designs) that might help SHERPA to locate productive regions of the design space more quickly, thus speeding up the overall search. For example, in addition to the baseline design, you might consider seeding the study with other potentially good designs that:
- you have investigated or produced in the past
- your competitors have used
- are feasible, but perhaps not optimal
- are high performing relative to one or more criteria, but not all of them
- have some desirable features, but don’t necessarily perform well
- you have a hunch may work well
- are from a previous HEEDS MDO study
One or more of these injected ideas might contribute to a more efficient search, while the cost of doing this is only the time to enter the variable values that define each of the designs. SHERPA will evaluate the injected designs when the search process is launched, so there is no need to simulate them before injection.
Will the search process suffer if the designs you suggest do not have a high performance? Not usually. You see, if you don’t seed the search with any injected designs, then the first set of designs evaluated by SHERPA are quasi-random ones selected from a type of Latin hyper-cube sampling procedure. Many of these designs may not be very good, so your suggestions are at least as good as random designs, and potentially much better.
The only potential disadvantage is losing some diversity in the initial set of designs. To alleviate this, the number of designs injected prior to the study should be kept relatively small, say less than five or ten percent of the total number of evaluations, and the injected designs should not be too similar unless you feel very confident about their performance.
An even more interesting approach is to inject designs during an exploration study. As you know, it is recommend that you monitor the progress of your studies as they run. While investigating an intermediate design or a trend in the search, your intuition will often be triggered. You may have new ideas for how to improve one or more of these intermediate designs, or how to create a new design concept that SHERPA has not found yet. These ideas can be shared with SHERPA through the same injection process described above.
As before, if your ideas do not lead to improved designs, you have not impaired the search progress in any significant way. Broad exploration is part of the SHERPA strategy, so your injected designs fit well within this approach. The risk is low, and the potential gains are high.
Given a sufficient exploration budget, HEEDS will be able to identify better performing designs in the design space on its own. But, by coupling its specialized search capabilities with your fundamental domain knowledge, high performing regions can be explored more quickly and effectively, leading to truly innovative breakthroughs.
This collaborative design exploration process couples and leverages highly developed human intuition with advanced mathematical and computational capabilities. This process is powerful, simple and … intuitive.
We hope these tips helps you to discover better designs, faster.