Best Designs and Pareto Frontier

As some of our CAESES® users may already know, we really love to squeeze in small pieces of useful func­tion­al­ity into our main­te­nance releases. Why wait for a major release if our users can benefit from them right now? This time, we added the visu­al­iza­tion and the high­light­ing of the best design can­di­dates for version 4.1.2 (to be released in autumn 2016). Typ­i­cally, our pro users run auto­mated design studies and formal shape opti­miza­tions. Of course, they want to know which of the gen­er­ated designs are the best can­di­dates. If you cope with several objec­tive func­tions, the so-called pareto frontier is also of interest (showing the best designs for con­flict­ing objec­tives). So far, our users could check the design results table and sort it accord­ingly by clicking on the header title. Well, it was ok, at least for the sit­u­a­tion where you have only one objec­tive. But this was not suf­fi­cient for multi-objec­tive opti­miza­tions, and we’ve had ideas for improve­ments on our list pretty much since CAESES® has existed. So finally we spent one sunny day in July 2016 on the tech­ni­cal imple­men­ta­tion and some new icon designs. Inter­ested in the outcome? Here you go: 

Object Tree

After a com­pleted run, the best designs receive a new icon. With this, you can imme­di­ately detect inter­est­ing and relevant designs, to analyze them or to use them e.g. as basis for another local optimization. 

Design Results Table

In the design results table, the best designs receive the same new icon as well. Note that we have sim­pli­fied the icon for valid designs — it lost the check mark, which gives you a cleaner overview. 

Charts

Now the pareto designs have their own distinct blue color in the charts. This allows you to create pareto-plots auto­mat­i­cally without any addi­tional user inter­ac­tion. Note that the valid designs are colored green, and the invalid ones (vio­lat­ing a con­straint) are colored red. 

Design Studies

The func­tion­al­ity described above is also avail­able for Design of Exper­i­ments (Sobol, Latin Hyper­cube, …) and simple para­me­ter studies. Just set and activate the objec­tives for the chosen design engine. CAESES® takes this infor­ma­tion into account and shows you the best designs of your study right after the run. Until now, users haven’t been able to set objec­tives for para­me­ter studies. 

More Infor­ma­tion

Thanks to our users who haven’t stopped pushing this topic again and again. We hope you’ll like this first version, any feedback or ideas on how to improve this are highly welcome.