You are on page 1of 9

ANSYS Tutorial: Design Optimization in DX

ANSYS Workbench Tutorial: Want to know more about you model? Run a What if scenario? How sensitive are you results for changes in the input variables? Maybe optimize your model in terms of weight? Then DesignXplorer is the tool for you. This article will first give you a short summation of what DesignXplorer is and then show how to set up a simple optimization.

What is DesignXplorer?
Design Explorer is based on a method called Design of Experiments (DOE). This together with various optimization methods helps the program to develop an optimized structure based on selected input and output parameters. Input parameters can either come from DesignModeler or from various CAD systems. These parameters can be in terms of thickness, length, etc. They can also come from Mechanical in terms of forces, material properties, etc. The output parameters are calculated in Mechanical and can for example be in terms of total mass, stress or frequency response. After setting up an analysis with a number of input parameters and out parameters there are the steps that can be run within DesignXplorer:

Design of Experiments Response surface Optimization Six Sigma Analysis

Design of Experiments
Design of experiments is the foundation that everything within DesignXplorer is built on. What this technique is about is to determine how many and for what input values the analysis shall be run. There are various techniques for this but the same goal for all is to get as good response surface as possible with as few input combinations as possible. So basically this step defines is how many analysis that will be run. Each combination that ANSYS solves for is referred to as a Design Points.

Response Surface
When the Design of Experiments is run the next step is to create a response surface based on these results. A response surface will be created for each output parameter. The response surface is basically created via curve fit through the Design Points. From this response surface you can then investigate output results for input variable combinations that hasnt been solved for.

Goal Driven Optimization


To help you to select the combination of input variables that satisfies your goals best, you can run a Goal Driven Optimization. Here you have the possibility to give all you parameters different objective functions on which you also can give different importance. ANSYS will then give you a number of candidates that satisfies your goals in the best way.

Six Sigma Analysis


The Six Sigma Analysis is a method that includes both a Design of Experiment and a Response Analysis. What resides is that you also have the possibility to set the input parameters as uncertainties parameters. This means that you can see how uncertainties in the input variables will affect the results for the output parameters.

How to r an O H run Optimiz zation in Desig n gnXplor rer.


In the tutorial below it wi be shown how to set up an optim t ill n t mization an nalysis in De esignXplore We will use er. two input vari o iables, a thic ckness and a radius, an two outpu variables, total mass and Princip Stress. T nd ut s pal The goa is to find the best com al mbination b between thickness and radius to minimize the mass of the structure a m e and to ensure that the stress w be below a defined value. e will w

Ste 1: Para ep ametric model m


The first step i to create a parametri model. In this case A is ic n ANSYS Des signModeler has been u used, but there are a number o different CAD syste that also can be use e of ems o ed.

Wi ithin Design nModeler th paramete are simply defined b checking the box be he ers by g etween the d desired par rameter. In this case we have two parameters. A thicknes and a rad e . ss dius.

Ste 2: Set u the ana ep up alysis to b run. be


The next step is to transfe the param er metric mode from Desi el ignModeler to Mechan r nical and set up the t ana alysis. In this case we w run a st will tatic analysi which is s is simply drag gged and dro opped onto the defined d geo ometry. Not that there now is a P te e Parameter Se associated with the g et geometry an the Static Structural nd c pro oject.

No you will have to def your materials, loa boundar condition mesh set ow fine ads, ry ns, ttings etc. a usual in as Me echanical. W resides from an or What s rdinary analysis is that you also w have to define you o t will d output par rameters. In this case w will use t n we total mass and Maximu Principa Stress as o a um al output parameters. Fir define loads, rst

the define ou en utput parame eters.

Ste 3: Add a Goal D ep d Driven Opt timization to you project. n


The next step is to add a G Driven Optimizat Goal n tion to the c current project. To do this, simply drag it from e h m the toolbox an drop it on the curre analysis e nd nto ent s.

On could also check the parameters set to see t all your parameters are proper defined. ne o s that r s rly

Ste 4: Defi Design of Exper ep ine n riments.


To define Des sign of Expe eriments, do ouble click Design of E D Experiments in the Proj Schema s ject atic. The go oal re up priate numb of analy to get as accurate re ber ysis s esponse sur rface as possible. (And of her is to set u an approp cou urse, the few paramet combina wer ter ations the be etter when i comes to calculation time.) For each input it par rameter you will have to define bo u t oundaries in between w n which they a allowed to vary. In t case the are this e rad are allo dius owed to vary between 2 and 3 mm. The thickn is allow to vary between 8 and 10 mm y ness wed m. The picture be elow shows how the bo oundaries ar defined fo the radiu together with the sele re for us w ected design n poi to be ru ints un.

Ste 5: Run the Desig of Expe ep n gn eriments.


This part of th analysis is usually th most time consuming part. Whe pressing the button U he i he e en Update Des sign of Experiment ANSYS w sequent ts will tially run an analysis fo each sugg n or gested desig point. In this case, gn n here s, s 0 un. ave put wh we have two input parameters ANSYS suggested 10 design points to be ru If we ha more inp par rameters we need more design poi to be ab to create an accurate response s e e ints ble e e surface whi will lead to ich d eve more ana en alyses to be run. If that is the case, ANSYS lu , uckily has a solution to speed up th process. If his you have a com u mputer or cluster with several CPU available and you combine tha with a pa c Us c at arallel licens se for ANSYS yo can run t analyses in parallel instead. Th will dras r ou the s l his stically spee up the ca ed alculation tim me com mpared to d doing a sequ uential run. The figure bel shows t results o the Desig of Experi low the of gn iments after all design points have been run. r p e

Ste 6: Calc ep culate Res sponse Sur rface.


Wh we have a solution for all desi points we can calcu hen e n ign w ulate a respo onse surface As mentio e. oned before e this is a curve fit through the results for each design point s this is a q so quite fast op peration.

Besides the ca apability to p the resp plot ponse surfa itself, there are here a lot of cap ace e pabilities to plot the o nnection bet tween input and output variables. Among oth things w have the p t t her we possibility t plot how to con sen nsitive outpu results ar based on changes in the input va ut re ariables.

Ste 7: Run Optimiza ep n ation.


No when we have a resp ow e ponse surfa we can u this to fi an optim design. When you opened ace use find mal Op ptimization y have th possibilit to set diff you he ty ferent objec ctives for yo paramete and also the possibil ou ers lity to weight thos in regard to importan In this case we hav an object for the Principal St w se nce. ve tive tress and th he Ge eometry Mass. For the s stress weve set a criter where w want to have a struct e ria we h ture that don experien nt nce any stresses ab y bove a targe value of 1 MPa. T is set to high impor et 180 This o rtance. For the mass weve set an obj jective to m minimize the mass with importance low. The i e e input variab are left with no obj bles jective since e the we are o ere only interest in that th are insid the boun ted hey de ndaries that we defined from the st tart. Any va alue in between tho boundar are ok. b ose ries

Wh weve d hen defined obje ectives it is j to run the optimiza just t ation and ra the different candid ate dates that AN NSYS gives us. In this case ANSY gives us the followin candidat to choose between. s YS ng tes

No we can s ow select the ca andidate tha suites us th best. Bas on the r at he sed rates (stars) given by A ANSYS, is sea like Can ams ndidate B is the best so s olution for us. When se u elected the appropriate candidate it could alwa a t ays be a good prac ctise to inse this as a d ert design poin and run th analysis again to ma sure that there is no nt he a ake t o bstanially di ifference be etween the r solution and that ta real n aken from th response surface. he e sub The result from the final run can be s m r seen in the f figure below w.