Jump to content

Design Explo­ration and Opti­miza­tion: Is Your CAD Tool Really up to the Task?

designexplorationblogpost

If you have been browsing the infor­ma­tion on our website, you might have seen that our tool CAESES® brings along several dif­fer­ent com­po­nents, such as a CAD modeling envi­ron­ment, process automa­tion, design explo­ration and opti­miza­tion methods. Out of these it is, however, really the (very spe­cial­ized) CAD modeling that is our main focus and strong point in the context of sim­u­la­tion-driven design. You, as well as several of your fellow engi­neers, might be won­der­ing, do I really need an addi­tional CAD tool?”. Obvi­ously, we think you do and here is why. But let’s start at the beginning… 

CFD Becoming a Design Tool

Due to the rapid devel­op­ment in computer tech­nol­ogy over the last decades, both in terms of com­put­ing power and afford­abil­ity, the use of CFD has increased sig­nif­i­cantly. Not only is it being utilized to a much larger extent, but also earlier and earlier in the product devel­op­ment process. As opposed to using CFD late in the design process, where it can merely serve for val­i­dat­ing a com­pleted design or give some guidance for late changes, employ­ing it early in the process turns it into a real design tool. It can be a valuable help in quickly gaining knowl­edge about the product’s behavior and, under strict con­sid­er­a­tion of the product’s per­for­mance, guide the design in the right direc­tion from the start, when critical deci­sions are taken. Apart from the avail­abil­ity of com­put­ing resources, another factor that has greatly con­tributed to the increased use of CFD is the improve­ment in ease-of-use. While it pre­vi­ously was a tool that was mostly limited to being applied by sim­u­la­tion spe­cial­ists, it is now avail­able to design­ers and engi­neers at large. Espe­cially the advances made in the automa­tion and robust­ness of the meshing process have fueled this development. 

Process Automa­tion and Optimization

As a next logical step, applying CFD early in the process and con­sid­er­ing its results to guide the devel­op­ment of the product, quickly leads to the wish of employ­ing it within an auto­mated process for the sys­tem­atic analysis of design variants – design explo­ration – and opti­miza­tion. Shape opti­miza­tion of com­po­nents and systems is becoming a standard pro­ce­dure in many indus­tries that deal with flow-related geome­tries, such as the auto­mo­tive, maritime, aero­space and power gen­er­a­tion indus­tries. A common denom­i­na­tor in all these very dif­fer­ent products is that small con­certed changes in shape often lead to sub­stan­tial improve­ments in per­for­mance. Fur­ther­more, even small improve­ments often yield impor­tant benefits for the producer, the consumer and the envi­ron­ment, e.g., when reducing energy con­sump­tion and emissions.

 Automate CFD analysis for all geometry variants

Apart from the pure CFD tools that, as pre­vi­ously men­tioned, are becoming increas­ingly suited for process automa­tion, addi­tional com­ple­men­tary CAE tools are needed, which take care of other tasks required for the complete automa­tion of the process: control of the opti­miza­tion process, includ­ing variant and data man­age­ment, and gen­er­a­tion of the geometry variants that should be eval­u­ated, pro­vid­ing the input for the sim­u­la­tions. While there are quite a few generic opti­miza­tion tools on the market that can fulfill the former role, the latter – i.e., ded­i­cated tools for shape vari­a­tion – seems to be a rather sparsely pop­u­lated niche. Tra­di­tional CAD systems often do not fulfill the given require­ments, well, at least when dealing with the complex, compound-cur­va­ture shapes that char­ac­ter­ize many flow-related geome­tries, and a spe­cial­ized CAD approach is called for instead. 

The Bot­tle­necks with Tra­di­tional CAD Tools

Tra­di­tional CAD packages are surely very powerful systems that accom­pany the complete design process and fulfill many dif­fer­ent tasks, e.g. con­struct­ing pro­duc­tion-level geometry models, creating complex assem­blies, pro­duc­ing BOMs and man­u­fac­tur­ing drawings, as well as managing PLM data. They are, however, often detail-centric, meaning that the geometry models include many details that are relevant for the final product, but not nec­es­sar­ily for the sim­u­la­tion. When pre-pro­cess­ing the models for the CFD grid gen­er­a­tion, they have to be de-featured, the wetted” surfaces have to be extracted and the geometry has to be cleaned up (e.g., making it water­tight). Also, these tools are often not pre­dis­posed for quick vari­a­tion of complex geometry. Changing prop­er­ties of the shape can involve a lot of manual effort and is not nec­es­sar­ily robust, leading to many infea­si­ble variants when attempts are made to automate the variation.

 The automated regeneration of complex geometries is often a challenge for traditional CAD tools

Another aspect that char­ac­ter­izes tra­di­tional CAD is the typical user group. Usually, a ded­i­cated CAD depart­ment will be in charge of oper­at­ing the CAD system, handing over geome­tries to the CFD depart­ment. When the people in the CFD depart­ment want to try vari­a­tions of the geometry or suggest some changes, they have to request a new geometry from the CAD depart­ments, which often leads to delays and an inef­fi­cient process. Obvi­ously, process automa­tion is hardly possible in this set-up.

 Efficient automated solutions are required to robustly create and also to assess large sets of design candidates

Design Explo­ration and Opti­miza­tion: The Case for a Spe­cial­ized CAD for CFD

A spe­cial­ized CAD like CAESES®, on the other hand, is a surface modeler that focuses on the modeling of the CFD-relevant geometry only, pro­vid­ing it in a state that can directly be used for sim­u­la­tion. It has a strong focus on geometry vari­a­tion, so that, once a model has been para­me­ter­ized accord­ingly, it can robustly provide geometry variants just by simply changing the model para­me­ters, be it in a manual or an auto­mated process.

Parametric model of a turbine wheel including the periodic CFD flow domain

The require­ments ful­filled by such a CAD tool for CFD can be sum­ma­rized as follows:

  • Geome­tries should be defined and con­trolled by as few para­me­ters as possible, thus reducing the degrees-of-freedom. The opti­miza­tion effort scales with the number of free vari­ables (often in a qua­dratic fashion); there­fore, a low number is highly desirable.
  • It should be easy and fast to vary the geometry by con­trol­ling the values of pre­vi­ously defined para­me­ters. These para­me­ters should be inde­pen­dent from each other and it should not be required to change multiple para­me­ters in a con­certed way, just to obtain a specific change in shape.
  • Geometry gen­er­a­tion should be robust with a minimal amount of failed variants.
  • The gen­er­ated geometry should be provided in a state and format that can directly be used for the sim­u­la­tion tool involved in the specific process.
  • The system should have the ability to manage con­straints and even build them right into the model, so that the creation of infea­si­ble variants is pre­vented or at least minimized.

In contrast to a tra­di­tional CAD package, this is meant as a tool for the CFD engi­neers, giving them the ability to generate their own geome­tries and variants, even at a stage where the design is not pro­gressed so far that the CAD depart­ment can actually hand over geometry files.

Parametric stator model including periodic CFD domain

What about Morphing Tools?

A typical alter­na­tive to using a CAD package in the design explo­ration or opti­miza­tion process is a morphing tool. Such tools usually act in the CFD domain by deform­ing the com­pu­ta­tional mesh. While this does bring along some advan­tages related to removing the neces­sity for remesh­ing, it often creates problems down­stream. The opti­mized geometry, that is now only existing in the CFD domain, has to be brought back to CAD, which usually means remod­el­ing it, trying to stay as close to the result as possible. This takes sub­stan­tial effort and often results in slight devi­a­tions from the opti­mized shape, with possible negative impact on the performance.

 RBF morphing in CAESES: Deforming the back of the car geometry that is given e.g. as STEP/IGES file 

CAESES® also brings along some powerful morphing capa­bil­i­ties to work with existing geome­tries, without having to build a new para­met­ric model from scratch. The dif­fer­ence to most other morphing tools on the market is that it stays in the CAD domain, applying the defor­ma­tions to the original surface geometry. After the opti­miza­tion, this geometry can easily be exported in a standard geometry format and brought back into the pro­duc­tion CAD system for further development. 

Bot­tle­necks with Your Tra­di­tional CAD System?

If you have encoun­tered one or more of the bot­tle­necks men­tioned before, when using your tra­di­tional CAD package for design explo­ration and opti­miza­tion, get in touch with us and find out if CAESES® is the right solution for you. 

Contact Us

More articles

Latest from the blog

All articles

Stay up to date

Receive latest news to your inbox.

Subscribe to newsletter