ME5634 Article-Avoiding Common FEA Mistakes.pdf

ME5634 Article-Avoiding Common FEA Mistakes.pdf - i a if I...

Info icon This preview shows pages 1–3. Sign up to view the full content.

View Full Document Right Arrow Icon
Image of page 1

Info icon This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document Right Arrow Icon
Image of page 2
Image of page 3
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: i a if I; i i l i i i a if: 2 ii i g h j LE 1 a ii ‘23 K i \ Avoiding common FEA mistakes Vince Adams CAD Geometry Figurel. SPRlNG Partial Detaii - Qtt. Symmetry Full Nonlinear Solid Model — No Contact Run Time - ~6 hours l998 here is always a beginning. Everyone involved in Finite Element Analysis (FEA) should remember their first real project, their first real mesh, their first presentation of results. “Real” should be defined as post—training. In training, all the problems build correctly and the answers are always as expected. That’s how most training classes are developed. Eventually, the rookie analysts must sit down at their work— station, fire up the software and start modeling without the comfort of an all—knowing instructor. It is in these early stages that novices will make the most mistakes and experience the most confusion. The opportunity to learn from mistakes is valuable but you must first recognize that a mistake has been made. In FEA, many significant mistakes can easily go unnoticed by the , untrained eye. This is the greatest danger facing the growing number of design analysts who are likely the only person in their company using finite element technology. FEA data is often used to drive design deci— sions. Consequently, incorrect results will cost your company money. lf‘you’re lucky, mistakes will be caught in the prototype stage and the only loss will be in time and fabrication costs The other end of the spec— trum could mean compromising consumer safety. Many organizations depend on accurate FEA results as the basis for safety-related decisions in structures that are too complex or costly to prototype. As the analyst, you must understand your tools and minimize the chances of error. The following observations have been made over the course of several years working with new FEA users in various industries It should not be considered a complete listing of the problems new users will encounter; those are as numerous as the students themselves. Additionally, this cannot replace one-onfone interaction with a mentor or coach who has sufficient experience in both using and teaching the technology. In fact, if you recog- nize your own'habits in this article or aren’t clear on some of these con— ' cepts, it is important that you L seek guidance from an expert. ‘Upon completing their first L solo model of a part or system, most new users get excited and go through Several cartridges on their color printer While documenting their success. This is a great reaction; the best users tend to be passion— ate abOut the technology and think that it’s- fun. HOWever, this excitement often clouds the cold objectivity required totruly qualify an analysis Plane Stress w! Contact Full Nonlinear Model Run Time - ~10 minutes ANALYSIS SOLUTIONS model and the corresponding results. To provide a base for improvement and growrh, a new user must overcome three key issues: 1. Poor initial implementation factors. 2. The tendency to model too much complexity. 3. Lack of understanding of the limitations of the technology. initial impiementation Unfortunately, many new users haven’t been properly prepared to evaluate the quality of FEA data. Most rookie analysts have only vendor training in their software tool of choice to fall back on. In general, however, most vendor training programs are problem—free: the models all mesh per— fectly, they solve quickly and they have mesh sizes and construction that have been pre-converged to match target results. lnevitably, there are shortcomings that require workarounds. These courses usually do not properly prepare students for the real world they must face on their own. Experience and constructive criticism by peers and mentors are the best means to move beyond the initial training. As stated earlier, most new users are indi— vidual islands of technology within their companies. Consequently, they don’t have the benefit of a mentor or peer to review their methods or models. This review is critical to ongoingg arowth. It 15 the best way to ensure that bad habits don t propagate. Also, it pro— vides an opportunity to learn more accurate or efficient techniques for fiitui‘e studies. A workshop attendee from Harley—Davidson Motorcycles once commented that experi- enced FEA users will readily admit they have a lot to learn, whereas users who think they know enough are probably in trouble. A last barrier to success facing many new users is their company’s choice of software. Design engineers often grasp for that most tempting offer, the CAD—embedded FEA system. ln their zeal to buy, many users pur— chase a tot—only tool. A case can be made that these tools provide a preview of a part or system’s performance and aren’t intended for serious analysis. Your company’s needs should be reviewed with an expert user who has no stake in the decision to ensure that the best solution is implemented. Too much complexity Another problem a new user will encounter is the desire to model too much complexity too quickly. This is a natural consequence of a few common characteris— tics: over—enthusiasm, poorly understood problems and over—reliance on CAD models. Over—enthusiasm—First of all, most new users are very enthusiastic, even anxious, to start justifying the technology. What’s more impressive, result contours on a single part or on an entire system, complete with the corporate logo? It is not unrealistic to expect that the sources of error increase exponentially with the complexity of the ' model. For single—part studies, correcting your approach for problem understanding and over—reliance on CAD, as described below, should help considerably. For assembly studies, we recommend an approach called component contribution analysis. This approach involves breaking a system into its components and developing boundary conditions for each individual analysis, which should represent that com— ponent’s interaction with the rest. This forces the user to consider the contribution that this part makes to the overall system response and provides a baseline for evaluat- ing the quality of the final system model. While it may take longer and the compo— nent results aren’t nearly as dramatic, the eventual outcome will be solid data with a minimum of unaccounted—for error. Poorly understood problems—~Multiplying the effects of this over—enthusiasm is the ten— dency towards starting with an incomplete ‘ understanding of the problem and the desired output. it is common for new users to over~rely on their new tool to provide answers to poorly defined questions. The assumptions and inaccuracies that are inher— ent in the technology make it even more crit- ical that the analyst understand the impact of every joint, load, constraint or property so that the results can be qualified correctly. FEA results can’t be taken at face value. They must be interpreted in light of the ana— lyst’s understanding of the material properties, manufacturing process, analytical assumptions and end use of the part or system. It is good practice to outline the assumptions and their expected effect on paper and review them with a peer or mentor. Even a non—user should be able to assess the effects of poorly understood assump~ tions and might even point out issues that might be overlooked by a regular user. “Measure Twice, Cut Once” is applicable here. Build models with only as much com— plexity as you are able to understand and qualify. Resist the attempt to model any part or system that you can’t visualize or predict ballpark results for. Over-reliance on CAD models—The ability to generate a good tetrahedral automesh on a CAD solid must be recog— nized as a significant advancement in the usability of FEA. This has made it accessi— ble to many more users, problems and geometries than ever before. Most analysts will admit that, when used correctly with appropriate error control, a second—order tet mesh can generate good results. However, this advancement may also facilitate mindless meshing. When it was more difficult to mesh 3. part and solution time was a serious issue, analysts took the time to ensure that an efficient mesh was built with the best elements for the task. They also defeatured their CAD model more carefully. It is not always enough, or always correct, to remove all fillets and rounds. Identify the load path and try to identify any sub—structures that might be broken off for a design component analysis. Always look for symmetry, axisymmetry or planar approximations. These might be derived from your CAD model. Figure 1 shows where an automeshed solid is clearly the wrong approach. wliar:lanology limitations It is important to always keep in mind that an PEA model is an approximation of a real system. A model’s ability to simulate real—world behavior is a direct function of the error introduced by the assumptions required to complete the analysis. These assumptions include material properties, geometry, element selection and density, code accuracy and boundary conditions. Understanding the effect of these assump- tions allows you to understand the differ— ence betvveenuvalid simulation results and mere pretty pictures. ’ teady and unsteady 3D fluid flow program LINFLOW with an pdule solves the unsteady fluid dynamics in the frequency ialysis {such as flutter predietions) “ Cover the surface of your structure with boundary elements and solve your 1 LINFLOW Rev 1. 0 IS available on UNIX Workstations and Intel PC(NT/4. CI) . i 'Ord now and get a limited version of LINFLOW'called LINFLOW- 1 Ed alian’al for only $99, excl shipping.LINFLOW- -Educati’onal may be Oscillating loudspeaker in a Steady Flow Field Courtesy of NOKIA Research ,3 used as a separate program or as a mod 6 in ANSYS ANSYS/ED (new ANSY commands and menus at included). For more information send or fax this coupon to: ANKER - ZEMER Engineering AB PO BOX 156, S — 691 23 KARLSKOGA, SWEDEN Phone: +46 — 586 - 52820 Fax: +46 — 586 — 56470, E—mail: [email protected] Web: —zemer.com Or contact your local ASD (AN SYS Support Distributor). Company: Name: Address: Phone: Fax: 'l'mdemarks are owned by their respective vendors. ANALYSIS SOLUTIONS SPRlNG I998 A better—looking mesh is usually a better mesh. This statement refers to the fact that a mesh with better shaped triangles and quadrilaterals, a gradual transition between different densities and better adherence to underlying geometry will be more aestheti— cally pleasing. More importantly, it will always produce more accurate results. While one may argue the benefits of improved aesthetics, this correlation makes it much easier to evaluate the quality of a mesh before the solution is even performed. How does this apply to over—reliance on CAD models? It is the tendency of new users to assume that any mesh is a good mesh and that as long as their geometry is correct, the analysis will be correct. Fz‘gureZ shows two meshes of the same part. The coarse mesh, using system-selected defaults, is unlikely to produce accurate stress data. The fine mesh requires more thought by the analyst but will provide significantly better results. Unfortunately, most PEA solvers don’t even understand the concept of geometry as defined in a CAD system. Therefore, users must focus on the resultant mesh and the three qualities ofa good mesh stated above before submitting a final model to the solver. Even the latest releases of P—ele— ment—based tools are not immune to inac— curacies based on poor mesh quality. Let CAD drive the mesh, but evaluate the mesh on} its own merits. Figure 3 shows one means of checking the quality of a mesh. Local error estimates highlight mesh regions that require refinement or must be interpreted in light of the error estimates . rm . a no e LAALLALL iur usually more accurate. it is the tendency of new users to want to tet mesh every- . Element Error Pint High Error Levels At This Stress Riser Indicate That More Mesh Refinement is Required " SPRlNG i908 Von Mises Stress Contours Figure 3. Checking the quality of a mes/7. thing. This is partly because it takes work, sometimes days ofwork, to build a good mid— plane shell model ofa thin—walled part. Most of the time, the CAD solid model is of little help in this task. However, thin’walled parts, especially in bend— ing, will be modeled much, more accurately with a shell representation than a solid mesh for two reasons. Solids require several elements through a thickness to accurately capture bending, and most new users tend to undermesh anyway. Take the time to model shells where shells are best used and beams where beams are best used. The results will be more accurate and, oftentimes, the model will be more flexible. Additionally, when results count, use see (ind—order tetrahedrons or elements with mid~side nodes. These provide a more accu— rate representation of displacements over a similar mesh of linear tets. Even higher order P—elements should be used, if avail— able, to capture higher stress gradients. Boundary conditions are often the hard— est assumption to quantify and the most casually applied input to an PEA model. We have developed a workshop dedicated to the practical application of boundary conditions because they are so important, yet so poorly understood. Boundary condi- tions, in simple terms, represent any part ‘ "n9 "nlifiirl" Ar Avfnrlunl in “on” ma m a U1 yALuLAAaL lnL r.) JL\IL LApuuuy uriuuilpfi modeled. Consequently, it is fair to assume that applied boundary conditions should allow or restrict deformation of the model as the parts they are repre— senting would. This seems intuitive but I frequently need to ask stu- dents the questions, “Can it really bend like that?” and “is it really that rigid?” One of the most basic tests of a boundary condition scheme is to review an ani— mation of the deformations and question the validity of the displacements. ANALYSIS SOLUTIONS unarse Mash Using catsuit size Fina Mesh Using Local Refinement Figure 2. Two meshes oftlje meme part. Other boundary condition mistakes ' made by new users include over—constrain— ing models, incorrect load balance and rash assumptions that should have been bracketv ed or proven out with test models. Corrections for these errors are problem dependent and must be addressed on a case—by—case basis in a model review with peers or a coach. Suffice it to say that no boundary condition, load or constraint should, be applied without an understand— ing of its effect on the displacement. fins‘teittaintt Recent advances in solid modeling, com— puter speed, graphics processing and auto— matic meshing have certainly made finite element analysis more accessible and effec— tive in the general design engineering envi- ronment. However, the jury is still out on Whether they have made the technology eas— ier. The common errors listed throughout rL: finial” nfiina 1‘..an an -L,...-Laf..i ._1n.. Luis at more p In“ uaeix LU Liiuugiiuui 1.11:111‘ ning and interpretation of the model and the results. Implementing correct boundary condi— tions and understanding the physical system being modeled have not yet been automated (and most likely won’t be in the near future). \X’hile making a mesh has certainly been streamlined, technology is not the only issue. Thoughtful modeling and, results interpretation are still the difference between good engineering data and pretty pictures. These skills will only come with experience, not a push—button interface. There is no substitute for experience. Vince Adam: is director of engineering at W/j/zeTe/e, [7213. in Sc/ermlaurg, [L He can be rear/Jed w’a emailarurncenwuzatekeum. % ...
View Full Document

{[ snackBarMessage ]}

What students are saying

  • Left Quote Icon

    As a current student on this bumpy collegiate pathway, I stumbled upon Course Hero, where I can find study resources for nearly all my courses, get online help from tutors 24/7, and even share my old projects, papers, and lecture notes with other students.

    Student Picture

    Kiran Temple University Fox School of Business ‘17, Course Hero Intern

  • Left Quote Icon

    I cannot even describe how much Course Hero helped me this summer. It’s truly become something I can always rely on and help me. In the end, I was not only able to survive summer classes, but I was able to thrive thanks to Course Hero.

    Student Picture

    Dana University of Pennsylvania ‘17, Course Hero Intern

  • Left Quote Icon

    The ability to access any university’s resources through Course Hero proved invaluable in my case. I was behind on Tulane coursework and actually used UCLA’s materials to help me move forward and get everything together on time.

    Student Picture

    Jill Tulane University ‘16, Course Hero Intern