Do FEA Tools Give The Same Answers?

Contents
1. 2. 3. 4. 5. 6. Executive Summary Software Tool Details Summary of Results Problem Descriptions Problems Appendix

A Comparison of Finite Element Analysis Software

Prepared By:
Vince Adams IMPACT Engineering Solutions, Inc. 1800 Nations Drive, Suite 121 Gurnee, IL 60031

EXECUTIVE SUMMARY

The overall conclusion from this investigation was that, in support of the premise, each of these popular FEA tools will allow a competent user to get to approximately the same response if used with a similar amount of care and knowledge of the technology.

Executive Summary When it comes to real parts or engineering problems, concern has been expressed within the general product development industry that some Finite Element Analysis (FEA) tools are advanced or expert products while others are basic, rst pass, or designer level products. The differentiation, both stated and implied, is that the latter group is acceptable for rough, seat-of-the-pants studies but if accurate results are needed, the former group of tools must be utilized. To explore this perceived difference, IMPACT Engineering Solutions, Inc. (IMPACT) was approached by SolidWorks Corporation to perform an independent comparison of their COSMOSWorks Finite Element Analysis (FEA) tool with other, similarly marketed, FEA tools. The tools chosen for this comparison were: COSMOSWorks Version 2005 ANSYS Version 8.1 NEi/Nastran Version 8.3 Pro/MECHANICA Wildre 2 The primary goal of this investigation was to determine if the same or similar results would be obtained if all these tools were used with the exact same boundary conditions, properties and geometry. In nearly all the 10 stress-displacement analyses performed, COSMOSWorks was found to produce results within 10% of the average response from all the FEA tools used in this investigation. This was comparable to the consistency of the other tools. Additionally, it should be noted that each of these tools allowed fast and easy set-up of the problems posed, allowing the user to focus on the engineering aspects of the problem versus the software interface. This represents a much-needed direction in these types of tools. Similarly, in the linear static cases, solution times were all fast enough to justify the use of these tools in an iterative design mode without fear of delay due to analysis. When

the problem called for contact conditions to be used, COSMOSWorks did complete their solution notably faster in all cases.
The overall conclusion from this investigation was that, in support of the premise, each of these popular FEA tools will allow a competent user to get to approximately the same response if used with a similar amount of care and knowledge of the technology. In general, an engineer seeking to make a design decision regarding system performance or part failure would have come to the same conclusion with the data generated by any of these tools. It is fair to say that if these four FEA tools are used by a competent user applying the same assumption set, the results will be identical from a practical standpoint, leading to identical conclusions and design decisions.

PACT IM PAC T E N G I N E E R I N G

DO FEA TOOLS GIVE THE SAME ANSWERS ?

PAGE 1

SOFTWARE TOOL DETAILS

Software Tool Details Four software tools were evaluated in this study, COSMOSWorks, ANSYS, NEi/Nastran, and Pro/MECHANICA. IMPACT Engineering Solutions, Inc. utilizes these tools regularly and provides support and education at an advanced level for each one of them. IMPACT does not sell any of these tools. A more detailed discussion of each tool follows:
FEA SOFTWARE TESTED COSMOSWorks Version 2005 ANSYS Version 8.1 NEi/Nastran Version 8.3 Pro/MECHANICA Wildre 2

ANSYS ANSYS Version 8.1, developed by ANSYS Inc, Canonsburg, PA, was used in this study. All the solutions in this study were completed using the ANSYS Structural conguration with the Workbench interface. The only exception was the Plastic Tub shell model. Due to problems importing clean geometry into Workbench, the part was meshed in a different tool and solved in the classic ANSYS environment. Workbench is associative with some CAD models in their native format and can import CAD les in neutral formats such as Parasolids, IGES, or STEP. COSMOS COSMOSWorks Version 2005, developed by SolidWorks Corporation, Concord, MA, was used in this study. COSMOSWorks is embedded inside the SolidWorks CAD tool and is available in three congurations; namely Designer, Professional and Advanced Professional. All the solutions in this study were completed using the Designer conguration with the exception of the modal analysis of the Mufer Guard. This required the Professional conguration. SolidWorks Corporation develops additional standalone COSMOS analysis products that are not exclusively linked to a single CAD software. NEi/Nastran NEi/Nastran for Windows Version 8.3 from Noran Engineering, Westminster, CA, was used in this study. NEi/Nastran uses Femap v8.3 from UGS PLM Solutions as the pre- and post-processor. The Femap interface allows a user to import CAD models directly from their native environments as well as neutral geometry formats such as Parasolids, IGES, or STEP. Noran Engineering also develops NeiWorks, an FEA tool that also works within the SolidWorks environment. Pro/MECHANICA The Wildre 2 Version of Pro/MECHANICA from PTC, Needham, MA, was used in this study. All the solutions in this study were completed using the Pro/ENGINEER Advanced Structural and Thermal Simulation conguration. Pro/MECHANICA can be used as an embedded tool within Pro/ENGINEER, much like COSMOSWorks/SolidWorks or as a stand-alone analysis tool. At the start of the investigation, the geometry for some of the parts was in native SolidWorks or Pro/ENGINEER format while others were in Parasolids, STEP, or IGES. However, the geometry for each study was brought into SolidWorks and then exported in Parasolids or Pro/ENGINEER format. This minimized the effect of translation inaccuracies between various CAD tools on the validity of the analysis comparison.

IM PAC T E N G I N E E R I N G

DO FEA TOOLS GIVE THE SAME ANSWERS ?

PAGE 2

SUMMARY OF RESULTS

While the results from the Pro/ MECHANICA solution are included in the summary tables, the average values computed consider only the h-element products, COSMOSWorks, ANSYS, and NEi/Nastran.

COSMOS, ANSYS, and NEi/Nastran use the same underlying h-element technology, which utilizes more, smaller elements to capture complex stress and displacement elds. Pro/MECHANICA uses p-elements, a technology which utilizes larger elements with the internal ability to adjust the mathematical complexity of these elements to improve the solution. It is reasonable to expect a somewhat different solution based on the difference in technology. Consequently, while the results from the Pro/MECHANICA solution are included in the summary tables, the average values computed consider only the h-element products, COSMOSWorks, ANSYS, and NEi/Nastran. Summary of Results Comparative Tables The following tables summarize the results of the study as described in the executive summary section of this report. Table 1 Displacement Response Summary for All Cases

Table 2 Von Mises Stress Response Summary for All Cases

Table 3 Modal Response Summary for Mufer Guard

IM PAC T E N G I N E E R I N G

DO FEA TOOLS GIVE THE SAME ANSWERS ?

PAGE 3

DETAILED SAMPLE PROBLEM DESCRIPTION AND RESULTS Detailed Sample Problem Descriptions and Results There were ten (10) sample problems used in this study. Nine (9) of these were linear static or static with contact solutions and one was a natural frequency (modal) solution.. The problems used are listed below: 1. Hinge Leaf a. No Contact b. With Contact 2. Bicycle Hub-Spoke System 3. Plastic Toy 4. Brass Chuck 5. Plastic Tub 6. Zinc Ratchet 7. Plastic Cover 8. Aluminum Housing 9. Plastic Pole Restraint 10. Mufer Guard These problems are described in more detail on the following pages along with the detailed results from each analysis tool. Additionally, any observations on the problem or the way any of the software tools being studied handled that problem follow the summary. All stresses are reported in pounds per square inch (psi) and all displacements are reported in inches (in).

IM PAC T E N G I N E E R I N G

DO FEA TOOLS GIVE THE SAME ANSWERS ?

PAGE 4

HINGE LEAF STUDY

Hinge Leaf This study highlights how important the way a code resolves contact is to the nal results. In the No Contact case, the variation in the stress from one tool to the other is much less than when contact is introduced. The Hinge is intentionally forced to slide about the Pin. This will be difcult to resolve if software doesnt update contact conditions to reect the changing geometry. Problem Set-Up
This study highlights how important the way a code resolves contact is to the nal results. This will be difcult to resolve if software doesnt update contact conditions to reect the changing geometry.

Summary of Results

IM PAC T E N G I N E E R I N G

DO FEA TOOLS GIVE THE SAME ANSWERS ?

PAGE 5

HINGE LEAF STUDY RESULTS

Results Plot

Deformation plot from ANSYS WORKBENCH 8.1

Stress plot from ANSYS WORKBENCH 8.1

IM PAC T E N G I N E E R I N G

DO FEA TOOLS GIVE THE SAME ANSWERS ?

PAGE 6

BICYCLE HUB-SPOKE SYSTEM

Bicycle Hub-Spoke System This turned out to be the most difcult problem in the test suite. Referring to the problem set-up image, the Spoke is initially contacting on the outer edge of the hole. However, as it loads up, it rocks due to the deformation of the shaft so that the in the nal solution, it is contacting the inner edge of the hole. Note, that the Pro/MECHANICA stress solution is 20% different than the other stress solutions. One explanation for this is that COSMOSWorks, ANSYS and NEi/Nastran use a Large Displacement solution which is truly nonlinear and will update the contact conditions for changing orientation and position of the various contact pairs (created between elements and nodes of opposing contacting surfaces.) It is believed that Pro/MECHANICA does not update the contact conditions in this manner thus generating a ctitious load distribution in the system.Another observation is that while the NEi/Nastran and ANSYS solutions took 3-5 hours to solve, the COSMOSWorks Large Displacement solution took less than 15 minutes. Problem Set-Up
Another observation is that while the NEi/Nastran and ANSYS solutions took 3-5 hours to solve, the COSMOSWorks Large Displacement solution took less than 15 minutes.

Summary of Results

IM PAC T E N G I N E E R I N G

DO FEA TOOLS GIVE THE SAME ANSWERS ?

PAGE 7

BICYCLE HUB-SPOKE SYSTEM RESULTS

Results Plot

Displacement plot from COSMOSWorks 2005

Stress plot and mesh from COSMOSWorks 2005

IM PAC T E N G I N E E R I N G

DO FEA TOOLS GIVE THE SAME ANSWERS ?

PAGE 8

PLASTIC TOY STUDY

Plastic Toy The noticeably higher stress solution calculated by Pro/MECHANICA is most likely due to the difference in element technologies. However, the three h-element solutions were converged over multiple iterations such that the stress presented is their nal response. Problem Setup

Summary of Results

Results Plot

Displacement and stress plots from Pro/MECHANICA Wildre 2

IM PAC T E N G I N E E R I N G

DO FEA TOOLS GIVE THE SAME ANSWERS ?

PAGE 9

BRASS CHUCK STUDY

Brass Chuck In this example, the contact initiates at the top of the Chuck but as it opens up, the bottom edge of the contact pad engages the Pin. The stress solution in the notch area required multiple convergence passes and highlighted the difference approaches each of these tools provide for implementing local mesh renement. Albeit different, all were satisfactory for the task in this case. Problem Setup

Summary of Results

IM PAC T E N G I N E E R I N G

DO FEA TOOLS GIVE THE SAME ANSWERS ?

PAGE 10

BRASS CHUCK STUDY RESULTS

Results Plot

Displacement plot from ANSYS WORKBENCH 8.1

Stress plot and mesh from ANSYS WORKBENCH 8.1

IM PAC T E N G I N E E R I N G

DO FEA TOOLS GIVE THE SAME ANSWERS ?

PAGE 11

PLASTIC TUB STUDY

Plastic Tub The IGES geometry that existed for this part did not read cleanly into either Pro/ENGINEER or in the standalone Pro/MECHANICA interface. A reasonable attempt was made to heal the geometry or stitch over gaps manually but it became too time consuming, therefore Pro/MECHANICA was not used for this problem Both ANSYS and NEi/Nastran solutions used linear, or 4-noded, quadrilateral elements for the solution while COSMOSWorks used a 2ndorder, 6-noded, triangular elements. On a positive note the higher order triangular mesh in COSMOS converged to a higher stress value than the lower order quadrilateral mesh.
On a positive note the higher order triangular mesh in COSMOS converged to a higher stress value than the lower order quadrilateral mesh.

Problem Setup

Summary of Results

IM PAC T E N G I N E E R I N G

DO FEA TOOLS GIVE THE SAME ANSWERS ?

PAGE 12

PLASTIC TUB STUDY RESULTS

Results Plot

Displacement plot from NEi/Nastran Version 8.3

Stress plot and mesh from NEi/Nastran Version 8.3

IM PAC T E N G I N E E R I N G

DO FEA TOOLS GIVE THE SAME ANSWERS ?

PAGE 13

ZINC RATCHET STUDY

Zinc Ratchet Again, the IGES geometry that existed for this part did not read cleanly into either Pro/ENGINEER or the standalone Pro/MECHANICA interface. A reasonable attempt was made to heal the geometry or stitch over gaps manually but it became too time consuming. Therefore Pro/MECHANICA was not used for this problem An interesting boundary condition in this problem was tying the front bore to a central point constrained in just Y-direction. This was extremely easy to do in COSMOSWorks using the Remote Load/Constraint option. It was fairly easy in NEi/Nastran using a single rigid RBE3 element. However, ANSYS version 8.1 or Workbench did not have an easy way to apply this constraint in the user interface. A few lines of classic ANSYS pre-processing commands were used. ANSYS version 9.0 now supports this constraint directly in the user interface. Problem Setup
An interesting boundary condition in this problem was tying the front bore to a central point constrained in just Y-direction. This was extremely easy to do in COSMOSWorks using the Remote Load/Constraint option. It was fairly easy in NEi/Nastran using a single rigid RBE3 element.

Summary of Results

IM PAC T E N G I N E E R I N G

DO FEA TOOLS GIVE THE SAME ANSWERS ?

PAGE 14

ZINK RATCHET STUDY RESULTS

Results Plot

Displacement plot from NEi/Nastran Version 8.3

Stress plot and mesh from NEi/Nastran Version 8.3

IM PAC T E N G I N E E R I N G

DO FEA TOOLS GIVE THE SAME ANSWERS ?

PAGE 15

PLASTIC COVER STUDY

Plastic Cover This case warranted the most convergence passes of all the problems in the study.

Problem Setup

Summary Results

IM PAC T E N G I N E E R I N G

DO FEA TOOLS GIVE THE SAME ANSWERS ?

PAGE 16

PLASTIC COVER STUDY RESULTS

Results Plot

Displacement plot from COSMOSWorks 2005

Stress plot and detail from COSMOSWorks 2005

IM PAC T E N G I N E E R I N G

DO FEA TOOLS GIVE THE SAME ANSWERS ?

PAGE 17

ALUMINUM HOUSING STUDY

Aluminum Housing As evidenced by the consistency of the results, this was one of the most straightforward problems in the study.

Problem Setup

Summary of Results

IM PAC T E N G I N E E R I N G

DO FEA TOOLS GIVE THE SAME ANSWERS ?

PAGE 18

ALUMINUM HOUSING STUDY RESULTS

Displacement plot from NEi/Nastran Version 8.3

Stress plot from NEi/Nastran Version 8.3

IM PAC T E N G I N E E R I N G

DO FEA TOOLS GIVE THE SAME ANSWERS ?

PAGE 19

PLASTIC POLE RESTRAINT STUDY

Plastic Pole Restraint This was a well behaved contact problem. Regions that come into contact initially stay in contact without much sliding. Problem Setup

Summary Results

IM PAC T E N G I N E E R I N G

DO FEA TOOLS GIVE THE SAME ANSWERS ?

PAGE 20

PLASTIC POLE RESTRAINT STUDY RESULTS

Results Plot

Displacement plot from COSMOSWorks 2005

Stress plot and mesh from COSMOSWorks 2005

IM PAC T E N G I N E E R I N G

DO FEA TOOLS GIVE THE SAME ANSWERS ?

PAGE 21

MUFFLER GUARD STUDY

Mufer Guard All mode shapes, across the four software packages, were identical. There was very little variation in the problem set-up or response. Product Setup

Summary Results

Results Plot

Frequency plot from COSMOSWorks 2005

IM PAC T E N G I N E E R I N G

DO FEA TOOLS GIVE THE SAME ANSWERS ?

PAGE 22

APPENDIX

Appendix Additional Study Details and Rationale IMPACT accepted this investigation as part of our on-going program to provide informative, objective and educational material to the FEA user base and to debunk the myths, if they are in fact myths, regarding the difference in accuracy between the various hierarchies of analysis tools. This appendix provides some additional detail on the study and the rationale behind the choices made. Background Many investigations of this nature have been undertaken in the past using textbook or highly simplied examples, both by independent sources and by the developers of the software themselves. Software benchmarks, most notably those developed by NAFEMS, typically indicate that tools in question will yield acceptable levels of accuracy on targeted problems with closedform or well-documented solutions. The authors experience with these benchmarks and the software used in this study is that they all perform reasonably well. Readers are encouraged to try these on their own. Our experience working with engineering companies and analysis groups around the world has been that most popular FEA tools, if used correctly and within the limits of their solver, element technologies, and physics capabilities will generate valid results for most linear static solutions and many nonlinear or dynamic problems. Similarly, while some of the basic category tools may lack the breadth of features to provide accurate results in some more complex problems, all of the advanced tools provide a multitude of ways to generate inaccurate results to even basic problems in the hands of an under-trained, unsupervised user. Where variation between tools or techniques becomes dangerous is when a user might draw erroneous conclusions regarding part acceptability from FE data that otherwise looks reasonable. It is always important to remind users and managers who are active in this technology that the validity of the results goes far beyond the accuracy or consistency of the tool being used. If the other assumptions that dene a problem are invalid, yet held consistent across multiple software products, consistent incorrect answers could foster a false sense of condence in the data. Consistency versus Correctness The intent of this study was not to document, or even attempt to obtain, a correct solution. Simply dening correctness is beyond the scope of this investigation. Additionally, a study of a tools ability to calculate a correct solution is best performed on simple, well-dened benchmark problems as described previously when the validity of a target solution is beyond question. Instead, the intent of this study was to utilize each of the subject FEA tools on actual, manufactured or manufacturable parts with properties and boundary conditions that mimic actual or reasonable use to determine if the results differ and by how much. Armed with this information and being cognizant of studies on accuracy performed in the preferred manner described previously, a potential user of this technology can be assured that failure to achieve desired results with any of these tools rests in the quality of use, not the tool itself. This is critical because users have, within their control, the ability to improve their skill level and quality of use but few have access to multiple analysis tools and even fewer can affect changes in the software algorithms themselves.

The authors experience with these benchmarks and the software used in this study is that they all perform reasonably well. Readers are encouraged to try these on their own.

IM PAC T E N G I N E E R I N G

DO FEA TOOLS GIVE THE SAME ANSWERS ?

PAGE 23

Representative, Reasonable, & Repeatable Conditions It was a guiding premise behind this investigation that the problems chosen, while actual parts from a variety of sources, not instigate a debate on correct or proper boundary conditions or properties for a particular application. For example, one of the problems used in this study is one leaf of a stamped hinge engaging a stop pin. Engineers familiar with hinge design could, quite rightly, take exception to the choice of constraints used. However, it is unlikely that any attempt to develop a more realistic set of boundary conditions would have satised all of the critics, thus obscuring the intent of the study. Consequently, the boundary conditions chosen for these problems will provide a representative and reasonable response in the part or assembly that is indicative of the intended use, not to attempt to validate the design or robustness of the part itself. In some cases, the boundary conditions were chosen for repeatability across all tools to ensure set-up consistency. In other cases, they were chosen arbitrarily to produce a non-trivial response in an area that might be of concern to a developer of similar parts. Similarly, the choice of material properties was either arbitrary or restricted to the narrow focus of this study which was primarily linear static analyses. Consequently, some of the results show stress levels that most likely would ag an engineer to consider nonlinear material behavior. This was ignored in favor of, again, results consistency versus accuracy. No attempt to engineer these parts was made so the concept of failure from a materials perspective is moot. A subsequent study of the consistency of these and/or other tools in a nonlinear material or advanced dynamic scenario may lead to conclusions that differ from those found in this study but any attempt to qualify this performance was beyond the scope of this investigation. Scope of Report With a primary goal of reporting consistency in results, detailed model information, such as mesh size, problem set-up time, & run time, was not diligently recorded. It is our belief that there is so much variation in the way people work, even within the same tool, that precise reporting of such details could diminish the more salient points of the study. Therefore, the reported data will focus on output values that are representative of the results generated by each software product. An indicative location for displacement and stress, Von Mises Stress in all cases, was chosen based on the performance of the part under the given boundary conditions and then reported consistently in all cases. As with the discussion of the rationale behind the problem set-up, the output locations selected were not necessarily chosen based on where an engineer might need to look to validate the design. They were chosen, rst and foremost, for the repeatability of the results at that location and then in an area of stress concentration so that any differences would not be trivial. However, throughout the course of the investigation, general relative observations were made about ease of setup, solution time, total problem duration, etc that, when one of the software products stood out from the crowd, in either a positive or negative way, will be noted.

IM PAC T E N G I N E E R I N G

DO FEA TOOLS GIVE THE SAME ANSWERS ?

PAGE 24

About the Author - Vince Adams Over the past 18 years, Vince Adams has established a reputation as an expert in the use of Finite Element Analysis as an integral part of the product development process. Mr. Adams is co-author of Building Better Products with Finite Element Analysis from OnWord Press and numerous articles on FEA. As an invited speaker at conferences on FEA and product design around the world, Mr. Adams brings a fresh perspective to both the use and management of the technology in a product design environment. Mr. Adams has been recognized by his peer group of engineering professionals both locally and internationally. He was invited by NAFEMS, the international organization for FEA quality and education, to serve as the inaugural chairman for their North American Steering Committee in 1999. His career in product development includes multiple successes as a Product Design Engineer and Project Manager, accumulating several US & international patents. Mr. Adams is currently the Director of Analysis Services at IMPACT Engineering Solutions, Inc., a leading provider of engineering support solutions since 1987, providing exible and scalable solutions that are tailored to meet the unique needs of progressive manufacturers in product design and development.

IM PAC T E N G I N E E R I N G

DO FEA TOOLS GIVE THE SAME ANSWERS ?

PAGE 26

IMPACT Engineering Solutions, Inc. Milwaukee Ofce 3315 N. 124th Street Suite N Brookeld, WI. 53005 Phone 262-317-8100 Fax 262-317-8101 www.impactengsol.com Chicago Ofce Hawkston Hall Ofce Center 1 1800 Nations Drive, Suite 121 Gurnee, IL. 60031 Phone 847-599-5635 Fax 847-599-6035 IMPACT Engineering Solutions serves YOU, the progressive manufacturer, with a diverse, vertically integrated line of services. Since our inception in 1987, IMPACT has utilized the latest CAD, CAE and collaboration technologies to help you reduce product development time and become more productive in the process. By working with IMPACT, you benet from the wide range of clientele and products weve been involved with over the years. Capitalize on that experience.