Verification Procedure for MSC-NASTRAN
Contents
1. NASA Contractor Report 4675 A Verification Procedure for MSC NASTRAN Finite Element Models Alan E Stockwell Lockheed Engineering amp Sciences Company Hampton Virginia National Aeronautics and Space Administration Prepared for Langley Research Center Langley Research Center Hampton Virginia 23681 0001 under Contract NAS1 19000 June 1995 Printed copies available from the following NASA Center for AeroSpace Information 800 Elkridge Landing Road Linthicum Heights MD 21090 2934 301 621 0390 National Technical Information Service NTIS 5285 Port Royal Road Springfield VA 22161 2171 703 487 4650 Acknowledgments The author wishes to acknowledge the contributions of Mercedes Reaves and Raymond Kvaternik of NASA Langley Research Center Both Ray and Mercedes offered several helpful suggestions and Ray supplied a stack of related publications from his personal files Table of Contents Acknowledgements iil 1 0 Introduction 1 2 0 Preprocessor Checks 1 3 0 Analytical Checks 2 4 0 Model Verification Procedure 9 5 0 Summary 17 6 0 References 18 PRECEDING PAGE BLANK NOT FILMED v y PAGELA INTENTIONALLY BLANK 1 0 Introduction Finite element models FEMs are used in the design and analysis of aircraft to mathematically describe the airframe structure for such diverse tasks as flutter analysis and actively controlled landing gear design FEMs are used to model the entire airplane as well as ai2. However since test analysis correlation is not an exact science care should be taken in interpreting the results 3 4 Checking Static Analysis Output After a static analysis has been executed NASTRAN provides several diagnostics that can be used to check the results The output is described briefly in the following sections Reference 6 contains a detailed description of these features 3 4 1 FBS Diagnostics MSC NASTRAN solves static analysis problems by decomposing the stiffness matrix and then using forward backward substitution FBS to solve for the displacement vectors The FBS module provides useful diagnostics to help the user determine if there were numerical conditioning problems during the solution First a Residual Load Vector is calculated by subtracting the applied load vector from the product of the stiffness matrix times the calculated displacement vector P Ku P The Residual Load Vector is not printed unless the user requests it by inserting PARAM IRES 1 in the Bulk Data Deck A better measure of the error is obtained by computing the ratio of the work done by the residual forces to the work done by the applied forces e UTE ul P This error measure is printed under the heading EPSILON NASTRAN flags epsilons larger than 0 001 however MSC suggests that epsilons in the neighborhood of 10 9 are generally considered acceptable The external work done by the applied loads is also reported 3 4 2 OLOAD Re
3. release all other single point constraints which represent a physical connection to ground and apply an enforced displacement at each restrained DOF one at a time For each of the six resulting analyses check the deformed shape both visually and numerically For example if a unit displacement is applied in the x direction then all x displacements should be equal to 1 For large structures this check is easier to assess by using a post processor to display the deformed shape The issue of boundary conditions is again an important consideration in this process The procedure calls for checking the structure in the free free completely unrestrained condition However for some structures e g symmetric structures the analyst may also 7 want to check the structure with SPCs applied to insure that strain free motion is still possible in the DOF that are not affected by the boundary restraints 3 3 3 Checks Against Reference Data Frequently there exists analytical or experimental data which can be used to validate the model For example if a model is delivered from one contractor to another then translated from one analysis program to another the results of the translation can be checked if a set of reference data is available This might take the form of a set of displacements or element forces caused by a given loading It is a relatively simple task to make such a comparison Test data if available can also be used to check a model
4. ABSTRACT 14 SUBJECT TERMS Finite element modeling verification MSC NASTRAN 19 SECURITY CLASSIFICATION OF ABSTRACT Unclassified 18 SECURITY CLASSIFICATION OF THIS PAGE Unclassified 17 SECURITY CLASSIFICATION OF REPORT Unclassified NSN 7540 01 280 5500 Standard Form 298 Rev 2 89 Prescribed by ANSI Std Z39 18 298 102
5. AT GRID POINT 5754 CHOSEN BECAUSE IT IS NEAR THE CG 100 5754 200 5754 300 5754 400 5754 500 5754 Check the output for the following NOTE NASTRAN will issue a series of warning messages UWM 3204 Ignore these They are left over from the older versions of MSC NASTRAN which required a dummy load card with enforced displacements e g SPCDs 1 OLOAD RESULTANT All terms should be zero since no loads were applied 2 Inspect the messages output by the DCMP module sparse decomposition User Information Message 4158 provides statistics which inform the user of the numerical quality of the stiffness matrix KLL 3 User Information Message 5293 output by the forward backward substitution FBS module provides more information about the numerical quality of KLL The error measure epsilon and the external work are printed 4 Check the spc force resultant the maximum spc forces maximum displacements and maximum applied loads Note that the spc forces will not be exactly zero especially for the rotational DOF 5 Finally a good visual confirmation of the results can be obtained by plotting the deformed shapes The plots should not show any visible deformation other than rigid body motion 16 4 6 Static Analysis Checks Against Reference Data If data is available from the finite element model developer a static analysis run can be set up to verify results such as displacements at key locations caused by prescribed
6. MSC NASTRAN automatically checks for mechanisms every time it performs a decomposition using the DCMP module Diagnostics are printed if any mechanisms are detected Unlike the one time grid point singularity check decompositions of various matrices are performed in several places in a typical solution For example if the user requests static condensation the static transformation matrix GOAT which relates the omitted degrees of freedom to the analysis degrees of freedom is obtained by a decomposition of the stiffness matrix KOO The DCMP module is used again in a typical static analysis in which the solution is obtained by a decomposition of the leftover stiffness matrix KLL followed by a forward backward substitution During decomposition each diagonal term in the stiffness matrix is compared to the corresponding term of the factor diagonal matrix D Ratios larger than the value set by the user parameter MAXRATIO are printed with the corresponding grid point and DOF identified The user should carefully inspect the output file If there are decomposition diagnostics determine the set level A set L set etc at which the problems occurred A more detailed discussion of mechanisms is contained in references 6 8 and 9 5 3 2 5 Multi Level Strain Energy Checks Multi level strain energy checks are another means of detecting modeling errors The checks are not built into the structured solution sequences however a DMAP alter is availa
7. errors it is not an exhaustive investigation of modeling details nor does it address the issue of whether the structure is modeled appropriately The assumption is made that the structure was carefully modeled by the developer The FEM user s task is to ensure only that the model makes mathematical sense and contains no obvious errors or omissions Several methods such as kinetic energy and effective mass can be used to evaluate the dynamic properties of FEMs These methods can also identify weaknesses or modeling errors However their primary purposes are to characterize the dynamics of the structure and to guide a dynamicist in selecting a valid set of structural vibration modes for a particular analysis A discussion of these methods is beyond the scope of this document 2 0 Preprocessor Checks 2 1 Introduction Preprocessor model checks are important in developing a FEM However some of the standard checks employed in the model development process e g element aspect ratio or taper may not be necessary or appropriate when a model created by another organization is being verified The discussion that follows focuses on those particular preprocessor checks which are valuable tools in verifying a model developed outside the user s organization 2 2 Visual Checks An analyst using a FEM developed by an outside source can employ preprocessor visual checks to become familiar with the model and to ensure that it looks reasonable Preprocess
8. forces The same care should be taken to methodically check the output of such a run however the objective is to simply compare results Since this procedure would vary depending on the model no example is given here 5 0 Summary Finite element models FEMs are the basis for a variety of engineering computations such as stress and stability analysis and vibration and dynamic load analysis Although a carefully developed thoroughly validated FEM is always desirable it is of utmost importance when the model is being used in the preliminary design stages of a large project Since no data is available to validate the model FEM developers must use their best engineering judgment to model the structure accurately FEM users may want to assess the suitability of the developer s modeling techniques however a user s primary validation task is to ensure that the model is mathematically correct and does not contain any inadvertent errors This document outlines a suggested procedure for accomplishing this task 17 6 0 References 1 PATRAN Plus User Manual PDA Engineering October 1989 2 MSC NASTRAN Reference Manual Version 68 Lahey R S Miller M P and Reymond M The MacNeal Schwendler Corporation 1994 3 I DEAS Master Series Structural Dynamics Research Corporation 1993 4 Development and Applications of a Multi Level Strain Energy Method For Detecting Finite Element Modeling Errors Hashemi Kia M Kilroy K and Pa
9. 3 Grid Point Singularities MSC NASTRAN s Grid Point Singularity Processor GPSP is automatically included in all solution sequences Grid point singularities are defined as zero or near zero terms in the stiffness matrix They are either the result of modeling errors such as missing elements or are caused by undefined DOF such as in plane drilling rotations of plate elements The GPSP module inspects each DOF in the N set independent degrees of freedom left after multipoint constraint elimination Principal stiffnesses are calculated for the three translational and three rotational DOF at each grid point and each grid point stiffness term is divided by the corresponding principal stiffness The resulting ratio e is compared to the minimum allowable value set by the user with PARAM EPZERO default 10 8 Singular degrees of freedom i e DOF whose e is less than EPZERO are listed along with their corresponding stiffness ratio If the parameter AUTOSPC is set to YES default the DOF are automatically constrained and the original set membership and new set membership after the DOF are constrained are listed The GPSP output should be carefully inspected Reference 6 contains a thorough discussion of the procedure 3 2 4 Mechanisms Mechanisms lead to stiffness matrix singularities involving two or more grid points An example of a mechanism is a section of a structure that is capable of rigid body motion in one or more directions
10. 995 Contractor Report 4 TITLE AND SUBTITLE 5 FUNDING NUMBERS A Verification Procedure for MSC NASTRAN C NAS1 19000 Finite Element Models WU 233 01 01 03 6 AUTHOR S Alan E Stockwell 8 PERFORMING ORGANIZATION REPORT NUMBER 7 PERFORMING ORGANIZATION NAME S AND ADDRESS ES Lockheed Engineering and Sciences Company Langley Program Office 144 Research Drive Hampton VA 23666 9 SPONSORING MONITORING AGENCY NAME S AND ADDRESS ES National Aeronautics and Space Administration Langley Research Center Hampton VA 23681 0001 10 SPONSORING MONITORING AGENCY REPORT NUMBER NASA CR 4675 11 SUPPLEMENTARY NOTES Langley Technical Monitor Howard M Adelman 12a DISTRIBUTION AVAILABILITY STATEMENT 12b DISTRIBUTION CODE Unclassified Unlimited Subject Category 05 Availability NASA CASI 301 621 0390 Finite element models FEMs are used in the design and analysis of aircraft to mathematically describe the airframe structure for such diverse tasks as flutter analysis and actively controlled landing gear design FEMs are used to model the entire airplane as well as airframe components The purpose of this document is to describe recommended methods for verifying the quality of the FEMs and to specifiy a step by step procedure to implementing the methods 15 NUMBER OF PAGES 24 16 PRICE CODE A03 20 LIMITATION OF
11. C1 SPC1 ENDDATA Notes AUTOSPC K6ROT POST NEWSEQ MAXRATIO BAILOUT GRDPNT 0 WTMASS 1 Gravity in 1 101 2 Gravity in 1 102 3 Gravity in 1 103 massl baf 101 102 103 0 0 0 YES 5 0 2 el 1 1 E7 Y and Z directions X Direction Symmetric BCs Y Direction Symmetric BCs Z Direction Symmetric BCs 386 1 386 1 386 1 oor oOo O Or o O O O He oO ooo 00259 Constraints for Static Test Loads The following constraints are used to support the model for the static load checks symmetric constraints symmetric normal modes They are used in addition to the free free They should be removed before calculating 3 5639 13 5754 1 The gravity loads are applied with symmetric boundary conditions enforced for the subcases shown SPC 1 Additional supports shown in the Bulk Data deck above 13 are required to remove rigid body motion in the free DOF Since Version 68 of MSC NASTRAN allows changes in SPC sets between subcases the anti symmetric boundary conditions could also be checked in the same run by adding three more subcases which refer to a different SPC set 2 PARAM GRDPNT turns on the Grid Point Weight Generator Although this is optional weight calculations are inexpensive and the output provides one more item that can be used to verify that the input data is correct Check the output for the following 1 Check the Grid Point Weight Generator ou
12. E MODEL CHECKS SUBTITLE Enforced Displacements ECHO NONE MAXLINES 200000 14 LINE 52 DISP PLOT ALL SPCFORCES ALL Model in Free Free State SUBCASE 1 LABEL X translation free free model SPC 3 LOAD 100 SUBCASE 2 LABEL Y translation free free model SPC 3 LOAD 200 SUBCASE 3 LABEL Z translation free free model SPC 3 LOAD 300 SUBCASE 4 LABEL X rotation free free model SPC 3 LOAD 400 SUBCASE 5 LABEL Y rotation free free model SPC 3 LOAD 500 SUBCASE 6 LABEL Z rotation free free model SPC 3 LOAD 600 Symmetric Boundary Conditions SUBCASE 11 LABEL X translation SYMMETRIC BCS SPC 13 LOAD 100 SUBCASE 12 LABEL Z translation SYMMETRIC BCS SPC 13 LOAD 300 SUBCASE 13 LABEL Y rotation SYMMETRIC BCS SPC 13 LOAD 500 Antisymmetric Boundary Conditions SUBCASE 21 LABEL Y translation ANTISYMMETRIC BCS SPC 23 LOAD 200 SUBCASE 22 LABEL X rotation ANTISYMMETRIC BCS SPC 23 LOAD 400 SUBCASE 23 LABEL Z rotation ANTISYMMETRIC BCS SPC 23 LOAD 600 BEGIN BULK PARAM AUTOSPC YES PARAM K6ROT 5 9 PARAM POST 2 PARAM OGEOM NO PARAM NEWSEQ 1 PARAM MAXRATIO 1 87 15 SPC SET SYMMETRIC BC CARDS LOCATED AT END OF BULK DATA ANTISYMMETRIC BC os CONSTRAIN GRID POINT 5754 NEAR CG IN ALL SIX DOF COMBINE SET 1 AND SET 3 COMBINE SET 2 AND SET 3 3 123456 5754 13 1 3 23 2 3 ENFORCED DISPLACEMENTS
13. TLE PARAM CHECKOUT YES ECHO NONE MAXLINES 200000 LINE 52 PARAM CHECKOUT YES BEGIN BULK Include mass data optional for this run include massl1 bdf Parameters used in all static and dynamic runs AUTOSPC YES K6ROT 5 0 POST 2 NEWSEQ MAXRATIO 1 E7 BAILOUT 1 Rest of Bulk Data Check the output for the following 1 Look at the printout of matrix Emh EMH Nonzero terms in EMH indicate improper multipoint constraint equations i e equations having internal constraints This could be caused by errors on MPC cards or rigid element cards such as RBARs Note that the use of MSC NASTRAN s rigid elements RBARs RBE2s etc instead of MPCs wherever possible will help prevent multipoint constraint errors because the program automatically generates the constraint equations for these elements If there are no internal constraint errors all columns of EMH will be null 2 Look for messages from the DCMP module NASTRAN decomposes the matrices Rim RGMM and RT RMMM to detect the presence of linearly dependent equations at both the G set level and the M set level If there are errors of this type NASTRAN will issue several error messages including the following USER FATAL MESSAGE 6137 rank deficient matrix e USER FATAL MESSAGE 5225 attempt to operate on a singular matrix USER INFORMATION MESSAGE 4158 decomposition statistics USER INFORMATION MESSAGE 4698 more decomposition statistics 10 U
14. ble The alter checka v68 is located in the misc subdirectory on the MSC NASTRAN delivery tape The checks are referred to as multi level because the computations are performed at various set levels such as the G set N set and A set level Since each level represents a significant step in the reduction process leading to the final equations the checks can be a useful means of determining both the location and the cause of errors For example an error that occurs at the A set level but does not occur at the N set level would be caused by a static or dynamic reduction error rather than a rigid body constraint problem An important issue to consider when using these checks is the application of specific boundary conditions If all single point constraints except those that involve zero stiffness DOF e g drilling rotations in QUADA plate elements are removed then the checks will be independent of specific boundary conditions This may be useful in identifying hidden problems such as grounding However the analyst may also want to check the effects of the single point constraint SPC elimination See the discussion in section 3 2 The multi level strain energy check procedure consists of computing a set of six rigid body displacement vectors then using them to compute forces rigid body strain energy and rigid body mass matrices The checks proceed as follows Stiffness Checks 1 At the G set level compute a set of rigid body v
15. checks may also be useful Differences in the results between the analysis program and the preprocessor may indicate translation problems 2 4 Summary This section has presented some brief guidelines for using a graphical preprocessor to verify a model A detailed discussion of commercial FEM preprocessors is beyond the scope of this document and it is left to the individual analyst to correctly use the features of a selected preprocessor in an effective manner 3 0 Analytical Checks 3 1 Introduction The purpose of the analytical checks described in this section is to ensure the mathematical soundness of the model and to uncover any gross modeling errors such as missing elements or incorrectly applied boundary conditions The material presented herein consists of generally accepted practices and procedures Most of the methods are described in references 4 5 and 6 3 2 Pre analysis Mass Stiffness and Matrix Reduction Checks Several analytical model checks can be performed prior to any static or dynamic analysis These checks are referred to as pre analysis checks because they are computations which are performed on the mass and or stiffness matrices and are independent of specific boundary conditions or applied loads 3 2 1 Constraint Checking MSC NASTRAN 2 provides the option in any of the Structured Solution sequences SOL 101 200 of requesting a superelement checkout run by including PARAM CHECKOUT YES in either t
16. e compared to the rigid body mass matrix MO calculated by the Grid Point Weight Generator For example at the G set level WGHT RBG MGG RBG The matrix WGHT should be equal to MO The matrix WGHTN calculated at the N set level should also be equal to MO However when the same calculations are performed at the A set level the matrix WGHTA will not necessarily be the same if the structure has any single point constraints applied Any mass associated with the constrained DOF will not be accounted for in WGHTA This problem can be avoided by temporarily removing the SPC card in the Case Control Deck The mass checks are a useful means of verifying that the proper mass of the structure was retained throughout the reduction process 3 3 Static Analysis Checks Simple static analyses can be performed to check a model If pre analysis checks have been made some of the static analyses may be somewhat redundant However a static analysis is generally a relatively inexpensive means of checking the soundness of a finite element model 3 3 1 1 9 Check Using an appropriate set of boundary conditions apply a unit gravity load to the entire structure The resulting displaced shape can be inspected for reasonableness For example are there any parts of the structure that show suspiciously large displacements Does the overall deformed shape look reasonable 3 3 2 Enforced Displacement Check Constrain a single grid point in all six DOF
17. e origin of the basic coordinate system which is the default The center of gravity is also calculated with respect to the reference point OUTPUT FROM GRID POINT WEIGHT GENERATOR REFERENCE POINT 150002 MO 3 719099E 00 0 000000E 00 0 000000E 00 0 000000E 00 0 000000E 00 0 000000E 00 0 000000E 00 3 719099E 00 0 000000E 00 0 000000E 00 0 000000E 00 2 179047E 03 0 000000E 00 0 000000E 00 3 719099E 00 0 000000E 00 2 179047E 03 0 000000E 00 0 000000E 00 0 000000E 00 0 000000E 00 4 126100E 04 0 000000E 00 0 000000E 00 0 000000E 00 0 000000E 00 2 179047E 03 0 000000E 00 2 125744E 06 0 000000E 00 0 000000E 00 2 179047E 03 0 000000E 00 0 000000E 00 0 000000E 00 2 166884E 06 S 1 000000E 00 0 000000E 00 0 000000E 00 Q 000000E 00 1 000000E 00 0 000000E 00 0 000000E 00 0 000000E 00 1 000000E 00 DIRECTION MASS AXIS SYSTEM 5 MASS X C G Y C G 2 C G x 3 719099E 00 0 000000E 00 0 000000E 00 0 000000E 00 Y 3 719099E 00 5 859073E 02 0 000000E 00 0 000000E 00 Z 3 719099E 00 5 859073E 02 0 000000E 00 0 000000E 00 I S 4 126100E 04 0 000000E 00 0 000000E 00 0 000000E 00 8 490239E 05 0 000000E 00 0 000000E 00 0 000000E 00 8 901639E 05 I Q 4 126100E 04 8 490239E 05 el a 8 901639E 05 Q 1 000000E 00 0 000000E 00 0 000000E 00 0 000000E 00 1 000000E 00 0 000000E 00 0 000000E 00 0 000000E 00 1 000000E 00 The Grid Point Weight Generator is described in detail in reference 6 3 2
18. ectors RBG using the VECPLOT module 2 Compute the reaction forces resulting from the rigid body motion and print the normalized non zero forces REACG KGG RBG 3 Compute and print the strain energy CHKKGG RBG REACG 4 Repeat steps 1 through 3 at the N set level the N set contains all independent DOF not eliminated by multipoint constraints 5 Repeat steps 1 through 3 at the A set level To obtain the A set stiffness matnx NASTRAN first partitions the N set into the F set free DOF and the S set DOF eliminated by single point constraints The F set is then partitioned into A set analysis set and O set omitted DOF and a reduced stiffness matrix KAA is typically computed by the process of Guyan static reduction generalized dynamic reduction or component mode synthesis If no reduction is requested by the user the F set and A set will be equivalent The diagonal terms of the strain energy matrix should all be nearly zero if there are no errors such as grounding problems in the stiffness matrix The reaction forces are normalized by dividing each term by the largest term in the vector If there are non zero terms the elements of the normalized reaction force matrix REACGNRM REACNNRM or REACANRM can be surveyed to find the largest forces Mass Checks Similar calculations are made by pre and post multiplying the mass matrices by the rigid body vectors This process results in a 6x6 matrix that can b
19. freedom 2 The product Rim RmgRihg is calculated and decomposed by the DCMP module During the solution process NASTRAN decomposes symmetric structural matrices into upper and lower triangular factors and a diagonal matrix e g K LDL where L is the lower triangular factor and D is called the factor diagonal matrix Note that the upper triangular factor for a symmetric matrix is equal to LT Symmetric decomposition followed by forward backward substitution is a computationally efficient alternative to matrix inversion An additional benefit of decomposition in MSC NASTRAN is the diagnostic messages that alert the user to problems in the matrices Each diagonal term of K is divided by the corresponding term of the factor diagonal matrix D and ratios larger than PARAM MAXRATIO are printed The number and location of any negative terms in 3 the factor diagonal matrix are also printed In the case of the constraint matrix Ring the terms flagged by the decomposition of Rim indicate the presence of linearly dependent rows in Ring i e redundant constraints This condition will probably cause singularities or poorly conditioned constraints if the problem is not corrected 3 The product Rim A LA is calculated and decomposed and factor diagonal terms larger than MAXRATIO are printed The results of this check may be compared to the results of step 2 A degree of freedom flagged here that was not flagged in step 2 indicates that a p
20. he Bulk Data Deck or Case Control Deck The checkout option triggers a series of checks in the Bookkeeping and Control Phase 0 subDMAP The run is automatically terminated before the matrix assembly generation and reduction operations begin in the Phase subDMAP While this option is primarily intended for checking superelement models it includes a sequence of multipoint constraint checks which are useful for any model These checks detect the presence of internal constraints grounding and ill conditioning The checks operate on the multipoint constraint equation matrix Rmg formed in module GP4 from the MPC and rigid body Bulk Data entries In order to perform the checks NASTRAN partitions Rmg into dependent m set and independent n set sub matrices i e Rng Ram Rmn As described in Section 9 4 1 of reference 2 three tests are performed 1 A matrix of rigid body vectors us is assembled using the VECPLOT module and the product Emh AmgUgh is calculated The terms of Emn larger than PARAM TINY are printed These terms usually indicate internal constraints although exceptions to this may occur if there are MPC equations involving scalar points A simple example of an internal constraint is an MPC equation involving two degrees of freedom in which the coefficient for the independent degree of freedom is inadvertently left blank NASTRAN will assume that the coefficient is zero thereby grounding the dependent degree of
21. lysis requests because this is a pre analysis check run 2 The Alter cannot be used with the scr yes option You must create a database The database files can be deleted after the run Check the output for the following 1 Check the Grid Point Weight Generator GPWG Output Specific items to be checked include Mass Center of Gravity Inertia Matrix I s and the Rigid Body Mass Properties Matrix MO N B The data is computed in weight units according to the value set on PARAM WTMASS and the properties are computed with respect to the reference point selected by PARAM GRDPNT If PARAM GRDPNT is not specified the reference point is taken as the origin ofthe Basic coordinate system Reference 6 contains a thorough discussion ofthe GPWG Inthe example above grid point 5754 was chosen because it is near the CG 2 At each set level G set N set and A set check the strain energy matrix CHKKGG CHKKNN or CHKKAA All diagonal terms should be very small The DMAP will issue a warning message if any of the diagonal terms are greater than 105 Experience has shown that this message may be safely ignored if translational terms are less than about 103 and rotational terms are less than about 101 Rotational terms are a function of the reference point used to compute the rigid body vectors Choosing a reference point outside the structure for example may cause the rotational strain energy to exceed 101 3 If the diagonal terms of a st
22. or plots of the model allow the analyst to verify the overall shape of the model as well as key dimensions If more than one coordinate system has been used to define model geometry the plots are an excellent method of determining whether key structural details are oriented correctly Most preprocessors allow the user to group grid points and elements according to criteria such as physical or material property number Although it may not be feasible to check every property in a model the plots offer a quick method of checking selected data Plots of loads and boundary conditions can also be used to quickly check that they are applied correctly It should be recognized that every time a translation is made from one analysis or preprocessor code to another e g PATRAN 1 to NASTRAN 2 NASTRAN to I DEAS 3 etc there is a potential for introducing errors The analytical checks described in Section 3 provide a good basis for ensuring that the results of some of the preprocessor checks are still valid after the model has been translated into an MSC NASTRAN input file 2 3 Element Checks Most preprocessors will perform element distortion checks that measure quantities such as taper skew angle and aspect ratios Modeling details that violate generally accepted guidelines may not necessarily be incorrect However it is useful for an analyst to be aware of the expected quality of results obtained from various parts of the model Weight property
23. rain energy matrix are not nearly zero as defined above inspect the reaction force matrices to determine the degree s of freedom causing the problem The reaction force matrices REACGNRM REACNNRM and REACANRM are normalized such that the largest force is 1 0 4 Inspect the rigid body mass matrices WGHT WGHTN and WGHTA and compare them to the matrix MO output by the Grid Point Weight Generator Possible reasons tor discrepancies include improper dynamic reduction and errors in rigid body elements or MPCs Some terms in the A set matrix WGHTA may be less than the corresponding terms in MO if boundary conditions e g symmetric antisymmetric have been included in the calculations by specifying an SPC set in the Case Control section It is a good idea to perform the checks for both the free free and constrained conditions for half models that use symmetric or antisymmetric boundary conditions to simulate the other half of the structure 4 4 Static Analysis 1 g Check Set up a Solution 101 Superelement Statics run similar to the one shown below SOL 101 TIME 300 DIAG 8 15 CEND TITLE MODEL CHECKS 12 SUBTITLE ECHO MAXLINES LINE 1 G Loads in X NONE 200000 52 DISP plot all SUBCASE LABEL SPC LOAD SUBCASE LABEL SPC LOAD SUBCASE LABEL SPC LOAD BEGIN BULK include GRAV GRAV GRAV PARAM PARAM PARAM PARAM PARAM PARAM PARAM PARAM YN UY 4 MN 4 Y Y gt SP
24. rframe components Model verification procedures are especially important for large scale FEMs of an entire airplane which are developed by an outside contractor or agency and are used for aeroelastic and dynamic analyses Since there is no test data to validate the FEMs during the preliminary design stage it is especially important for both model developers and users to understand the limitations of the models and to ensure that they are used correctly The purpose of this document is to describe recommended methods for verifying the quality of the FEMs and to specify a step by step procedure for implementing the methods The procedure has been successfully applied to large scale FEMs of preliminary design concepts for the NASA High Speed Civil Transport HSCT aircraft The document is divided into four sections Section 1 is the Introduction Section 2 Preprocessor Checks briefly describes suggested procedures for checking a model using a graphical preprocessor Section 3 Analytical Checks describes methods of verifying the mathematical correctness of the model using the MSC NASTRAN finite element code Section 4 Model Verification Procedure presents a step by step procedure for implementing the analytical checks described in Section 3 Section 4 is intended to be a working document containing a cookbook procedure to facilitate the model checking process and help ensure a consistent level of quality Although this procedure may uncover modeling
25. rker G NASA CR 187447 October 1990 5 Diagnostics in Finite Element Analysis Haggenmacher G W and Lahey R S First Chautauqua on Finite Element Modeling September 15 17 1980 6 MSC NASTRAN Linear Static Analysis Users Guide Version 68 Caffrey J P and Lee J M The MacNeal Schwendler Corporation 1994 7 MSC NASTRAN Programmers Manual Version 64 The MacNeal Schwendler Corporation 1986 8 MSC NASTRAN Numerical Methods User s Guide Version 67 Komzsik L The MacNeal Schwendler Corporation 1992 9 MSC NASTRAN Handbook for Numerical Methods Version 66 Komszik L The MacNeal Schwendler Corporation 1990 18 REPORT DOCUMENTATION PAGE Public reporting burden for this collection of information is estimated to average 1 hour per response including the time for reviewing instructions searching existing data sources gathering and maintaining the data needed and completing and reviewing the collection of information Send comments regarding this burden estimate or any other aspect of this collection of information including suggestions for reducing this burden to Washington Headquarters Services Directorate for Information Operations and Reports 1215 Jefferson Davis Highway Suite 1204 Arlington VA 22202 4302 and to the Office of Management and Budget Paperwork Reduction Project 0704 0188 Washington DC 20503 1 AGENCY USE ONLY Leave blank 2 REPORT DATE 3 REPORT TYPE AND DATES COVERED June 1
26. roblem exists in the dependent partition Rmm but not in the matrix containing all DOF Ring Therefore an error was made in specifying the dependent degrees of freedom The checkout option automatically stops the solution process after the constraint checks Additional model verification may be accomplished by using either specially developed Direct Matrix Abstraction Programs DMAPs or static analyses 3 2 2 Grid Point Weight Generator GPWG After assembling the mass and stiffness matrices NASTRAN can print out a summary of the structure s weight properties including center of gravity total weight and inertia matrix These checks are performed before the mass matrix is converted from weight units to mass units The user should check the location of the center of gravity the total weight in each direction the principal mass axis directions and the inertia matrix The weight should be the same in all three directions unless scalar masses are used Note that the inertia matrix s is not in tensor form The off diagonal terms of l s must be multiplied by 1 0 to convert the matrix to tensor form The inertia tensor is referred to in the MSC NASTRAN documentation 7 as the intermediate inertia matrix I In the sample GPWG output shown below the reference point is taken as grid point 150002 in the model Therefore the rigid body mass matrix MO represents the mass properties of the structure with respect to grid 150002 not th
27. ser Information Message 4698 is the most useful of these because it lists the grid point number and degree of freedom at which the error occurs Errors in RGMM will propagate down to RMMM However an error in RMMM that does not appear in RGMM indicates that a change in the M set exists which will fix the problem see discussion in section 3 4 3 Multi Level Strain Energy Checks Set up another static analysis run Include the DMAP Alter checka v68 for Version 68 or checka v675 for Version 67 5 located in the misc sssalter directory or misc sssalter on VMS machines Set the parameters to the values shown in the following example SOL 101 TIME 300 DIAG 8 15 echooff include checka v675 echoon CEND TITLE MODEL CHECKS SUBTITLE MSC ALTERS ECHO NONE MAXLINES 200000 LINE 52 kkk Comment out SPC selection to check model in free free condition SPC 1 BEGIN BULK Parameters for multilevel mass and stiffness checking CHKMASS 1 gt check mass matrices CHKSTIF 1 gt check stiffness matrices PARAM CHKSTIF 1 PARAM CHKMASS 1 Other analysis parameters SPARAM AUTOSPC YES PARAM K6ROT 5 0 SPARAM POST 2 PARAM NEWSEQ 1 PARAM MAXRATIO 1 E7 PARAM BAILOUT 1 PARAM WTMASS 00259 PARAM GRDPNT 5754 Include mass data or use editor to include in bulk data include massl bdf Rest of Bulk Data ENDDATA 11 Notes 1 There are no output or ana
28. sic coordinate directions for each load vector No location information is given however the table is another useful means of checking that the magnitudes of the displacements make sense for each direction 4 0 Model Verification Procedure 4 1 Introduction The model verification procedure described in this section is based on the techniques outlined in the previous sections The procedures have been used for checking a non superelement High Speed Civil Transport HSCT model generated outside of LaRC delivered in a foreign FE code format and translated into MSC NASTRAN format As described in Section 1 the procedures are intended primarily to verify that the model is mathematically correct Modeling issues such as mesh density element type and usage and connection details are difficult to check in an objective sense and are beyond the scope of this document It is assumed in this section that the model has already been translated into MSC NASTRAN form and has been checked with a graphical preprocessor see the discussion on analytical checks in Section 3 for more detailed descriptions of the MSC NASTRAN procedures and calculations 4 2 Constraint Checks Set up a static analysis run SOL 101 and include PARAM CHECKOUT YES in the Case Control or Bulk Data deck A sample deck is shown below NASTRAN BUFFSIZE 4096 SPARSE 25 NOTE SPARSE 25 not needed for Version 68 INIT DBALL LOGI DBALL 500000 TITLE MODEL CHECKOUT RUN SUBTI
29. sultant The OLOAD Resultant is automatically calculated for each applied load vector It represents the resultant of all applied loads referenced to the origin of the basic coordinate system or to the grid point specified by PARAM GRDPNT Although this computation is 8 an applied loads check and is not really a model check it is an important consideration when static loads are being used to check out a model OLOAD output at the grid point level can be requested by using the Case Control OLOAD card 3 4 3 SPCFORCE Resultant The SPCFORCE Resultant is the summation of all forces of single point constraint with respect to the reference point As in the OLOAD summation the reference point is either the grid point specified by PARAM GRDPNT or the origin of the basic coordinate system SPCFORCES can also be printed for individual grid points A useful equilibrium check can be made by summing the SPCFORCE Resultant and the OLOAD Resultant 3 4 4 Maximum Load NASTRAN automatically prints a summary of the maximum load in each direction of the basic coordinate system for each load vector Care must be taken when reading this output The maximum load may not occur at the same grid point for any of the six directions and the table gives no information as to where the maximum loads occurred This information can be useful as a sanity check 3 4 5 Maximum Displacement The table of maximum displacements is also printed for each of the six ba
30. tput Look for correct weight C G and moments of inertia 2 OLOAD RESULTANT The resultant of forces in the translational directions for each subcase should be compared to the known weight of the structure If the Grid Point Weight Generator used the origin of the basic coordinate system as a reference point PARAM GRDPNT 0 then the resultant OLOAD moments should be the same as the corresponding terms of the matrix MO 3 Inspect the messages output by the DCMP module sparse decomposition User Information Message 4158 provides statistics which inform the user of the numerical quality of the stiffness matrix KLL 4 User Information Message 5293 output by the forward backward substitution FBS module provides more information about the numerical quality of KLL The error measure epsilon and the external work are printed 5 Check the maximum displacements and applied loads Remember that these are a function of possibly arbitrary boundary conditions 6 Finally a good visual confirmation of the results can be obtained by plotting the deformed shapes Look for excessive local displacements or kinks that might indicate missing elements or constraints 4 5 Static Analysis Enforced Displacement Check Set up a Solution 101 Superelement Statics run similar to the one shown below Note that MSC NASTRAN versions prior to version 68 do not allow boundary condition SPC set changes between subcases except for SOL 24 TITL
Download Pdf Manuals
Related Search