ACCURATE ENFORCED MOTION ANALYSIS USING MSC/NASTRAN SUPERELEMENTS, Christopher C. Flani; SDRC Engineering Services Division, Inc. ‘San Diego, California ABSTRACT. The standard approach for performing an enforced motion analysis in MSC/NASTRAN uses very large masses and forces to obtain the desired motion at selected locations. This approach can lead to inaccurate results if the large masses are too large or too small. An alternate approach for enforced motion analysis is presented in this paper. The alternate method uses the Craig-Bampton superelement capability in MSC/NASTRAN to form the required matrices for a direct solution of the equations of enforced motion. The need for large ‘masses is eliminated, resulting in improved accuracy. In addition, the enforced motion analysis is performed directly, eliminating the need for Lagrange multipliers. A rigid format alter for performing the new enforced motion analysis method is included in the Paper. An example problem is presented to demonstrate the new method and to illustrate some of the pitfalls of enforced motion analysis. 1994 MSC/NASTRAN World Users Conference Orlando, Florida June 20-24, 1994ACCURATE ENFORCED MOTION ANALYSIS USING MSC/NASTRAN SUPERELEMENTS Ciristopher C. Flanigan SDRC Engineering Services Division, Inc. ‘San Diego, California ‘Nomenclature ‘Acronyms DOF Degrees of freedom DMAP Direct matrix abstraction programm DRM Data recovery matrix MSC MacNeal-Schwendler Corporation NASTRAN NASA Structural Analysis Program Matrices B Damping I Tdentity K Stiffness Mo Mass P Applied loads P—-Prcudo loads U Displacement © Velocity Acceleration Subscripts { —__ fset (free DOF: g-m-s) £ —g-set(all DOF) 'm —_m-set (DOF constrained by MPC) | —_q-set (component mode DOF) S _s-et (DOF restrined by SPC) t ‘-set (physical boundary DOF) Introduction Enforced motion transient analysis is a very impor- ‘ant capability for the design of dynamic components. In enforced motion transient analysis (also known as “base shake”), motion histories are prescribed at selected locations in a component. The responses at other locations caused by the prescribed motion are applications for enforced motion transient analysis include spacecraft coupled to a launch vehicle and road vehicles traveling over rough terrain. The bi shake method is often used to perform trade studi for modified components using the interface motion histories from a previous system coupled transient analysis. (MSCINASTRAN has the ability to perform enforced motion analysis using the “seismic mass” ap- proach 1]. In this method, extremely large masses or inertias are placed atthe enforced motion locations. ‘Extremely large forces are applied to the large masses to cause the desired motion histories. The seismic ‘mass approach has traditionally been prone to numer- ical error. If the seismic masses are not sufficiently large, dynamic feedback from the component causes the motion ofthe seismic masses to deviate from the prescribed histories. Ifthe seismic masses are too large, numerical ill-conditioning can occur in the ‘mass matrix and eigensolution. ‘This paper presents an alternate formulation for enforced motion transient analysis. The alternate ‘method is based on a simple explicit algorithm that eliminates the need for seismic masses, thereby improving the accuracy of the enforced motion solution. ‘The alternate method is implemented usi superelement methods in MSCINASTRAN to easily {generate the required matrices. The alternate method ig illustrated using an example problem. Finally, some of the limitation of enforced motion analysis are presented, Theory The derivation of the alternate method for enforced ‘motion begins with the component equations of ‘motion: Ke X:+ByX;+MyX=P) (1) Using MSC/NASTRAN superelement methodology {2}, the equations of motion can be reduced from the Fett the eset: [i s]fifefae mafti} De wesley) ‘The form of (2) assumes that standard MSCINASTRAN superelement capabilites ae used, Since MSC/NASTRAN uses an enhanced version of the Craig-Bampton modal synthesis method (3), theoff-diagonal partitions ofthe stiffness matrix are null ‘The form of (2) would be different if any other modal synthesis method wero used such as the MacNeal- Robin esidual flexibility method [4,5]. ‘The equations for enforced motion analysis can be significantly simplified using the following assump- tions and limitations: + Component modal damping only (By= By = 0) + Nointerally applied forces (P,=0) [ eltede eles) [ws wes}-fo} ‘The lower partition of (3) can be written as Kg Uy +B Uy +My Gh +My t=O Ky, Ug +Byj Ug +My) Uy =-Mg, (6) The accelerations of the tset DOF are prescribed using the values from the original coupled loads analysis. This relationship forthe t-st accelerations can be added to (5) to form the equations of enforced ‘motion for the a-set DOF: fo w}tedfeLo melfte}> [se Fl-Ltof © (6)is inthe standard form for a modal transient anal ysis. The solution of (6) will be very efficient and extremely accurate if modal (uncoupled) damping is used. Non-diagonal damping will couple the equa~ tions of motion, thereby requiring a longer a slightly less accurate solution using the Newmark-Beta method, where Internal responses such as element forces and stresses can be recovered using standard MSC/NASTRA’ data recovery capabilities. Alternatively, beter eff ciency and accuracy can be obtained using data recovery matrix methods (6), ‘Implementation ‘The alternate method for enforced motion analysis is implemented in MSC/NASTRAN using a rigid for- mat alter. The rigid format alter for SOL 72 is i cluded in Appendix A To use the alternate method, the user must comply with the following requiee- meats + The enforced motion component must be defined as a single superelement or as a multiple superelements assembled into a single “collector” Superelement. ‘+ The enforced motion DOF must be exterior tothe component, ‘+ Fixed-interface component modes must be caleu- lated (do not use fre or mixed-inerface modes). +The residual structure must include only the exter Jor DOF of the upstream component. No addi- tional grids or elements may be added to the residual structure + The enforced motion DOF must be listed on SUPORT entries in the residual structure. +The acceleration histories for the set DOF must be defined as “applied loads” using TABLED] ‘cards and related input. + Component modal damping may be defined using a TABDMPI table, + Standard Case Control and Bulk Data input must be defined for performing a modal transient analy ‘The rigid format alter forms the required matrices for the enforced acceleration transient analysis (6,7). A modal transient analysis is performed using the pre~ scribed accelerations and the user-specified modal damping. If needed, nonzero initial conditions could he added by two methods: + Special rigid format alters 7.8] + Changing the approach code from “MODES" to ‘DIRECT’ for the TRDI transient response DMAP module and manually defining initial con- ditions using IC and TIC entries. ‘The use ofthe rigid format alter and the required user ‘operations ae illustrated in the following section. Example Problem ‘The example problem was atypical aerospace appli: cation including a spacecraft coupled t0 2 rocket‘motor as shown in Figure 1. The system was excited by thrust transients applied to the rocket nozzle. A baseline coupled loads analysis was performed standard methods to obtain the accelerations at the spacecraft interface. The interface accelerations were converted to TABLEDI statements to perform the enforced motion analysis. ‘The input file forthe enforced motion analysis ofthe spacecraft is shown in Figure 2. The spacecraft was defined a a single superclement with the interface DOF exterior to the superelement. Fixed-interface ‘component modes were calculated to 75 Hz. 1% modal damping for the component modes was defined using a TABDMPI table. The acceleration histories were defined using DLOAD, TLOADI, DAREA, and TABLED! statements. ‘The results from the enforced motion analysis were ‘compared to those of the baseline coupled loads analysis. In addition, a “seismic mass” analysis was performed using the standard capabilities in MSCINASTRAN, The acceleration histories of the ‘enforced motion DOF exactly matched the histories ‘prescribed from the coupled loads analysis as shown in Figure 3 and Table 1. ‘The interior accelerations were reasonably accurate as shown in Table 2. However, there were substantial variations in the ‘element loads as shown in Table 3. For most ofthe ‘element forces, similar results were oblained from the seismic mass and enforced acceleration methods. ‘The reasons forthe differences between the standard analysis and the enforced motion analyses are dis- ‘cussed in the following section. ‘Limitations While this paper presents an alternate method for ‘more accurate enforced motion analysis, there are basic accuracy limitations of the enforced motion approach. These limitations are especially significant for coupled system solutions such as the example problem showa in Figure 1. ‘As shown in Tables 2 and 3, the interior results from the enforced motion analysis did not match those {rom the baseline coupled analysis eventhough there ‘were no changes tothe spacecraft model. For some of the element loads, the differences were extremely large. ‘There were thee major causes forthe response differences. First, the modal damping of 1% applied to the system modes is not numerically equivalent to 1% damping applied to the component modes. The differences between system and component mode damping can be even mote significant when the oe toe] — dss — Oe Fay oe 9% | —d.s0a] — 00 aa [05 toe | —0.00s| — Ps [hor oe | poe] — [6 [00 oa | 000] — 08 a toe | nso — a8 Bo toe | psa] — oe ps tor coe | 01a] — a | oa toe | — 000s — 008 Som noe —0b0al — 908 oe COCO ‘aoe | 0.000 — 008 Table 2 nev acelin Gad [OF] Santad | Eatooed TERT Aes] Sesaie Anas | “Accel | Dione | Mass TO ots“ | 0 opr} —orns | oo 5% [on ops tose “oer aon | — 7 1s asso — 30 or iPass oe oe Pos is] 06s | ase 04% | — ose 1+ [6s | oss | 09% | — ones Lone oes bse ss Pst 55s 85s 08 | os 30-1 [anne [oss [3am [ oss 30 aner [—oncs [0 [es 3b [ oro] oer a4 | — 0.7 a1 — i] — ona] oe a [sr — 8s] aos CV 10‘Table 3. Imerior element forces. Tianen [en ] Sonal] Entaesd Ent Are] Seale] Ses Ww Coio| Ansys | ‘Acs | Ditinnee | Mass | Diflewace ee Toa ama — Tea ae] te] sa Taf 30 | an | — 96.3] a9 7 seer sre | — Sas | aro eC ET ae 860 10 | aT YE Tse te] 0 af ars aon em iL aban asta | — toe | ass a TL a sora ar | ase | TBs os aso on | aa] as feat ease |e TELE sa est ies BL a[ oe aoe | ase aon] eof — se ser — tie en Ss ae |e | Gol —s| 3002827 | se | oe a HL a| aso] ase3s] nse | ease a a| — pas 9033 — ore | — 23.6] — 08 Lato ss ase ss La a0 — 203 ore — 903 0 Hal ss] 9a ae saa aeL— a — oa ors] 80 | — son] hm SLL stare st Lapa] aie se aan als] 9929s ore | — 8] — 09% asf — 2007 aes] —Ta9@ | —a145 098 as] 1-07 07] are] — 207 — 038 ia] — ar] ae a0 Sas] — Boos | goto] ine | — on 108 Sela] — a6] — ge] 07% | — 650] 107% so — esses el a3 | —an | a0 | a0] — a0 Saat aT] ae] BL a[ arses] ate | 3s] a eT YO sols — mn — ase} sre] — as 573% aAPPENDIX A Rigid Format Alter for SOL 72 w[ENFORCED ACCELERATION TRANSIENT ANALYSIS Rigid Format 72 - Modal Transient Analysis with Superelements NSC/MASTRAN Version 87 ‘This alter performs an enforced acceleration transient analysis See the referenced technical paper for more information. Reference: ‘Accurate Enforced Motion Analysis using NGe/NAgTSAN Superelenenta, "1994 HSC/NASTRAN World User's Conference, Orlando, Florida, Sune 20°24, 1994 Requirements to use this alter - sou 72 COMPILE 90L72, souIN=MsCSOU, NOLTST, NOREP Tnelide this alter ienediately before the “CEND* card (CASE CONTROL DECK: Standard sequests for a modal transient analysis (METHOD, DLOAD, STEP, and SDAMP) ‘The METHOD requests for the upstrean superelenent and the Fesidual structure should speci#y the same frequency range. BULK DATA. DECK. ‘The physical exterior {T-set) DOF of the component must be sntered on SUPOR? etatenents ‘The accelerations at the component T-set DOP must be defined as ‘applied loads" ‘The DDRIO and node acceleration options mist be deactivated EXAMPLE NAGTRAN DECK: Sou 72 time 30 bing 8 COMPrLE SOL72, SOUIN=MSCSOU, NOLTST, NORE INCLUDE RPT2D339. SUBTITLE = ENFORCED ACCELERATION TRANSIENT ANALYSIS 8 Sea = ALL $ ALL superelement operations $ SuBcase 10 SUPER 10 METHOD = 75 ‘3 Component modes to 75 Hz supease 10000 EAREL = RESIDUAL STRUCTURE Merucp "75 '$ Component modes to 75 Hz BuGAD = 1 § bynanie loads (ent. accel.) eee 1 Znteazation-stens B§ ‘BEGIN BULK Bulk data for structural model i f Deactivate pomiat and wooAcE PARAM, DORM, 1 ‘PARAM, MOOACC, = 5 § Extorced motion Dor Supore, a4, 123456 SUPORT, 45, 123456 ‘SUPORT, 48, 123456 gurons49;123456 3 Define enforced accelerations 8 BLOAD, 2,2, ,22,1/1-,2/203 pherdideySeds 8 3 TOAD, 1,2, 1 DAREA,1/100/1,1 ‘TABLEDL,1 10, ,0+, -002, 106, .00 Renaining enforced acceleration data 327, .006, .763, é § 18 damping on component modes ‘TARDNPL, 1, CRIT 20. -01,100., -02, NDE $ $ integration steps 8 ‘TSTEP, 1,100, .001,1 5 ‘ENDDATA HISTORY DOCUMENTATION O7-Feb-94 Chris Flanigan ~original version lauren 934. v67. CMEANA/CHLANENT/ <1 learn GlAMA,,. /MOQQDTAGL, 1/4 § JMATMCD CMEAMAT, |! /KOQDIAGL, 2/5 § lmarcen — /onunL/7 /hogsEr/1 Jan QNULL, MQQDTAGI /MQODTAG § Jan GNULL, KOUDEAGL /xQQDIAG § lato MQQDTAG,,., MOQ, /28§ lmaruon —KQQDIAG, "OQ, /28 lwec | USER/VAgty'A°/"Qi/7R" § lMATGEN —, /OQNULL/? /ROOSET/NOQSET § SgwdLL, PuzoZ/PHZOO $ aw Sop = i 3 Damping for component modes 23456789012345676901254567090125456749012365678901236567890123456709012 2 3 4 3 6 7 Tom ‘clasoic™ craig-sanpcon component matrices into full matrix into full matrix ‘coupling matrix[MERGE QQ... VAOT, /MLAR? $ Symmetric merge [ADDS Mra, MLAAZ,, /MLARL $ ada partitions IMopra, MLAAL////6_§ Eabel as symmetric IMenae XQ,» -VAGT, /KLAA $ Symmetric merge lsum MAREGOA 5. Govon €0 make GOA Aurer 953. V6? Before forming GOA BABEL MAKBGOA'S Make COA. $ & 3 Peior co caloulating system modes, renove the R-set partitions I$ of the system ctiffness and mass matrices. This will cause the fg Ssysten nodes" to be identical co the fied-interface component 5 5 nodes of the upstream superelenent. JAUrER — 1022,1052 § v67 Remove auto-OMIT PAREN NEAA|VALCOMP, 13, / § Syonetric partition [PART YOR, VALCOMP, MEK, § Syonetric partition JALTeR —1087,1087 $ V67 Replace READ READ. Souk, ace, ,, RED, CASES, /LAMA, PHIK, 42 081G5/ WAN READAPP=" HODES! /S,N,NBIGV $ Modes lumen 1081, 1061 § v6T Replace RETGL [RErGL fx, 100%, DYNAMICS, CASES.» /LAMA, PHIX AE, BIOWMAT, CUVEC/V/N,READAP®/S,N,NEIGV'§ ‘odes lauezr 106e,3071§ ver Renove auto-expand [MaRGE PEK, ,,, , VALCOMP/PHIA/1 ¢ Row merge Is is tutta avset matrices for the enforced acceleration solution § we | 0 Lo Sone = Be eet 3 o | mRR eyo lauren 1131, 1131 $_v67 Replace TRDL luarcen | /ERR/1/NORSET.$ Reset identity matrix IMATGEN | /NULLLL/7/NOUSED/HOLSET $ Null Leset sq. matrix Ibo NULLUD,xiGH/ aL § apo NULL aL [Abb NULLLL BHEL/ BEL & Iuence HL, ,,, VALCOMD, /KAAENFA/ $ IMeRGe oath | IRR, VALCCMP, /URRENPA/ § [MERGE BHAI |, VALCOMP, /BAAENEA/ § BAR for enforced accel is 5 JS puild A-set forces for the enforced acceleration solution 5 S is cme Pe i Ge 5 Be b es PAREN — POT, ,VALCOMP/, PRT, /1 $ Row partition INpvaD —MLRSPRT,/2Ut/ /—2 8 Smk ta MERGE -PLP|PRT,, ,, VALCOMP/PAENPA/1 ¢ ow merge 2 perform the transient solution 5 NPRD1 CASES, TRL, NLET, DIT, KAAENFA, BAABNFA, MARENPA, PABNPA/ ‘HVE, PNLH) "MODAL /hoUE/V, x/nONeUP=“1/0§ Nodal transient $ § Renove solution set output Auer 133,149 $ v67 Remove MODOUT, HSOR?2 5 Renan the transient response output JALTER 1155,1155 § v67 Replace MP¥AD lap UHV, JUD Romane 5 [Seeznd of RP720339: a5