Paper r Internatinal Cnerence INMARECH008, France Multidiciinary Deign Optimizatin Methd Apied t a HO Deign Ca Anxi, Cui eicheng State Key Labratry Ocean Engineering, Shanghai Jia ng Univerity, Shanghai, 00030, China China Ship Scientiic Reearch Center, P.O.Bx 6, uxi, Jiangu, 408, China Abtract In thi paper, a Multidiciinary Deign Optimizatin (MDO prcedure i apied t a HO deign. Multidiciinary decmpitin and analye have been develped r thi cmex ytem that include hydrdynamic, tructure, prpulin, weight & vlume. he Multi-Objective Cllabrative Optimizatin (MOCO methd i elected t cnduct the preliminary cnceptual deign the HO. hi apprach wa able t identiy Paret rnt deign. he reult al demntrate that MDO apprache are mre uitable r deign the HO and mre lexible and advanced cmpared with the traditinal deign apprach. Keywrd: MDO, HO, Cnceptual Deign, Multi-Objective Cllabrative Optimizatin (MOCO. Intrductin he cnceptual deign tage a cmex engineering deign prce i the mt critical t it ucce r ailure, a at thi tage 75% the inal ct and perrmance metric are determined. he deign a Human Occupied ehicle (HO i a cmex and multidiciinary tak, and the tak i ten divided int a et maller and eaier tractable deign prblem. A cmete deign require analye hydrdynamic, tructure, prpulin, weight, cntrl, peratin, ct and the ther. It i imprtant r each thee apect t be addreed at the cnceptual deign phae. he traditinal apprach r the deign a HO i a equential rder. A hwn in Fig., in thi apprach, deign begin with the irt diciine team, where the value certain deign variable are ixed and paed t the ecnd diciinary team, and n until a cmete deign emerge rm the lat diciine. he traditinal apprach may lead t nn-ptimal ytem deign r many rean including: [] ( Analyi in uptream diciine may depend n reult determined in dwntream diciine. In the equential rder apprach, uptream diciine mut aume value which may nt match the actual value when they are inally determined in the dwntream diciine. ( he ytem bjective (ct, perrmance, etc. may depend heavily n reult determined in dwntream diciine where deign reedm n lnger exit t make igniicant change. (3 he deign may be ixed by time and it reache dwntream diciine that may be
impible r them t atiy their cntraint. It i bviu that the traditinal apprach t a HO deign i nt uitable r the develpment mdern HO which will becme mre and mre cmex. Hydr- dynamic Prpulin Energy Structure eight & lume Fig. he traditinal cnceptual deign architecture a HO he ield multidiciinary deign ptimizatin (MDO ha emerged t develp apprache r ptimizing the deign large cued ytem []. MDO i cncerned with hw t eiciently analyze and ptimally deign a ytem gverned by the multie cued diciine r made up cued cmpnent. It i a part the cncurrent engineering technlgy that may well be an enabling technlgy r cmex advanced ytem [3]. ith the rapid grwth MDO ver the pat decade, MDO ha been widely dicued and ued nt nly in aerpace and aernautical indutrie [4-6], but al in ther cmex engineering ytem uch a autmbile [7], underwater vehicle [8, 9], hip [0] etc. and reulted in a mre reliable and better deign. Fr the purpe attaining the verall perrmance ptimizatin a HO and imprving prcedure a HO cnceptual deign, MDO technique ha been emyed. he purpe thi paper i t exre hw Cllabrative Optimizatin (CO, ne the MDO methd, can be apied in the cnceptual deign a HO.. Multi-Objective Cllabrative Optimizatin (MOCO Cllabrative Optimizatin (CO, ne the multidiciinary deign ptimizatin methd, ha been develped t prmte autnmy while prviding a crdinating mechanim that guaranteeing prgre tward an ptimum and maintaining interdiciinary cmpatibility [, ]. It baically cnit a tw-level ptimizatin tructure. he riginal tructure CO i hwn in Fig.. ithin CO, the deign tak i accmihed by everal diciinary team a well a by a ytem-level team. he diciinary team are ree t deine their wn lcal deign. he tak the diciinary-level team i t ind a lcal deign that atiie lcal cntraint and cme a cle t that peciied by ytem-level ptimizer a pible. he ytem-level team i in charge
adjuting the ytem variable with the gal minimizing r maximizing the ytem-level bjective. hi prblem i ubjected t the interdiciinary cmpatibility cntraint equal t zer. Sytem-Level ptimizer Gal: Deign bjective.t.: Interdiciinary cmpatibility cntraint Subpace ptimizer Subpace ptimizer Subpace ptimizer N Gal: Interdiciinary cmpatibility Gal: Interdiciinary cmpatibility L Gal: Interdiciinary cmpatibility S.t.: Analyi S.t.: Analyi S.t.: Analyi N cntraint cntraint cntraint Analyi Analyi Analyi N Fig. he Baic Cllabrative Optimizatin architecture CO ha been widely dicued and apied in practical engineering prblem. Sme reearcher have apied it t bth ime tet prblem [-5] and mre cmex engineering deign prblem. hee apicatin invlve launch vehicle deign [4], aircrat deign [5,6,6,7], underea vehicle deign [8,9], cnceptual hip deign [0],and turbine engine deign [8]. CO ha been judged highly advantageu in it apicatin t practical engineering deign prblem. At the ame time, many reearcher have cued n extenin r mdiicatin t CO aimed at imprving verall eiciency, permitting their ue n prblem with high dimeninality cuing and imiying their imementatin. In rder t relieve the numerical diicultie caued by certain mathematical manipulatin, the ue an apprximatin mdel ha been prped in ace the diciinary deign in CO [7,9]. relve the cnvergence prblem CO, Kr and Manning [] adpted the direct earch methd uch a Hke and Jeeve methd r the prbabilitic earch methd uch a genetic algrithm intead the gradient-baed methd. A variety extenin have been made t CO including relving the ytem including mixed cntinual and dicrete deign variable [], and intrductin multi-bjective rmulatin [0, ]. In thi tudy, a MOCO ha been elected t handle multibjective ytem. In the MOCO, the gal the ytem level ptimizer i t minimize a ytem level multibjective unctin target variable while atiying cmpatibility cntraint uing a Paret Genetic Algrithm baed n (PGA. PGA lve ytem level ptimizatin prblem with repect t ytem deign variable. Fr each generatin at the ytem level, the diciine are ptimized r each candidate deign rm the ppulatin. he ytem level ptimizatin prblem i decribed a equatin (-3 3
Min N :{,, L, } ( S. t.: J 0, i =,, L, N ( i D.. : x = x, x, x, x, Λ, x,] (3 [ h aux m Equatin (4-7 decribe the ubytem level ptimizatin prblem r a typical ubytem, in thi cae ubytem : x N N h aux i j = ( + ( + ( + ( h x i= aux i i= aux j S. t.: g 0 (5 min x max (6 D.. : x = [ h,( x aux i, x] (7 he MOCO architecture r HO deign that ha been develped r cnceptual deign i briely dicued, a mre detailed decriptin can be und in Re. []. y (4 3. HO deign Apicatin A MOCO methd ha been cmeted r a deep ea HO which i hwn cnceptually in Figure 3. he hape, type prpulin, acent depth, preure hull tructure and material have been identiied. he deign pace thi vehicle ha been decribed in able. he cruie and peratin time c and the Rati R w are the bjective attribute. D max θ θ D max L a L m Fig.3 HO cniguratin able Deign variable HO general deign variable unit Baeline Lwer Bund Upper Bund 4
L m 4.45 4 5 m L m 3.0.8 3. a 3 D m 3.0.6 3. 4 u kn.5.5 3.0 5 kg 0 00 300 6 E kh 0 90 40 7 kg 6374 5500 7000 bu Firtly, we decmpe the deign prblem int a ytem mdule and ur diciinary ptimizatin mdule: gemetry & hydr, tructure, prpulin, weight & vlume. Figure 4 decribe the MOCO architecture the HO deign. Sytem ptimizer D, L, R, D, L, R,, u, E pr, pr,, bu,, pr,, pr J Diciine Gemetry & hydr J Diciine Structure J 3 Diciine 3 Prpulin J 4 Diciine 4 eight & vlume Fig.4 MOCO architecture the HO deign Equatin (8-0 decribe the rmulatin ytem-level ptimizatin prblem, and Equatin (- decribe the ur ubytem ptimizatin prblem. ( Sytem-level tand rmulatin: Min : { R, } (8 w c S.t. : J i ε, ( i =,,3,4 ; (9 0 D..: x = [ D, u, E, L, R,,,,,, ] ; (0 pr pr 5
( Diciine, Gemetry and hydr ubpace tand rmulatin: D L R = ; ( ( + ( + ( + ( D L R S.t. : 4 L m m 5 ;.8 L a 3. m ;.6 D 3. m ; ( D..: x = [ La, Lm, D] ; (3 (3 Diciine, tructure ubpace tand rmulatin: D L = (4 ( + ( + ( + ( + ( D L S.t. :.6 D 3. m ; (5 D..: x = [ D, L, ] ; (6 (4 Diciine 3, prpulin ubpace tand rmulatin: u E R = (7 pr pr 3 ( + ( + ( + ( + ( u E pr pr R S.t. :.5 u 3. 0kn ;90 E 40kh (8 D..: x = [ u, E, R ] ; (9 (5 Diciine 4, weight & vlume ubpace tand rmulatin: bu 4 = ( + ( + ( + ( + ( bu pr pr + ( + ( + ( (0 pr pr S.t. : 00 300 kg ;5500 7000kg ;40 ρ 80kN ; t ( bu D.. : x = [,,,,,, ] ; ( bu pr pr 5.. Reult and Dicuin A Multi-Objective Cllabrative Optimizatin ha been run r 38 generatin with a ppulatin 50 HO. In ytem-level ptimizatin prblem, the relaxatin actr cmpatible cntraint i et t 0.000, the crver prbability, the mutatin prbability and the maximum generatin are et t 0.9, 0. and 350. Fr the ub-pace ptimizatin prblem, the equence quadratic prgramming (SQP i ued t attain the diciine ptimizatin lutin. he dierent w 6
ub-pace ptimizatin are lved in-parallel. Reult are preented in Figure 5. Nne thee HO can be identiied a "the bet". Selectin the preerred deign i up t the deigner. Figure 5 prvide the deigner with imprtant inrmatin t make thi electin. able lit ur deign which i elected rm the Paret ptimal lutin Parat rnt 7.0 6.5 HO HO 6.0 c (h 5.5 5.0 4.5 HO3 HO4 4.0 0.00 0.05 0.030 0.035 0.040 0.045 0.050 R w Fig. 5 Paret ptimal lutin HO multi-bjective multidiciinary deign ab. he ur deign in Paret rnt. deig D /m L m /m L a /m u /k E / k h bu /k /k c / R w n n g g h HO.93 4.60.95.50 3.99 7000.00 65.70 7.00 0.03 HO.87 4.63.93.50 9.9 6845.00 300.00 6.46 0.04 HO3.8 5.07 3.08.5 0.47 6656.00 300.00 5.9 0.046 HO4.83 4.60.93.5 98.83 648.00 300.00 4.39 0.048 HO and HO4 are lcated at the end the Paret rnt, and HO and HO3 are lcated at the middle the Paret rnt. HO ha the lnget cruie and peratin time which i up t 7 hur, but the alternative ha the minimum rati which i jut 0.03. In cntrat, HO4 ha the maximum rati and the hrtet cruie and peratin time. I the time i the mt imprtant perrmance r a HO, the HO between HO and HO are excellent chice. 7
4. Summary and Cncluin An apicatin Multidiciinary Deign Optimizatin t a HO cnceptual deign ha been preented. In thi apicatin, the MOCO architecture wa ued. he methd integrate the multi-bjective ptimizatin methd within the cllabrative ptimizatin ramewrk, which remain the main metric CO architecture and ability PGA t eeking nn-inerir lutin et. S the methd i eective in that it make a chance t execute in-parallel r diciinary deign, and it i mre lexible in that it enable the deigner t elect the ittet lutin amng the Paret ptimal et in accrding with their preerence and the nature the deign prblem. hee practical advantage make the architecture well-uited r the deign HO. Reerence: [] Balling R J, Gale D L. Cllabrative Optimizatin Sytem Invlving Dicrete Deign at the Diciine Level [J]. Jurnal Mechanical Deign, 998(0:3-39. [] Sbiezczanki-Sbieki J, Hatka R. Multidiciinary Aerpace Deign Optimizatin: Survey Recent Develpment [J]. Structural Optimizatin, 997, 4(:-3. [3] Balling R J, Sbiezczanki Sbieki J. Optimizatin cued ytem: A critical verview apprache [J]. AIAA Jurnal, 996, 34(:6 7. [4] Braun R D, Mre A A, Kr I M. Cllabrative architecture r launch vehicle deign [J]. Jurnal Spacecrat and Rcket, 997, 34(4:478-486. [5] Sbieki I P, Kr I M. Cllabrative Optimizatin Apied t an Aircrat Deign Prblem[C]. AIAA Paper 96-075, the 34 th AIAA Aerpace Science Meeting and Exhibit. Ren, Nevada, January 5 8, 996. [6] Batill S M, Stelmark M A, Yu X Q. Multidiciinary deign ptimizatin an electric-pwered unmanned air vehicle [J]. Aircrat Deign, 999(:-8. [7] Kdiyalam S. Evaluatin methd r multidiciinary deign ptimizatin, Phae I [R]. ech. Reprt, NASA/CR-000-033, Natinal Aernautic and Space Adminitratin, 998. [8] Belegundu A D, Halber E, Yukih M A, Simpn. Attribute-baed multidiciinary ptimizatin underea vehicle [C]. AIAA Paper 000-4865, the 8 th AIAA/USAF/NASA/ISSMO Sympium n Multi- diciinary Analyi and Optimizatin. Lng Beach, CA, Sept. 6-8, 000. [9] McAlliter C D, Simpn, Kurtz P H, Yukih M. Multidiciinary deign ptimizatin tet baed n autnmu underwater vehicle deign[c]. he 9 th AIAA/ISSOM ympium n Multidiciinary Analyi and Optimizatin. Atlanta, Gergia, Sept. 4-6, 00. [0] Kdiyalam S, Sbiezczanki-Sbieki J. Bi-level integrated ytem ynthei with repne urace [J]. AIAA Jurnal, 000, 38(8: 485-497. [] Braun, R. D. Cllabrative Optimizatin: An architecture r large-cale ditributed deign [D]. Ph.D. thei, Stanrd Univerity, Department Aernautic and Atrnautic, 996. [] Kr I, and Manning, Cllabrative Optimizatin Statu and Directin [C]. AIAA Paper 000-47, the 8 th AIAA/USAF/NASA/ISSMO Sympium n Multidiciinary Analyi and Optimizatin. Lng Beach, CA, Sept. 6-8, 000. [3] DeMiguel A, Murray. An Analyi Cllabrative Optimizatin Methd[C]. AIAA Paper 000-470, the 8 th AIAA/USAF/NASA/ISSMO Sympium n Multidiciinary Analyi and Optimizatin. Lng Beach, CA, Sept. 6-8, 000. 8
[4] Braun R D, Gage P, Kr I M. Imementatin and perrmance iue in cllabrative ptimizatin[c]. AIAA-96-407, the 6 th AIAA/USAF/NASA/ISSMO Sympium n Multidiciinary Analyi and Optimizatin. ahingtn, September, 996. [5] Alexandrv N M, Lewi R M. Analytical and Cmputatinal Apect Cllabrative Optimizatin r Multidiciinary Deign [J]. AIAA Jurnal, 00, 40(:30-309. [6] Manning. High Speed Civil ranprt Deign via Cllabrative Optimizatin [D]. Ph.D. thei, Stanrd Univerity. 999. [7] Jun S, Jen Y, Rh J, D Lee. Apicatin Cllabrative Optimizatin Uing Repne Surace Methdlgy t an Aircrat ing Deign [C]. AIAA-004-444, the 0 th Multidiciinary Analyi and Optimizatin Cnerence. Albany, New Yrk, Aug. 30-, 004. AIAA/ISSMO [8] Rhl P J, He B, Finnigan P M. A cllabrative ptimizatin envirnment r turbine engine develpment[r]. AIAA Paper, N. 98-4734, 998. [9] Sbieki I P, Kr I M. Cllabrative Optimizatin uing Repne Surace Etimatin [J]. AIAA Jurnal, 000, 38 (0: 93-939. [0] appeta R, Renaud J E. Multibjective cllabrative ptimizatin [J]. Jurnal Mechanical Deign, 997, 9(3: 403-4. [] Ca An-xi, Cui ei-cheng. Multi-Objective cllabrative ptimizatin in multidiciinary deign r ubmerible [J]. Jurnal Ship Mechanic, l., N., 008.4. 9