scieee Open visual document viewer

The meccano method for simultaneous volume parametrization and mesh generation of complex solids

Montenegro, R.,Cascón Barbero, José Manuel,Escobar, J. M.,Rodriguez, E.,Montero, G.

Abstract

The meccano method is a novel and promising mesh generation method for simultaneously creating adaptive tetrahedral meshes and volume parametrizations of a complex solid. We highlight the fact that the method requires minimum user intervention and has a low computational cost. The method builds a 3-D triangulation of the solid as a deformation of an appropriate tetrahedral mesh of the meccano. The new mesh generator combines an automatic parametrization of surface triangulations, a local refinement algorithm for 3-D nested triangulations and a simultaneous untangling and smoothing procedure. At present, the procedure is fully automatic for a genus-zero solid. In this case, the meccano can be a single cube. The efficiency of the proposed technique is shown with several applications.

Full text

The meccano me hod o simul aneous olume pa ame iza ion and mesh gene a ion o complex solids R Mon eneg o1,JMCasc´on2,JMEscoba 1, E Rod ´ıguez1and G Mon e o1 1Ins i u e o In elligen Sys ems and Nume ical Applica ions in Enginee ing, Uni e si y o Las Palmas de G an Cana ia, Campus Uni e si a io de Tafi a, Las Palmas de G.C., Spain 2Depa men o Economics and Economic His o y, Facul y o Economics and Managemen , Uni e si y o Salamanca, Spain E-mail: { mon eneg o,jmescoba ,e od iguez,gmon e o}@siani.es, [email p o ec ed] Abs ac . The meccano me hod is a no el and p omising mesh gene a ion me hod o simul aneously c ea ing adap i e e ahed al meshes and olume pa ame iza ions o a complex solid. We highligh he ac ha he me hod equi es minimum use in e en ion and has a low compu a ional cos . The me hod builds a 3-D iangula ion o he solid as a de o ma ion o an app op ia e e ahed al mesh o he meccano. The new mesh gene a o combines an au oma ic pa ame iza ion o su ace iangula ions, a local efinemen algo i hm o 3-D nes ed iangula ions and a simul aneous un angling and smoo hing p ocedu e. A p esen , he p ocedu e is ully au oma ic o a genus-ze o solid. In his case, he meccano can be a single cube. The efficiency o he p oposed echnique is shown wi h se e al applica ions. 1. In oduc ion The au oma ic and adap i e mesh gene a ion is a c ucial aspec in fini e elemen applica ions [4, 16, 17, 27, 28, 29, 31]. Along he pas , he main objec i e has been o achie e high quali y adap i e meshes o complex solids wi h minimal use in e en ion and low compu a ional cos , bu he 3-D p oblem is s ill open [1]. I is well known ha mos mesh gene a o s a e based on Delaunay iangula ion and ad ancing on echnique, bu p oblems, ela ed o mesh quali y o mesh con o mi y wi h he solid bounda y, can s ill appea o complex geome ies. In addi ion, an app op ia e defini ion o elemen sizes is demanded o ob aining good quali y elemen s and mesh adap ion. Pa icula ly, local adap i e efinemen s a egies ha e been employed o mainly adap he mesh o singula i ies o nume ical solu ion. These adap i e me hods usually in ol e emeshing o nes ed efinemen . In his di ec ion, we ha e in oduced he new meccano echnique [2, 3, 24, 25] o cons uc ing adap i e e ahed al meshes o solids. We ha e gi en his name o he me hod because he p ocess s a s wi h he cons uc ion o a coa se app oxima ion o he solid, i.e. a meccano composed by connec ed polyhed al pieces (a pa icula case is when meccano is composed by connec ed cubes, i.e. a polycube.). The idea o he me hod is o build a olume mesh o he solid as a de o ma ion o an app op ia e e ahed al mesh o he meccano. WCCM/APCOM 2010 IOP Publishing IOP Con . Se ies: Ma e ials Science and Enginee ing 10 (2010) 012018 doi:10.1088/1757-899X/10/1/012018 c 2010 Published unde licence by IOP Publishing L d 1 The new au oma ic mesh gene a ion s a egy uses no Delaunay iangula ion, no ad ancing on echnique, and i simplifies he geome ical disc e iza ion p oblem o 3-D complex domains, whose su aces can be mapped o he meccano aces. The meccano me hod combines a local efinemen /de efinemen algo i hm o 3-D nes ed iangula ions [21], a pa ame e iza ion o su ace iangula ions [9] and ou simul aneous un angling and smoo hing p ocedu e [6]. A p esen , he meccano echnique has been implemen ed by using he local efinemen /de efinemen o Kossaczky [21], bu he idea could be implemen ed wi h o he ypes o local efinemen algo i hms [18]. We no e ha he esul ing adap i e e ahed al meshes wi h he meccano me hod ha e good quali y o fini e elemen applica ions. Ou app oach is based on he combina ion o se e al o me p ocedu es ( efinemen , mapping, un angling and smoo hing) which a e no in hemsel es new, bu he o e all in eg a ion is an o iginal con ibu ion. Many au ho s ha e used hem in diffe en ways. T iangula ions o con ex domains can be cons uc ed om a coa se mesh by using efinemen /p ojec ion [26]. Adap i e nes ed meshes ha e been cons uc ed wi h efinemen and de efinemen algo i hms o e olu ion p oblems [8]. Mappings be ween physical and pa ame ic spaces ha e been analyzed by se e al au ho s. Significan ad ances in su ace pa ame iza ion ha e been done in [9, 11, 12, 23, 30, 32], bu he olume pa ame iza ion is s ill open. Floa e e al [13] gi e a simple coun e example o show ha con ex combina ion mappings o e e ahed al meshes a e no necessa ily one- o- one. La ge domain de o ma ions can lead o se e e mesh dis o ions, especially in 3-D. Mesh op imiza ion is hus key o keeping mesh shape egula i y and o a oiding a cos ly emeshing [19, 20]. In adi ional mesh op imiza ion, mesh mo ing is guided by he minimiza ion o ce ain o e all unc ions, bu i is usually done in a local ashion. In gene al, his p ocedu e in ol es wo s eps [15, 14]: he fi s is o mesh un angling and he second one o mesh smoo hing. Each s ep leads o a diffe en objec i e unc ion. The meccano me hod uses he imp o emen p oposed by [6, 7], whe e a simul aneous un angling and smoo hing guided by he same objec i e unc ion is in oduced. Some ad an ages o he meccano me hod a e ha : solid su ace iangula ion is au oma ically cons uc ed, he final 3-D iangula ion is con o ming wi h he objec bounda y, inne su aces a e au oma ically p ese ed ( o example, in e ace be ween se e al ma e ials), node dis ibu ion is adap ed in acco dance wi h he objec geome y, and pa allel compu a ions can easily be de eloped o meshing he meccano pieces. Howe e , ou p ocedu e demands an au oma ic cons uc ion o he meccano and an admissible mapping be ween he meccano bounda y and he objec su ace mus be defined. In his pape , we conside applica ions o he meccano me hod o complex genus-ze o solid, i.e. a solid whose bounda y is a su ace ha is homeomo phic o he su ace o a sphe e. In his case, we assume ha he solid geome y is defined by a iangula ion o i s su ace and ha he use only has o fix a meccano composed by a single cube and a ole ance ha fixes he desi ed app oxima ion o he solid su ace. In o de o define an admissible mapping be ween he cube aces and pa ches o he ini ial su ace iangula ion o he solid, we ha e in oduced an au oma ic me hod o decompose he solid su ace iangula ion in o six pa ches ha p ese es he same opological connec ions han he cube aces. Then, a disc e e mapping om each su ace pa ch o he co esponding cube ace is cons uc ed by using he pa ame e iza ion o su ace iangula ions p oposed by M. Floa e in [9, 10, 11, 12]. The shape-p ese ing pa ame iza ions, which a e plana iangula ions on he cube aces, a e he solu ions o linea sys ems based on con ex combina ions. In he nea u u e, mo e effo should be made in de eloping an au oma ic cons uc ion o he meccano when he genus o he solid su ace is g ea e han ze o. Cu en ly, se e al au ho s a e wo king on his aspec in he con ex o polycube-maps, see o example [23, 30, 32]. They a e analyzing how o cons uc a polycube o a gene ic solid and, simul aneously, how o define a con o mal mapping be ween he polycube bounda y and he solid su ace. Al hough ha monic WCCM/APCOM 2010 IOP Publishing IOP Con . Se ies: Ma e ials Science and Enginee ing 10 (2010) 012018 doi:10.1088/1757-899X/10/1/012018 2 maps ha e been ex ensi ely s udied in he li e a u e o su ace pa ame e iza ion, only a ew wo ks a e ela ed o olume pa ame iza ion, o example a p ocedu e is p esen ed in [22]. In he ollowing Sec ion we p esen a b ie desc ip ion o he main s ages o he me hod o a gene ic meccano composed o polyhed al pieces. In Sec ion 3 we in oduce applica ions o he algo i hm in he case ha he meccano is o med by a simple cube. Finally, conclusions and u u e esea ch a e p esen ed in Sec ion 4. 2. The algo i hm o he meccano me hod The main s eps o he gene al meccano e ahed al mesh gene a ion algo i hm a e summa ized in his sec ion. A de ailed desc ip ion o he me hod can be analyzed in [2, 3, 24, 25]. The inpu da a a e he defini ion o he solid bounda y ( o example by a gi en su ace iangula ion) and a gi en ole ance (co esponding o he solid su ace app oxima ion). The ollowing algo i hm desc ibes he whole mesh gene a ion app oach. Meccano e ahed al mesh gene a ion algo i hm (i) Cons uc a meccano app oxima ion o he 3-D solid o med by polyhed al pieces. (ii) Define an admissible mapping be ween he meccano bounda y aces and he solid bounda y. (iii) Build a coa se e ahed al mesh o he meccano. (i ) Gene a e a local efined e ahed al mesh o he meccano, such ha he mapping o he meccano bounda y iangula ion app oxima es he solid bounda y o a gi en p ecision. ( ) Mo e he bounda y nodes o he meccano o he objec su ace wi h he mapping (ii). ( i) Reloca e he inne nodes o he meccano. ( ii) Op imize he e ahed al mesh wi h he simul aneous un angling and smoo hing p ocedu e. The fi s s ep o he p ocedu e is o cons uc a meccano app oxima ion by connec ing diffe en polyhed al pieces. Once he meccano app oxima ion is fixed, we ha e o define an admissible one- o-one mapping be ween he bounda y aces o he meccano and he bounda y o he objec . In hi d s ep, he meccano is decomposed in o a coa se and alid e ahed al mesh by an app op ia e subdi ision o i s ini ial polyhed al pieces. We con inue wi h a local efinemen s a egy o ob ain an adap ed mesh which can app oxima e he bounda ies o he domain wi hin a gi en p ecision. Then, we cons uc a mesh o he solid by mapping he bounda y nodes om he meccano aces o he ue solid su ace and by eloca ing he inne nodes a a easonable posi ion. A e hose wo s eps he esul ing mesh is angled, bu i has an admissible opology. Finally, a simul aneous un angling and smoo hing p ocedu e is applied and a alid adap i e e ahed al mesh o he objec is ob ained. We no e ha he gene al idea o he meccano echnique could be unde s ood as he connec ion o diffe en polyhed al pieces. So, he use o cuboid pieces, o a polycube meccano, a e pa icula cases. 3. Applica ions o he meccano me hod om a cube In his sec ion, we p esen he applica ion o he meccano algo i hm in he case o he solid su ace being genus-ze o and he meccano being o med by a single cube. We assume as da um a iangula ion o he solid su ace. We in oduce an au oma ic pa ame iza ion be ween he su ace iangula ion o he solid and he cube bounda y. To ha end, we au oma ically di ide he su ace iangula ion in o six pa ches, wi h he same opological connec ion ha cube aces, so ha each pa ch is mapped o a cube ace. These pa ame iza ions ha e been done wi h GoTools co e and pa ame iza ion modules om SINTEF ICT, a ailable in he websi e h p://www.sin e .no/ma h_so wa e. This code implemen s Floa e ’s pa ame iza ion in C++. In he ollowing applica ion we ha e used he mean alue me hod o he pa ame iza ion o he inne nodes o he pa ch iangula ion, and he bounda y nodes a e fixed wi h cho d leng h pa ame iza ion [9, 11]. WCCM/APCOM 2010 IOP Publishing IOP Con . Se ies: Ma e ials Science and Enginee ing 10 (2010) 012018 doi:10.1088/1757-899X/10/1/012018 3 We ha e implemen ed he meccano me hod by using he local efinemen o AL- BERTA. This code is an adap i e mul ile el fini e elemen oolbox de eloped in C, see h p://www.albe a- em.de. This so wa e can be used o sol e se e al ypes o 1-D, 2-D o 3-D p oblems. ALBERTA uses he Kossaczky efinemen algo i hm [21] and equi es an ini ial mesh opology [26]. The ecu si e efinemen algo i hm could no e mina e o gen- e al meshes. The meccano echnique cons uc s meshes ha e i y he imposed es ic ions o ALBERTA in ela ion o opology and s uc u e. The minimum quali y o efined meshes is unc ion o he ini ial mesh quali y. The pe o mance o ou no el e ahed al mesh gene a o is shown in he ollowing applica ions. The fi s co esponds o an a madillo and he second o a bone. We ha e ob ained a su ace iangula ion o hese objec s om in e ne . 3.1. Applica ion o an a madillo The o iginal su ace iangula ion o he A madillo has been ob ained om he S an o d Compu e G aphics Labo a o y, h p://g aphics.s an o d.edu/da a/3Dscan ep/,andis shown in Figu e 1. I has 30000 iangles and 15002 nodes. The bounding box o he solid is defined by he poin s (x, y, x)min =(−60,−50,−26) and (x, y, z)max =(68,66,90). (a) (b) Figu e 1. O iginal su ace iangula ion o A madillo, (a) ini ial subdi ision in connec ed sub iangula ions using Vo onoi diag am and (b) compa ible pa i ion {T i S}5 i=0 ob ained a e applying ou educ ion algo i hm, emo ing di iding edges and smoo hing he in e aces be ween pa ches. We conside a uni cube as meccano. I s cen e is placed inside he solid a he poin (7.5,17.5,55.5). We ob ain an ini ial subdi ision o A madillo su ace in ele en maximal connec ed sub iangula ions using he Vo onoi diag am associa ed wi h he cen e s o he cube aces, see Figu e 1(a). In o de o ge a compa ible decomposi ion o he su ace iangula ion, WCCM/APCOM 2010 IOP Publishing IOP Con . Se ies: Ma e ials Science and Enginee ing 10 (2010) 012018 doi:10.1088/1757-899X/10/1/012018 4 we use he i e a i e p ocedu e p oposed in [3] o educe he cu en ele en e ices o he g aph GS o six. Figu e 1(b) shows he esul ing compa ible pa i ion {T i S}5 i=0. (a) (b) Figu e 2. (a) Floa e ’s pa ame iza ion o {T i S}5 i=0 on co esponding cube aces, (b) cube e ahed al mesh ob ained by he meccano me hod a e efinemen . We map each su ace pa ch Σi S o he cube ace Σi Cby using he Floa e pa ame iza ion [9]. The defini ion o he one- o-one mapping be ween he cube and A madillo bounda ies is s aigh o wa d once he pa ame iza ion o he A madillo su ace iangula ion is buil , see Figu e 2(a). A de ail o he each pa ame iza ion o Ti Son he co esponding cube ace Σi Cis shown in Figu e 3. The cube is di ided in o six e ahed a. In o de o app oxima e he A madillo su ace wi h a ole ance ε2=0.1, see [3], we efine he cube mesh by applying 68 Kossaczky ecu si e bisec ions, see Figu es 2(b) and 5(a). This mesh con ains 144964 e ahed a and 33889 nodes, wi h 31316 iangles and 15660 nodes on i s bounda y. The mapping o he cube ex e nal nodes o he A madillo su ace p oduces a 3-D angled mesh wi h 10871 in e ed elemen s, see Figu e 4(a). The eloca ion o inne nodes by using olume pa ame iza ions educes he numbe o in e ed e ahed a o 186. We use he e ahed al mesh op imiza ion, p esen ed in [6], such ha he mesh is un angled in 6 i e a ions. The mesh quali y is imp o ed o a minimum alue o 0.08 and an a e age qκ=0.71 a e 6 smoo hing i e a ions. We no e ha he meccano echnique gene a es a high quali y e ahed al mesh, see Figu e 4(b): only 4 e ahed a ha e a quali y o less han 0.1, 150 less han 0.2 and 1046 less han 0.3. In o de o show he efficiency o he mesh op imiza ion echnique inside he A madillo, we display in Figu e 5 wo c oss sec ions be o e (b) and a e (c) i s applica ion. The loca ion o he cube can be obse ed in Figu e 5(b). We also no e ha , due o he high quali y su ace iangula ion ob ained wi h ou me hod, he mesh imp o emen is no significan i we p e iously apply he smoo hing su ace iangula ion algo i hm in oduced in [7]. The CPU ime o cons uc ing he final mesh o he A madillo is app oxima ely 112.78 seconds on a Dell p ecision 690, 2 Dual Co e Xeon p ocesso and 8 Gb RAM memo y. Mo e p ecisely, he CPU ime o each s ep o he meccano algo i hm is: 5.45 seconds o he subdi ision WCCM/APCOM 2010 IOP Publishing IOP Con . Se ies: Ma e ials Science and Enginee ing 10 (2010) 012018 doi:10.1088/1757-899X/10/1/012018 5 (a) (b) (c) (d) (e) ( ) Figu e 3. De ail o Floa e ’s pa ame iza ion o A madillo su ace sub iangula ions {T i S}5 i=0. o he ini ial su ace iangula ion in o six pa ches, 1.01 seconds o he Floa e pa ame iza ion, 33.11 seconds o he Kossaczky ecu si e bisec ions, 2.25 seconds o he ex e nal node mapping and inne node eloca ion, and 80.96 seconds o he mesh op imiza ion. WCCM/APCOM 2010 IOP Publishing IOP Con . Se ies: Ma e ials Science and Enginee ing 10 (2010) 012018 doi:10.1088/1757-899X/10/1/012018 6 (a) (b) Figu e 4. (a) Tangled mesh a e he mapping o he ex e nal nodes o he cube o he A madillo su ace and (b) esul ing alid mesh a e inne node eloca ion and mesh op imiza ion. (a) (b) (c) Figu e 5. C oss sec ions o he cube (a), and he A madillo e ahed al meshes be o e (b) and a e (c) he applica ion o he mesh op imiza ion p ocedu e. 3.2. Applica ion o a bone The o iginal su ace iangula ion o he Bone has been ob ained om h p://www-c.in ia. /gamma/download/a ichage.php?di =ANATOMY&name=ballJoin , and i can be ound in he CYBERWARE Ca alogue. This su ace mesh con ains 274120 iangles and 137062 nodes. WCCM/APCOM 2010 IOP Publishing IOP Con . Se ies: Ma e ials Science and Enginee ing 10 (2010) 012018 doi:10.1088/1757-899X/10/1/012018 7 S eps o he meccano echnique a e shown in Figu e 6. The esul ing mesh has 47824 e ahed a and 11525 nodes. This mesh has 11530 iangles and 5767 nodes on i s bounda y and i has been eached a e 23 Kossaczky efinemen s om he ini ial subdi ision o he cube in o six e ahed a. A angled e ahed a mesh wi h 1307 in e ed elemen s appea s a e he mapping o he cube ex e nal nodes o he bone su ace. The node eloca ion p ocess educes he numbe o in e ed e ahed a o 16. Finally, ou mesh op imiza ion algo i hm p oduces a (a) (b) (c) (d) Figu e 6. The meccano (a) and a c oss sec ions o he Bone e ahed al mesh be o e (b) and a e (c) he applica ion o he mesh op imiza ion p ocedu e. (d) Resul ing e ahed al mesh o he Bone ob ained by he meccano me hod. WCCM/APCOM 2010 IOP Publishing IOP Con . Se ies: Ma e ials Science and Enginee ing 10 (2010) 012018 doi:10.1088/1757-899X/10/1/012018 8 high quali y e ahed a mesh: he minimum mesh quali y is 0.15 and he a e age quali y is 0.64. We no e ha he meccano me hod uses o cons uc mo e han 1000 e ahed a pe second on a Dell p ecision 690, 2 Dual Co e Xeon p ocesso and 8 Gb RAM memo y. Finally, we summa ize a compa ison be ween ou me hod and s anda d e ahed al mesh gene a ion echniques [27, 28, 29]. On he one hand, one o he mos impo an con ibu ions o he meccano me hod is he esul ing olume pa ame iza ion o he solid. I can ha e in e es ing applica ions in solid modeling and nume ical simula ion. Fo example, he applica ion o isogeome ic analysis [1, 5] can be easie . On he o he hand, ou olume meshes can be u ilized in adap i e fini e elemen applica ions by using he Kossaczky’s algo i hm. The local efinemen s eps a e e y as because he sequence o solid meshes is defined om he coa se mesh o he meccano, i.e., he di iding edge o e ahed on bisec ion is known s aigh o wa d. In addi ion, we no e ha he minimum mesh quali y is bounded du ing all he mesh adap a ion p ocess, because simila elemen s appea a e h ee consecu i e bisec ions. Fo a gi en solid su ace iangula ion, we ha e checked ha a cons ained Delaunay e ahed aliza ion [29] can be as e han ou me hod, bu he esul ing meshes ha e lowe quali y. I he admissible minimum quali y is inc eased, many poin s can be added and he numbe o e ahed a inc eases o each he objec i e. I only a con o ming Delaunay e ahed aliza ion is desi ed, he numbe o e ahed a can be much highe han in ou me hod. A g ea ad an age o ou me hod is ha he adap i e node dis ibu ion on he bounda y and he inne o he solid is mo e s uc u ed, because he posi ions a e fixed wi h a sequence o nes ed meshes. In he case o ad ancing on [27], he esul ing mesh depends on he quali y o he gi en su ace iangula ion. We ha e seen ha he op imiza ion o his iangula ion, changing he mesh opology, can be oo cos ly. 4. Conclusions and u u e esea ch The meccano echnique is a e y efficien adap i e e ahed al mesh gene a o o solids whose bounda y is a su ace o genus ze o. We ema k ha he me hod equi es minimum use in e en ion and has a low compu a ional cos . The p ocedu e is ully au oma ic and i is only defined by a su ace iangula ion o he solid, a cube and a ole ance ha fixes he desi ed app oxima ion o he solid su ace. A c ucial consequence o he new mesh gene a ion echnique is he esul ing disc e e pa ame iza ion o a complex olume (solid) o a simple cube (meccano). We ha e in oduced an au oma ic pa i ion o he gi en solid su ace iangula ion o fixing an admissible mapping be ween he cube aces and he solid su ace pa ches, such ha each cube ace is he pa ame ic space o i s co esponding pa ch. The main ideas p esen ed in his pape can be applied o cons uc ing e ahed al o hexahed al meshes o complex solids. In u u e wo ks, he meccano echnique can be ex ended o meshing a complex solid whose bounda y is a su ace o genus g ea e han ze o. In his case, he meccano can be a polycube o cons uc ed by polyhed al pieces wi h compa ible connec ions. A p esen , he use has o define he meccano associa ed o he solid, bu we a e implemen ing a special CAD package o mo e gene al inpu solid. Acknowledgmen s This wo k was pa ially suppo ed by he “Sec e a ´ıa de Es ado de In es igaci´on” o Spanish Go e nmen , “Minis e io de Ciencia e Inno aci´on”, and FEDER, g an con ac s: CGL2008- 06003-C03 and UNLP08-3E-010. Re e ences [1] Bazile s Y, Calo V M, Co ell J A, E ans J, Hughes T J R, Lip on S, Sco M A and Sede be g T W 2008 Isogeome ic analysis: Towa d unifica ion o compu e aided design and fini e elemen analysis T ends in Enginee ing Compu a ional Technology (S i ling: Saxe-Cobu g Publica ions) chap e 1 pp 1–16 WCCM/APCOM 2010 IOP Publishing IOP Con . Se ies: Ma e ials Science and Enginee ing 10 (2010) 012018 doi:10.1088/1757-899X/10/1/012018 9