Construction process optimisation - review of methods, tools and applications

G ađe ina 7/2018 593 GRAĐEVINAR 70 (2018) 7, 593-606 DOI: h ps://doi.o g/10.14256/JCE.1719.2016 Cons uc ion p ocess op imisa ion – e iew o me hods, ools and applica ions P imljen / Recei ed: 7.7.2016. Isp a ljen / Co ec ed: 24.11.2017. P ih aćen / Accep ed: 12.4.2018. Dos upno online / A ailable online: 10.8.2018. Au ho s: Václa Venk bec, MSc. CE B no Uni e si y o Technology Facul y o Ci il Enginee ing Ins i u e o Technology, Mechaniza ion and Cons uc ion Managemen Czech Republic [email p o ec ed] Assis .P o . Ma io Galić, PhD. CE Josip Ju aj S ossmaye Uni e si y o Osijek Facul y o Ci il Enginee ing Osijek Depa men o O ganiza ion, Technology and Managemen [email p o ec ed] Assoc. P o .U oš Klanšek, PhD. CE Uni e si y o Ma ibo , Slo enia Facul y o Ci il Enginee ing, T anspo a ion Enginee ing and A chi ec u e Chai o Cons uc ion Managemen , Technology and Economics [email p o ec ed] Subjec e iew Václa Venk bec, Ma io Galić, U oš Klanšek Cons uc ion p ocess op imisa ion – e iew o me hods, ools and applica ions A e iew o cu en heu is ic echniques and ma hema ical p og amming me hods is p esen ed in he pape . Mode n op imisa ion modelling ools a e p esen ed, common cons uc ion op imisa ion p oblems a e desc ibed, and an o e iew o hei ecen applica ion is gi en. I is also s essed ha he e is ample oom o u he esea ch on ac i e BIM-based op imisa ion models, aimed a be e managemen o cons uc ion p ocesses. Accele a ed de elopmen o decision-making models, which combine op imisa ion me hods and in o ma ion sys ems, can be clea ly p edic ed o he nea u u e. Key wo ds: BIM, cons uc ion p oduc ion, modelling, op imisa ion me hods, p ocess op imisa ion P egledni ad Václa Venk bec, Ma io Galić, U oš Klanšek Op imizacija g ađe inskih p ocesa – me ode, ala i i p imjena U adu je dan p egled su emenih heu is ičkih ehnika i me oda ma ema ičkog p og ami anja. P eds a ljeni su mode ni ala i za op imizacijsko modeli anje, izlažu se najčešći p oblemi op imizacije u g ađe ina s u i daje p egled njiho ih neda nih aplikacija. Rad je is aknuo p os o is aži anja na ak i nim modelima op imizacije, koji su pod žani BIM-om s namje om boljeg up a ljanja g ađe inskim p ocesima. Ub zani az oj modela odluči anja, koji kombini aju op imizacijske me ode i in o macijske sus a e, jasno se može p ed idje i za blisku budućnos . Ključne iječi: BIM, g ađe inska p oiz odnja, modeli anje, op imizacijske me ode, op imizacija p ocesa Übe sich sa bei Václa Venk bec, Ma io Galić, U oš Klanšek Op imie ung on Baup ozessen – Me hoden, We kzeuge und Anwendung In de Abhandlung wi d eine Übe sich übe die mode nen heu is ischen Techniken und Me hoden de ma hema ischen P og ammie ung da geleg . Vo ges ell we den mode ne We kzeuge ü die op imie e Modellie ung, es we den die häu igs en P obleme de Op imie ung im Bauwesen da geleg und es wi d eine Übe sich ih e ak uellen Anwendungen gegeben. Die Abhandlung be on den Un e suchungsbe eich ak i e Modelle de Op imie ung, die du ch BIM un e s ü z we den, mi dem Ziel eines besse en Baup ozessmanagemen s. Eine beschleunig e En wicklung de En scheidungsmodelle, welche die Op imie ungsme hode und die In o ma ionssys eme kombinie en, kann ü die nahe Zukun kla o ausgesehen we den. Schlüsselwö e : BIM, Baup oduk ion, Modellie ung, Op imie ungsme hode, Op imie ungsp ozess G ađe ina 7/2018 594 GRAĐEVINAR 70 (2018) 7, 593-606 Václa Venk bec, Ma io Galić, U oš Klanšek 1. In oduc ion Op imisa ion has been p esen o qui e a long ime in indus ial p oduc ion. Howe e , i can be no ed ha op imisa ion has once again become he subjec o highly in ensi e wo ldwide esea ch, ollowing a huge heo e ical g ow h o his ield in he pe iod be ween 1950s and 1970s. Reasons why his ield o science is s ill so p opulsi e can be ound in an inc eased common awa eness abou limi ed a ailabili y o esou ces, wha happens o be he main igge ha pushed he op imisa ion o wa d in he lis o public’s c i e ia. Addi ional mo i a o s a e wide a ailabili y o compu a ional packages and s eadily g owing ha dwa e capaci ies. Hence, he almos o go en ma hema ical models a e now being e ised, expanded, modi ied and e i ied h ough hei use in sol ing ac ual op imisa ion p oblems in indus ial p oduc ion. The quan i a i e suppo o decision making p ocesses in indus y is especially highligh ed as being o c ucial signi icance in mode n managemen ends. Howe e , he speci ici y o managing cons uc ion p ojec s is hei s ochas ic and dynamic en i onmen , despi e common epe i i e ope a ions, in which op imisa ion cons i u es one o he phases. A his poin , he low o sui able in o ma ion is highly impo an o pe o ming op imisa ion in cons uc ion p ocesses and, he e o e, i is necessa y o iden i y which o he a ailable solu ion echniques is he mos app op ia e o a pa icula p oblem. Al hough cons uc ion indus y has always been p ojec -o ien ed [1], as well as a ec ed by ine iciency and ine ec i eness [2], limi ed a en ion has been paid in p ac ice o ma hema ically- based app oaches o p ocesses op imisa ion. The e o e, be e awa eness abou con empo a y echniques o p ocesses op imisa ion can be gene ally bene icial o all decision-make s in cons uc ion indus y, especially o p ojec manage s and expe s o p oduc ion p epa a ion. I is wo h men ioning ha imp o emen o business p ocesses can also be achie ed in indus y h ough en e p ise esou ce planning (ERP) sys ems. Va ious ERP sys ems o manu ac u ing and ading companies a e known and some o hem ha e been suppo ed by a ionalized wo k lows [3] and schedules [4]. Subjec e iews ha e also been p esen ed in li e a u e [5] and some au ho s, e.g. [6], in oduced a sys ema ic app oach o ERP implemen a ion in cons uc ion sec o . I has o en been no ed ha ERP sys ems can be good ehicles o business p ocess eenginee ing and in oduc ion o IT sys ems [7]. Ne e heless, his a icle does no a emp o p o ide a e iew o in eg a ion and in e ope abili y be ween di e en in o ma ion sys ems, e.g. ERP and building in o ma ion modelling (BIM), since li e a u e on ha subjec can be ound in ele an da abases, and some ecen con ibu ions ha e e en indica ed ha he e s ill exis s a as sca ci y o knowledge in he men ioned ield, see e e ence Howe e , BIM is s eadily being in oduced in cons uc ion indus y, and i is necessa y o e iew cu en s a e in he sphe e o p ocess op imisa ion, since bo h concep s can be ope a i ely connec ed in o de o imp o e business pe o mance. To he bes o ou knowledge, such a e iew has no as ye been made. Au ho s [8] s a ed ha BIM has e ol ed and ha The cons uc ion communi y is seeing a shi om he 3D o isualiza ion aspec o BIM o wo k low-speci ic ools ha a e being di ec ly applied o sol e eal-wo ld p oblems, such as ins alla ion e i ica ion, sequencing, and es ima ing. The indus y dialogue is now mo ing o a gene al ques ioning o how we op imize he e ec i e cap u e, analysis, and dissemina ion o in o ma ion in eal ime o make p ojec s mo e success ul. The e o e, his pape gi es an o e iew o ecen achie emen s in he men ioned ields, and p o ides some indings ha a e in ended o ill gaps in li e a u e and open possibili ies o u he esea ch. Conside ing he g ea a ie y o p oblems encoun e ed, and in he ligh o p e ious li e a u e e iews conce ning speci ic solu ion me hods and modelling ools, he au ho s p esen a e iew o hose conce ning he op imisa ion o cons uc ion p oduc ion in o de o p o ide he expe and academic communi ies wi h cu en in o ma ion ela ing o hese a eas. 2. Op imisa ion me hods Op imisa ion me hods can be used o sol ing a wide ange o di e en enginee ing p oblems. E en hough op imisa ion p oblems may come om a ious ields and di e en sys ems, hey can basically be o mula ed in a su p isingly simila way. Gene ally, an op imisa ion p oblem (OP) can be exp essed in he ollowing way: minimize (x), subjec o: h(x) = 0 and g(x) ≤ 0 whe e (x) is he objec i e unc ion o be minimized o e he ec o o decision a iables x, while h(x) = 0 ep esen s equali y cons ain s and g(x) ≤ 0 deno es inequali y cons ain s. The objec i e unc ion de ines he c i e ion o selec ing an op imum solu ion, while he cons ain s de e mine bounda ies delimi ing he space o all easible solu ions. I should be no ed he e ha he objec i e unc ion may also be maximized o e he easible space i necessa y. As o decision a iables, hey a e usually calcula ed be ween hei lowe and uppe bounds, xLO ≤ x ≤ xUP, and hey can be con inuous, x R, whe e R is he se o eal numbe s; o in ege , x Z, whe e Z is he se o in ege s. In ege a iables can also appea as bina y decision a iables, i.e. x {0,1}m. An app op ia e me hod o sol ing a speci ic op imisa ion p oblem should be selec ed a en i ely in o de o ob ain aluable esul s. Be o e selec ing a me hod o solu ion, he op imisa ion p oblem should be analysed om he s andpoin o i s unc ions, cons ain s, and decision a iables. Solu ion app oaches o single-c i e ia op imisa ion p oblems can be oughly di ided in o wo main g oups, i.e. heu is ic me hods and ma hema ical p og amming me hods. Heu is ic echniques can be used o sol e a wide a ie y o op imisa ion asks and hei main ad an age is ha mos o G ađe ina 7/2018 595 GRAĐEVINAR 70 (2018) 7, 593-606 Cons uc ion p ocess op imisa ion – e iew o me hods, ools and applica ions hese echniques con e ge easonably as , and can handle p oblems ha con ain non-di e en iable unc ions. Howe e , a he end o he sea ch, heu is ic algo i hms o en o e only app oxima ely op imal, i.e. sub-op imal solu ions. Ne e heless, heu is ic me hods ha e been p o en sui able o sol ing a a ie y o op imisa ion p oblems in ci il enginee ing, and he mos equen ly used ones a e: di ec sea ch (DS) [9], e olu ion s a egies (ES) [10, 11] and gene ic algo i hms (GA) [12,13], abu sea ch (TS) [14], simula ed annealing (SA) [15], neu al ne wo ks (NN) [16], an colony op imisa ion (ACO) [17], pa icle swa m op imisa ion (PSO) [18], di e en ial e olu ion (DE) [19], and ha mony sea ch (HS) [20]. As ex ensions o main heu is ic me hods, he me a- and hype -heu is ic s ochas ic echniques a e mos ly inclined o bio-inspi ed compu ing algo i hms like bac e ial o aging op imisa ion (BFO) [21], cuckoo sea ch (CS) [22], a i icial bee colony (ABC) [23], i e ly algo i hm (FA) [24], ba algo i hm (BA) [25], lowe pollina ion algo i hm (FPA) [26], a i icial plan op imisa ion (APO) [27], wol sea ch algo i hm (WSA) [28], e c. I should be no ed he e ha a ious hyb id combina ions o he a o esaid echniques ha e also been epo ed in li e a u e. Ma hema ical p og amming me hods ha e also been widely ecognized as ad an ageous ools o op imisa ion in ci il enginee ing (examples suppo ed wi h e e ences will be gi en in sec ions 4 and 5). The majo bene i o ma hema ical p og amming me hods is ha hey p o ide an expec ed exac op imum esul , al hough he sea ch p ocess i sel can, in some cases, equi e longe amoun o ime. As is commonly known, slowe con e gence o esul s is ypical in he case o op imisa ion o highly combina o ial, disc e e, nonlinea and, pa icula ly, noncon ex p oblems. In gene al, he ield o ma hema ical p og amming co e s linea p og amming (LP), nonlinea p og amming (NLP), mixed-in ege linea p og amming (MILP) and mixed-in ege nonlinea p og amming (MINLP) me hods. LP is a well de eloped a ea o ma hema ical p og amming. Solu ion echniques o LP can gene ally be applied o a ious op imisa ion p oblems, which include linea objec i e unc ions and cons ain s wi h con inuous a iables. Well known simplex algo i hm [29] is p esumably he mos commonly used me hod o sol ing LP p oblems. Howe e , i should be emphasized ha he in e io poin me hod [30] is o en ound o be mo e e icien when i comes o la ge-scale LP asks. When he con inuous op imisa ion p oblem con ains nonlinea unc ions, NLP app oach is equi ed o each he solu ion. Many enginee ing asks can be ansla ed in o NLP p oblems. A sui able me hod o sol ing a pa icula NLP p oblem should be selec ed by conside ing i s size and cha ac e is ics o nonlinea i y. Cu en ly, he e is no s anda d me hod ha can equally well sol e all ypes o NLP p oblems. Howe e , as he e a e many e icien algo i hms, an app op ia e one can be selec ed o sol e a pa icula NLP p oblem. A his poin , he gene alized educed g adien me hod [31], augmen ed Lag angian me hod [32, 33], and successi e quad a ic p og amming [34], can p obably be men ioned as he mos equen ly used NLP me hods. MILP is an ex ension o he LP app oach. While LP algo i hms can be used o sol e con inuous linea p oblems, MILP me hods can handle linea op imisa ion asks wi h con inuous and disc e e decision a iables. In MILP p oblems, disc e e a iables can be decla ed as in ege a iables o as bina y 0-1 a iables. S anda d me hod o MILP op imisa ion is he b anch and bound me hod [35] al hough he cu ing plane me hod [36], as well as he b anch and cu me hod [37], can also be e icien ly applied in a numbe o cases. MINLP me hods a e necessa y o disc e e op imisa ion p oblems ha include nonlinea e ms in hei o mula ion. Al hough high quali y exac solu ions can be expec ed om he s a e-o - he-a MINLP me hods, i should be no ed he e ha he ield o nonlinea (disc e e) op imisa ion is higly complex and ha i has no ye eached he le el o ma u i y a ained by linea (con inuous) op imisa ion. Ne e heless, a numbe o e icien MINLP me hods a e a ailable o sol ing nonlinea disc e e op imisa ion p oblems in ci il enginee ing, such as he gene alized Bende s decomposi ion [38], nonlinea b anch and bound me hod [39], easibili y echnique [40], sequen ial linea disc e e p og amming [41], ex ended cu ing plane me hod [42], augmen ed penal y/ou e -app oxima ion/ equali y- elaxa ion algo i hm [43], b anch and educe me hod [44], mixed-in ege alpha BB algo i hm [45], hyb id algo i hm [46], and o he s. 3. Op imisa ion modelling ools A a ie y o comme cial so wa e o compu e -based modelling o op imisa ion p oblems a e cu en ly a ailable. Syn axes o algeb aic modelling languages a e o en lexible and allow o mula ion o la ge-scale compac op imisa ion models ia indexing. In his way, algeb aic modelling languages, such as AIMMS [47], AMPL [48], CAMPS [49], GAMS [50], LINGO [51], LPL [52], MPL [53], OPL [54] and UIMP [55], may be applied o complex and unique op imisa ion p oblems, which may need se e al e isions be o e an accu a e model is es ablished. They a e pa icula ly usable in case o conside able numbe o cons ain s o he same ype ha ollow a simila pa e n. In pa icula , he algeb aic modelling language can simul aneously o mula e all cons ain s o he same ype by simul aneously handling decision a iables o each ype. The applica ion o he algeb aic modelling language accele a es nume ous model managemen ac i i ies, such as ans o ming he da a in o model pa ame e s, modi ying he model, accessing he da a, and analysing model esul s. Sp eadshee s a e also e y popula ools o modelling op imisa ion p oblems. Mic oso ’s compu e package Excel (Mic oso Visual Basic), wi h add-ins like E ol e [56], Sol e [57], o Wha ’sBes [58], as well as WinQSB [59], can be de ined as sp eadshee based compu e p og ams ha a e o en used o modelling op imisa ion p oblems. Sp eadshee s a e use - iendly ools and enable op imisa ion model o mula ion wi hin a amilia so wa e en i onmen . On he o he hand, op imisa ion models gene a ed by sp eadshee so wa e a e less anspa en han G ađe ina 7/2018 596 GRAĐEVINAR 70 (2018) 7, 593-606 Václa Venk bec, Ma io Galić, U oš Klanšek he ones o mula ed using he algeb aic modelling languages. Tabula en y o model en i ies is also mo e ime consuming, and so he sp eadshee -o ien ed modelling so wa e is no mally used o c ea ing small- and medium-sized models wi h a easonable numbe o pa ame e s. The e a e also specialized s and-alone so wa e p og ams including in e alia Gu obi [60], TORA [61] and LIONsol e [62]. In e ac i e compu e languages o nume ic and symbolic calcula ions can also be applied o op imisa ion modelling. So wa e packages such as Ma hema ica [63] and Ma lab [64] can also be men ioned as in e ac i e compu e languages ha a e widely used o op imisa ion modelling, al hough ha is no hei main pu pose. Bo h o hese so wa e packages p o ide an in e ac i e en i onmen o da a s uc u ing, and o modelling and sol ing op imisa ion p oblems, while also enabling p esen a ion o esul s in he o m o compu e p in ou s and cha s. Clea signs ha e ecen ly eme ged abou he g owing in e es among esea che s in he use o in e ne applica ions o op imisa ion. One o he ways o pe o ming op imisa ion o e he in e ne is o use open p og ams loca ed a he NEOS se e [65]. A his se e , i is also possible o ind links o many comme cial op imisa ion p og ams, manuals o hei use, es esul s, and publica ions/jou nals wi h ecen op imisa ion esea ch da a, e c. 4. Use o op imisa ion in cons uc ion p ocesses 4.1. Gene al o ms o op imisa ion p oblems common o cons uc ion p ocesses This sec ion p o ides a b ie o e iew o o igins and de ini ions o some well-known op imisa ion p oblems ha can be ecognized as mos common in cons uc ion p ocesses. Thei o iginal name, basic desc ip ion and objec i es, as well as hei o igin, a e shown in alphabe ical o de in Table 1. Op imisa ion p oblems a e g ouped by hei main op imisa ion objec i es, and hey o m wo basic domains o asks common o cons uc ion p ocesses: a) op imisa ion o esou ces b) op imisa ion o layou and ou e. Domain Op imisa ion p oblem P oblem desc ip ion O igin Resou ce o ien ed op imisa ion p oblems Assignmen p oblem (AP). AP deals wi h he ques ion o how o assign a gi en numbe o asks o a gi en numbe o agen s, whe e each agen can pe o m any asks and, a ha , his agen c ea es a cos ha changes depending on he ask, and all asks mus be comple ed, p o ided ha only one agen can be assigned a single ask, and he o al cos s mus be minimised. [66] P ojec scheduling p oblem (PSP). PSP ep esen s a wide se o p oblems, which a e commonly summa ized by one dominan objec i e unc ion o minimiza ion o o al p ojec cos s. I expanded wi h objec i e unc ions o minimiza ion o p ojec ime and cos s, hey a e known as ime-cos ade-o p oblems (TCTO), and, i including also he op imisa ion o o he esou ces, hey a e iden i ied as ime-cos - esou ce op imisa ion p oblems (TCRO). [67] T anspo p oblem (TP). TP is de ined as a p og am o sol ing anspo o goods om mul iple sou ces o mul iple des ina ions wi h an objec i e unc ion aimed a minimising anspo cos s. TP is gene ally a ma e ial ne wo k- low op imisa ion p oblem. F om he iewpoin o TP o mula ion, he anspo cos s, supply and demand quan i ies, a e o en inpu pa ame e s while anspo ing lows ep esen decision a iables. [68] Layou and ou e o ien ed op imisa ion p oblems A c ou ing p oblem (ARP). ARP is a connec ed g aph cons uc ed o wo se s o poin s (o igins and des ina ions) o which i is necessa y o ind he closed ou e ha isi s each des ina ion poin a leas once, o o de e mine ha such a ou e does no exis . [69, 70] Capaci a ed a c ou ing p oblem (CARP). CARP has he objec i e o ind a numbe o ou es such ha each a c wi h posi i e demand is se iced by exac ly one ehicle and, a ha , he sum o demand o hose a cs, se iced by each ehicle, should no exceed gi en capaci y and he o al cos o he ou es is minimized. [71] Chinese pos man p oblem (CPP). CPP is de ined by he connec ed undi ec ed g aph wi h he known dis ance ma ix, and he p oblem is o ind a ou e ha passes h ough each poin o he g aph a leas once in he sho es possible way. [72] T a eling salesman p oblem (TSP). The TSP is de ined as an op imisa ion p oblem o he salesman who needs o a el om home loca ion o each loca ion speci ied on he lis and, a e execu ing all isi s, o e u n o home loca ion while aking in o accoun he objec i e o he sho es o al ou e o he minimum o al a el ime. [73] Vehicle ou ing p oblem (VRP). VRP is de ined as a p oblem o how o op imally ou e a lee o iden ical ehicles om a cen al poin o supply des ina ions wi h known demands subjec o ehicle capaci y cons ain s. [74] Table 1. Op imisa ion p oblems common o cons uc ion p ocesses G ađe ina 7/2018 597 GRAĐEVINAR 70 (2018) 7, 593-606 Cons uc ion p ocess op imisa ion – e iew o me hods, ools and applica ions Table 2. Resou ce o ien ed op imisa ion p oblems applied in cons uc ion Pe iod Desc ip ion o applied p oblem Me hods, models and ools Main indings and conclusions O igin 1997 - 2000 TCTO p oblem o an eigh een- ac i i y cons uc ion p ojec . Au ho s de eloped a new algo i hm by combining gene ic algo i hm (GA) and Pa e o on app oach. The p esen ed new algo i hm has p o en o be e icien and accu a e in sol ing he add essed p oblem by sea ching only a small ac ion o he o al sea ch space. Fu he mo e, he au ho s de eloped a compu e p og am TCGA ha uses he MS Excel p og am, and au oma es he execu ion o he p oposed algo i hm. [75] TCTO p oblem in ol ing gene a ion o six een cons uc ion p ojec s each consis ing o ele en ac i i ies. Machine lea ning gene ic algo i hm sys em (MLGAS). P e ious analy ical echniques we e known o limi he usage o GAs as i was necessa y o manually en e da a o he ime-cos cu es and his in linea o m only. MLGAS o e comes hese limi a ions by inco po a ing machine lea ning ia GAs. [76] Mul i-objec i e op imisa ion o esou ce alloca ion and le elling based on a case s udy in ol ing wen y ac i i ies and six esou ces. Gene ic algo i hm was applied by means o MS Visual Basic (VBA) p og amming language. Main con ibu ions o he men ioned app oach a e: e ec i e imp o emen o esou ce alloca ion heu is ics using andom ac i i y p io i ies; p ac ical modi ica ion o esou ce le elling heu is ics using a double-momen app oach; and mul i-objec i e op imisa ion o bo h esou ce alloca ion and le elling using GAs. [77] Resou ce le elling p oblem on a nine-ac i i y cons uc ion p ojec wi h mul iple esou ces, aimed a minimising esou ce u iliza ion a ia ion wi hin a ixed du a ion p ojec . GA (GARLS) based esou ce le elling sys em. GARLS model does no necessa ily need o commi o any speci ic heu is ic ule; hence, i is mo e lexible o sol ing complex le elling and scheduling p oblems. GARLS p o ides se e al easible o nea -op imal solu ions ha may assis in p ojec decision-making. In his s udy, he au ho s success ully used a a ie y o so wa e p og ams o GARLS implemen a ion (i.e. MS P ojec , MS Access, MS Excel, and VBA). As a ecommenda ion o u he de elopmen , he au ho s sugges modi ica ion o he p oposed model o include TCTO p oblem sol ing as well. [78] 2001 – 2010 Repe i i e scheduling p oblem wi h sha eable esou ce cons ain s in case o p ecas p oduc ion. GA-based esou ce-cons ained epe i i e scheduling model, using he VBA p og amming language. The au ho s poin ou ha hei model inco po a es an e icien compu a ional echnique o esou ce alloca ion and a mo e sui able way o modelling he esou ce sha ing in epe i i e scheduling, compa ed o he linea scheduling echnique. GA-based model does no ha e o commi o any pa icula heu is ic ules and hus is mo e lexible. In addi ion, he GA-based epe i i e scheduling model can explo e and use se e al nea -op imal solu ions, which a e no mally no a ailable using con en ional epe i i e scheduling echniques. [79] Repe i i e scheduling p oblem o a p ojec consis ing o ou simila sec ions o uni s, and each includes epe i i e ac i i ies wi h inish o s a connec ions wi hou ime lags. Au oma ed model based on dynamic p og amming o mula ion. I was concluded ha , since he model is au oma ed, i alle ia es he need o he use o p o ide a se o in e up ion ec o s in an a bi a y manne p io o scheduling. In addi ion, i signi ican ly educes he numbe o in e up ion ec o s in a a ional manne , making he op imisa ion p ocess easible; i also enables gene a ion o an op imum solu ion. [80] Resou ce scheduling and TCTO p oblem o a wel e-ac i i y cons uc ion p ojec wi h he assump ion ha he e a e no limi a ions on p ecedence ela ionships be ween succeeding ac i i ies Resou ce le elling using he augmen ed Lag angian GA model. The sugges ed model allows any linea o nonlinea unc ion o he p esen a ion o cos -du a ion and esou ce du a ion ela ionships. The model can also handle a wide ange o p ojec sizes including la ge cons uc ion p ojec s in ol ing a la ge numbe o ac i i ies. The model was e icien ly implemen ed in FORTRAN p og amming language in o de o sol e esou ce scheduling p oblems on se e al cons uc ion p ojec s. [81] Mul i-objec i e TCTO p oblem ha akes in o accoun adap i e weigh s among op imisa ion c i e ia on a case s udy consis ing o se en ac i i ies. GA based model wi h a modi ied adap i e weigh app oach. Au ho s no e ha his app oach p o ides GAs wi h g ea e eedom du ing sea ch in he mul i-objec i e space, which o e comes d awbacks o he single objec i e TCTO. Howe e , because he model uses GAs in e ms o a sea ch engine, i s andomness could a ec eliabili y o esul s. [82] Resou ce-cons ained PSP (RCPSP) aimed a minimizing p ojec du a ion. Me aheu is ic PSO based me hod. Based on compu a ional analyses, au ho s concluded ha he pe mu a ion-based PSO ou pe o ms he p io i y-based PSO, and ha he PSO-based me hodology has good pe o mance, which is compa able o o he me aheu is ic me hods such as GA and SA in sol ing he RCPSP. [83] S ochas ic mul i-objec i e TCRO p oblem wi h he ime and cos a iables conside ed as uzzy. S ochas ic mul i-objec i e op imisa ion model based on he use o he non-domina ed so ing GAs (NSGA-II). This model imp o es he p e iously de eloped weigh ing app oaches by p o iding a h ee-dimensional Pa e o on . I also adop s FSs o inco po a e he unce ain ies in ime and di ec cos s o p ojec ac i i ies. The model can also conside di e en le els o unce ain y by changing he α-cu le els. Au ho s sugges ed ha u he de elopmen o he model could be ocused on p o iding he capabili y o spli ing ac i i ies in he model. [84] PSP in case o mul iple shi s on cons uc ion p ojec s, he aim being o minimize p ojec du a ion, cos , and nega i e impac s o e ening and nigh shi s. Op imisa ion model consis s o h ee modules: i) ini ializa ion module o scheduling op imisa ion compu a ions; ii) scheduling module; and iii) mul i-objec i e GA module The model was p o en capable o e alua ing and iden i ying op imum shi sys ems o p ojec s and his al eady in he i s i e a ion. In addi ion, he model o e s an op imum solu ion wi h minimisa ion o p ojec ime and cos s, and wi h app op ia e dis ibu ion o wo ke s in o e ening and nigh shi s, whe e e e y solu ion iden i ies an op imum schedule and mul iple shi wo k plan o each ac i i y. The model also gene a es op imum plans o dis ibu ing labou among compe ing shi s o minimize nega i e impac s o labou cons ain s on p ojec pe o mance. [85] G ađe ina 7/2018 598 GRAĐEVINAR 70 (2018) 7, 593-606 Václa Venk bec, Ma io Galić, U oš Klanšek He e i should be no ed ha he e a e many o he a ia ions o op imisa ion p oblems, wi h a ious objec i es and cons ain s, ha would be wo hy o conside a ion. Howe e , hey a e no p esen ed he e due o he limi ed space a ailable o his pape . 4.2. O e iew o ecen op imisa ion applica ions in cons uc ion p ocesses Due o complex na u e o cons uc ion p oblems, and cyclic ela ions o planning and op imisa ion p ocesses, i is o en di icul o iden i y only one ype o he abo e p esen ed o iginal op imisa ion p oblems, o e en o comple ely di e en ia e one p oblem om ano he . Mos op imisa ion p oblems in cons uc ion appea o be a combina ion o a ious o iginal op imisa ion p oblems. I is he e o e ha d o summa ize all me hods ha a e used o sol ing such p oblems. Ne e heless, in his sec ion, he au ho s p o ide an o e iew o ecen and mos known (in he majo i y o cases mos ci ed) applica ions, me hods and models o op imisa ion p oblems in cons uc ion sec o . The o e iew is gi en in o m o wo ch onologically s uc u ed ables di ided by hei op imisa ion o ien a ions: esou ce o ien ed cons uc ion op imisa ion p oblems (Table 2) and layou and ou e op imisa ion p oblems (Table 3). The layou and ou e o ien ed op imisa ion asks applied in cons uc ion (gi en in Table 3) a e ch onologically s uc u ed s a ing om 1995. Since hey a e no as nume ous as he p e ious se o op imisa ion p oblems, hey a e no di ided by longe pe iods. 5. Op imisa ion and BIM 5.1. Gene al connec ions be ween op imisa ion and BIM Op imisa ion can ce ainly be conside ed as po en ially bene icial o many a eas in cons uc ion indus y. The basic aim o his sec ion is o p esen connec ions es ablished be ween op imisa ion and BIM ha ha e been ecen ly epo ed in epu able scien i ic li e a u e. Namely, o he bes o ou knowledge, such a e iew has no ye been ca ied ou no wi hs anding he ac ha i conce ns a apidly de eloping ield. O e iew o ecen achie emen s in he con ex o connec ing op imisa ion and BIM is especially needed o iden i y new pe spec i e opics o esea ch and o ill he li e a u e gap. 2011 – 2016 Non-linea quad a ic AP (QAP) o owe c ane and ma e ial supply loca ions. MILP model. Based on a nume ical example, au ho s ound ha he esul s gained by MILP a e be e han he esul s ob ained by GA, wi h almos 7% less in he o al ma e ial anspo cos . In addi ion, MILP o mula ion was ound o be mo e lexible in e ms o including addi ional design cons ain se s o modelling ac ual on-si e condi ions. [86] Nonlinea disc e e TCTO (NDTCTO) p oblem o a cons uc ion p ojec consis ing o wen y-nine ac i i ies. MINLP model. The p oposed model is mo e complex and equi es g ea e analy ical/ compu a ional e o han he MILP model. Howe e , he au ho s no ed ha he ad an age o he MINLP-NDTCTO model compa ed o he MILP model lies in i s modelling capabili ies. In addi ion, he p oposed model yields he exac op imum NDTCTP solu ion, while heu is ic models calcula e app oxima e op imum solu ions. [87] PSP o he p ojec wi h modula sca olding. Mul i-objec i e cons ained op imisa ion model based on he disc e e i e ly algo i hm (DFA). Resul s p o ed c edibili y o he op imisa ion model by p oducing a be e solu ion o wo k o ce alloca ion enabling p ope ime and cos balance. [88] TCTO p oblem o a cons uc ion p ojec wi h a smalle numbe o p ojec - signi ican ac i i ies. PSO me hod o op imizing global c i ical pa h diag ams using he MATLAB p og amming sys em. The au ho s success ully applied PSO me hod o op imizing ealiza ion o cons uc ion p ojec s. The p oposed model p o ides good esul s and p esen s se e al ad an ages compa ed o me hods based on he use o simplex algo i hms o linea p og amming, and o he adi ional ma hema ical p og amming me hods. [89] NDTCTO p oblem o a cons uc ion p ojec wi h a noncon ex unc ion o cos s. MINLP model. MINLP op imisa ion model o handling noncon ex dependencies was ound o educe he use e o in dealing wi h la ge-size da a and upda ing he model when ci cums ances unde which he p ojec scheduling was done ha e changed. The use o nonlinea exp essions can enable a mo e compac model o mula ion as well as a mo e apid execu ion o model managemen asks, such as ans o ma ion o da a in o model pa ame e s and model modi ica ions. [90] AP and op imal esou ce alloca ion p oblem on mul iple ongoing p ojec s. Bina y mul i-objec i e scena io simula ion model. The model was p o en as a use ul ool o sol ing small and medium scale p oblems. I is adap able o changes o inpu da a. I also enables compa ison o op imum and sub-op imum scena ios wi h he co esponding ou pu da a, and i has hus ul illed he main expec a ions. [91] PSP unde es ic ed cos s. MINLP model. The p oposed MINLP model comp ises o al p ojec cos s, gene alized p ecedence ela ionship cons ain s, p ojec du a ion es ain s, logical condi ions, and cos es ic ions. [92] TP o a ho asphal mix u e. E olu iona y algo i hm o he mul iple c i e ia sol e (MCS). Resul s ha e p o ed ha he e olu iona y algo i hm o he MCS is a use ul ool o sol ing he p oblem o planning anspo o ho asphal mix when he o al p ojec ealisa ion ime is no limi ed. The au ho s poin ou ha he model needs o be modi ied o sol ing big p oblems and o aking in o accoun ime and echnological ela ionships be ween sub p ocesses in he chain. [93] Table 2. Resou ce o ien ed op imisa ion p oblems applied in cons uc ion - ex ension G ađe ina 7/2018 599 GRAĐEVINAR 70 (2018) 7, 593-606 Cons uc ion p ocess op imisa ion – e iew o me hods, ools and applica ions Table 3. Layou and ou e o ien ed op imisa ion p oblems applied in cons uc ion Desc ip ion o applied p oblem Me hods, models and ools Findings and conclusions O igin Planning he cons uc ion-si e access ou es o la ge ehicles. The ou e-planning sys em was achie ed using he expe sys em Nexpe Objec and he geog aphic in o ma ion sys em (GIS) A c/In o in conjunc ion wi h he compu e -aided design (CAD) package Mic os a ion, Excel sp eadshee , and o he cus om p og ams. Alongside he "wha -i " analysis, he au ho s unde line ha he use can employ he p esen ed sys em o selec modi ica ion pa ame e s, modi y selec ed pa ame e s, and es modi ica ion esul s. In u he wo k, au ho s sugges de elopmen s in he ollowing sphe es: model unce ain y and abili y o iden i y a eas o isk on po en ial ou es; quan i ica ion o use p e e ences using a uzzy-logic app oach; and ep esen a ion o s a egies and knowledge equi ed o scena io modi ica ion in o de o acili a e access o si e. [94] Cons uc ion si e layou p oblem (CSLP) aimed a minimising ma e ial anspo cos s. GA gene a es an ini ial popula ion o layou s h ough a sequence o mu a ion ope a ions, and e ol es popula ion layou s h ough a sequence o gene ic ope a ions aimed a inding an op imum si e layou solu ion. Au ho s emphasize ha he key ea u e o he p oposed algo i hm is ha i uses a la ge numbe o di e en GA ope a o s o a y posi ions o objec s a ound he si e. The GA ope a o s a e p og ammed in such o de ha he chance o inding a easible posi ion o a selec ed block is maximized wi h a unc ion ha inds and s o es se s o easible posi ions o a selec ed objec . Ano he key ea u e is ha i main ains in each gene a ion he ch omosomes ep esen ing pa ial layou solu ions. These so-called "bad" ch omosomes a e kep o help he e olu ion p ocess ge ou o local op ima. [95] Scheduling, esou ce planning, and cos op imisa ion p oblem in he scope o la ge cons uc ion and main enance p og ams ha in ol e mul iple dis ibu ed si es wi h he objec i e o inding an op imum se o cons uc ion me hods and an op imum ou ing o de among si es. The au ho s p opose a dis ibu ed scheduling model (DSM) ha uses GAs o de e mine an op imum se o cons uc ion me hods and an op imum ou ing o de among si es. The DSM has p o en o be an indispensable addi ional ool du ing de e mina ion o an app op ia e echnology and an eligible and easible lis o acili ies o be included in municipal cons uc ion/main enance p og ams. I s bene i s include de e mina ion o he numbe o equi ed c ews and hei de ailed wo k plan, and conduc o sensi i i y analysis s udies o de e mining he mos p ope ime o s a execu ion o a cons uc ion p og am; i can also p o ide a dynamic en i onmen needed o mee cons ain s and decide on p ope co ec i e ac ions du ing execu ion. [96] Logis ic and dispa ching p oblems occu ing du ing disas e elie ac i i ies. They a e cha ac e ized by empo a y escape ou e p oblems, ehicle di ec ing p oblems, and he mul i-commodi y dispa ch p oblems. Me a-heu is ic me hod o an colony op imisa ion (ACO), which decomposes he o iginal eme gency logis ics p oblem in o wo sequen ial phases and i e a es be ween hem. In compa ison o he CPLEX solu ion, he quali y solu ion gained by ACO is achie ed wi hin a minu e o un ime, which is especially signi ican in eme gency si ua ions in ol ing con inuous unce ain y and in o ma ion dynamism. Howe e , he local sea ch based me a-heu is ics, such as abu sea ch, needs u he s udy o p o e i s e iciency in sol ing his p oblem. I should be no ed ha in oduc ion o local sea ch in o he pos -op imisa ion p ocedu e does no enhance he o e all solu ion e iciency, al hough some p o isional solu ion quali y is imp o ed in he p ocess. Compu a ional esul s sugges ha his decomposi ion app oach may be e icien o o he complex combina o ial p oblems wi h in e dependen decision a iables. [97] T a ic delays and scheduling p oblem based on he ou e- changing beha iou o oad use s. Mic o-simula ion model in ol ing eam ACO (TACO) sea ch o a nea -op imal scheduling, based on he VISSIM simula ion so wa e. Resul s indica e ha he o al a ic delay can signi ican ly be educed by means o he p oposed model. Compa ed wi h o he ma hema ical me hods, he mic oscopic simula ion equi es mo e compu a ional ime, bu i is close o he eal- ime si ua ion, wha makes he p edic ion and es ima ion mo e eliable. [98] VRP model o uck mixe s a elling o e a wo king day om conc e e plan o conc e e-demanding cus ome s and ice e sa. Combina ion o MILP model wi h gene ic local sea ch app oach using he CPLEX p og am. Gene al MILP sol e is inadequa e o sol ing eal ime la ge-scale p oblems. In o de o sol e such p oblems, he au ho s combined MILP wi h he local sea ch app oach. The sugges ed app oach is based on he assump ion ha all inpu da a a e a ailable o analysis. The e o e, he au ho s ecommend his app oach as a use ul ool o baseline planning. [99] CSLP. In eg a ed simula ion sys em consis ing o he in ica e sea ch and GAs, based on he Symphony pla o m. The p oposed simula ion sys em can easily be ex ended o accommoda e mo e disciplines and s a egies in o de o p oduce mo e ad anced ou pu s. The de eloped modelling sys em enables unnelling wo k expe s o c ea e models, and o expe imen wi h di e en scena ios wi hou he sys em de elope ’s ins uc ions. [100] Dynamic, mul i-objec i e CSLP wi h unequal-a eas. Max–Min an sys em (MMAS) and modi ied Pa e o-based ACO algo i hm. The in ui ionis ic uzzy The pe o mance o CSLP decision-making sys em is e i ied on he case s udy o a esiden ial building. I has p o en o be a use ul ool o p ojec manage s and planne s in he design o cons uc ion si es cha ac e ized by mul iple con lic ing o cong uen objec i es. In addi ion, i helps use s o design a cons uc ion si e layou , including o he quali a i e ac o s, such as ease o supe ision and con ol. [101] TOPSIS me hod was used o he e alua ion and le el selec ion s ages. The MTCARPTW was ans o med in o TSP in o de o apply a heu is ic ACO algo i hm. Resul s sugges ha he p oposed app oach e ec i ely wo ks wi h a MTCARPTW con aining less han i y nodes, yielding a good solu ion wi hin a sho compu ing ime. Au ho s sugges ha u u e s udies should explo e expanding he p esen p oblem in o p ac ical-sized ne wo ks in o de o examine whe he he ACO algo i hm ou pe o ms all known heu is ics on he MTCARPTW, and o de e mine in wha way he algo i hm's e iciency is a ec ed by an inc ease in he ypes o se ices. [102] CSLP as a owe c ane layou p oblem wi h he ma e ial supply and demand op imisa ion. Au ho s compa ed ela i e e iciency o he PBA, PSO, and BA o sol ing he add essed p oblem. Resul s show ha PBA pe o ms be e han he o he wo algo i hms. Al hough he PBA pe o ms well in op imizing loca ion o owe c anes, he algo i hm is unable o minimize ope a ing cos s o supply in case o mul i-dimensional p oblems. [103] Ready-mixed conc e e dispa ching and deli e y p oblem. The au ho s compa ed obus GA and Column Gene a ion (CG) o sol ing eal p oblems o di e en sizes. Resul s show ha , on an a e age, CG ob ains solu ions a 20% lesse cos . Howe e , obus GA con e ges 40% as e han CG, while he numbe o unassigned cus ome s is almos he same o bo h echniques. [104] Ready-mixed conc e e deli e y p oblem. The au ho s p esen ed a scena io simula ion model using he En e p ise Dynamics so wa e. The p esen ed model has p o en o be a use ul ool o sol ing disc e e small and medium size p oblems ela ed o he deli e y o eady-mixed conc e e. I p o ides an op imum solu ion and also he se o sub op imal solu ions, which imp o es he decision making p ocess o he planne . Howe e , model’s dependence on web maps is a majo cons ain hinde ing applicabili y o he model. [105] G ađe ina 7/2018 600 GRAĐEVINAR 70 (2018) 7, 593-606 Václa Venk bec, Ma io Galić, U oš Klanšek The BIM pa adigm is one o he mos p omising de elopmen s in he a chi ec u e, enginee ing, and cons uc ion indus ies [106]. The mos equen ly used de ini ion is: BIM is essen ially alue c ea ing collabo a ion h ough he en i e li e-cycle o an asse , unde pinned by he c ea ion, colla ion and exchange o sha ed 3D models and in elligen , s uc u ed da a a ached o hem [107]. Building in o ma ion model is da a- ich, objec -o ien ed, in elligen and pa ame ic ep esen a ion o physical and unc ional cha ac e is ics ans o med in o a mul i-dimensional digi al compu e -based model. All cha ac e is ics abou acili ies ealised h ough BIM-based ools a e a ailable o he use . BIM applica ions can be ca ego ized acco ding o hei basic ocus. Mos o BIM-based ools manually ans o m accu a e syn he ic da a in o i ual eali y and usually con ain 3D geome ical model wi h o he isualized in o ma ion (i.e. dimensions). This app oach can be named as passi e BIM because he analy ical pa o he model is absen . The analysis o s uc u al da a, such as an op imum ime scheduling and cons uc ion si e wo kspace planning, isk assignmen con ol, heal h and sa e y con ol o cons uc abili y e iew, also in ol es he use o o he ools, which mus be sui ably backed by compu e skills and expe knowledge o he use . Viewed in he s ic ma hema ical sense, BIM as a concep is no an op imisa ion app oach un il a leas one o he op imisa ion me hods is applied. This pu s in o pe spec i e he di e ences be ween he simula ed solu ion and he op imized one. The main pu pose o his sec ion is o place emphasis on ecen ly published BIM-based app oaches, ope a ionally named as ac i e BIM app oaches. In hese cases, models a e wo king wi h analy ical da a h ough algo i hms. The ollowing subsec ions concen a e on only hose ac i e BIM applica ions ha employ op imisa ion echniques and a e mos equen ly iden i ied in cons uc ion managemen li e a u e, i.e. pa icula ly hose o APs, CSLPs, and PSPs. 5.2. Ac i e BIM applica ions The aim o he ollowing able is o show ac i e BIM applica ions (i.e. in eg a ion o op imisa ion echniques and BIM) de eloped o he pu pose o gene a ing op imum esul s. Mos equen ly iden i ied applica ions a e namely he ones o APs and CSLPs. Table 4 shows con ibu ions in a ime-so ed s uc u e. Ac i e BIM applica ions o PSPs a e shown by ch onological o de in Table 5. Yea Desc ip ion o applied p oblem Me hods, models and ools Main indings and conclusions O igin 2003* AP o inding op imal posi ions o empo a y acili ies in CSLP. Au ho s de eloped a new algo i hm by using GA on he CAD-based pla o m. The pape concluded ha he GA by i sel does no ensu e an op imum CSLP solu ion, bu may o e a nea op imum solu ion. By minimizing he objec i e unc ion, he GA accomplishes he complex ask o assigning empo a y acili- ies o posi ions consis en wi h hei espec i e p oximi y needs. [108] 2005* AP o op imizing loca ion o c anes and acili ies. GA me hod was used o posi ioning c anes and acili ies. Signi ican ime sa ings can be made using he GA model o c ane posi ioning on cons uc ion si es. The se ing o wo ypes o ch omosomes has been ound use ul o gene a ing he GA model. The i s ch omosome indica es he c ane posi ion code while he second one indica es he numbe o c anes. Fu u e wo k can be ex ended o in eg a e and hyb idize he GA-based model wi h he 3D isualiza ion echnique. [109] 2010* CSLP o ai po expan- sion p ojec s. GA me hod was used o posi ioning c anes and acili ies. Signi ican ime sa ings can be made using he GA model o c ane posi ioning on cons uc ion si es. The se ing o wo ypes o ch omosomes has been ound use ul o gene a ing he GA model. The i s ch omosome indica es he c ane posi ion code while he second one indica es he numbe o c anes. Fu u e wo k can be ex ended o in eg a e and hyb idize he GA-based model wi h he 3D isualiza ion echnique. [110] 2014 BIM-based CSLP ocusing on ac ual a el pa hs. GA was employed o gene a e dynamic CSLP model. Au odesk Re i was used as he BIM ool and Mic oso P ojec was employed o scheduling. The au ho s de eloped hei own so wa e ha eads excel sp eadshee s. Resul s show ha he linea dis ance based op imisa ion o minimize he anspo dis ance o si e pe sonnel could lead o he gene a ion o sub- op imal layou s, and hence he ac ual a el dis ance was used. The me hod gene a ed layou s ha educed o al on-si e a el dis ances by 16.5%. Fu u e esea ch on his model could be o ien ed owa d in eg a ion o he 4D con- s uc ion simula ion. [111] 2015 BIM-based au oma ed CSLP o conges ed cons uc ion si es. GA was applied o gene a e dynamic CSLP models. Calcula ions we e based on BIM gene a ed da a. Resul s show a educ ion o app oxima ely 13.5 % in he ac ual a el dis- ance on cons uc ion si e compa ed o con en ional me hods. The au ho s concluded ha he con en ional algo i hm, which uses di ec dis ances in he op imisa ion p ocess, is no su icien . The model could be ex ended o include eal ime cons uc ion schedules and ma e ial logis ics models. [112] 2015 BIM-based au oma ed CSLP o cons uc ion si e cos op imisa ion GA, BIM and RFID a e combined o op imise gene a ion o cons uc ion si e layou s in eal ime. The esul s show ha he p oposed sys em can au oma ically ack empo a y acili ies, and model a ailable cons uc ion si e spaces, in o de o minimise cos s. [113] Table 4. Ac i e BIM applica ions o APs and CSLPs G ađe ina 7/2018 601 GRAĐEVINAR 70 (2018) 7, 593-606 Cons uc ion p ocess op imisa ion – e iew o me hods, ools and applica ions Table 4. Ac i e BIM applica ions o APs and CSLPs - ex ension Table 5. Ac i e BIM applica ions o PSPs 2014 2015 AP o gene a ing an op imal owe c ane layou plan. A BIM applica ion is used o au oma ically gene a e ma e- ial quan i ies needed o on-si e anspo . FA-based op imisa ion is employed o de e mine owe -c ane loca ions, as well as ma e ial sou ce and des ina ion poin s. Resul s show ha less ime is needed o c ea e a owe c ane layou scheme in compa ison wi h adi ional me hods, especially when mo e han one owe c ane is used. The me hod also gene a es sa ings in o al ma e ial anspo cos s and possible collision cos s as, due o isualisa ion o he p ocess, he wo ke s can easily unde s and and implemen he owe c ane layou scheme. [114] [115] 2015 AP o op imum owe c ane posi ioning and c ane ype selec ion. The analy ical hie a chy p ocess (AHP) is u ilized as a mul i-c i e ia decision making (MCDM) me hod o c ane ype selec ion. GA is used o de ining an op imum c ane layou , sho es ail leng h, and an op imum li ing assignmen plan. BIM model is u ilized o ex ac ing quan i ies equi ed o GA op imisa ion. Resul s show ha he use o AHP me hod o c ane ype selec ion is qui e signi ican and ha i p esen he highes le el o sensi i i y by o e ing he g ea es numbe o c i ical selec ion c i e ia. The p esen ed example shows ha he hamme head owe c ane is he mos sensi i e c ane ype since i has ob ained he highes sensi i i y coe icien s. Au ho s concluded ha he owe c anes BIM model can addi ionally be de eloped by adding mo e unc- ionali ies such as enabling ad anced spa ial que y unc ions, o in oducing mobile c ane modelling capabili ies. [116] 2016 AP o es ima ing a el equencies in CSLP. Objec i e unc ion mini- mizes a el cos s. In eg a ion o BIM inpu da a and a el equency scheduling is p e- sen ed. The VBA analyse is coded o linking in o ma ion ga he ed om bo h BIM and p ojec schedule. Es ima ed a el equencies a e se as inpu da a o MILP based CSLP model. Resul s show capabili ies o he p oposed model o au oma ing he a el equency es ima ion p ocess. The au ho s concluded ha a el equency alues used om o he p ojec s can gene a e some e o s in he op imisa ion p ocess. I is sugges ed ha u he de elopmen and imp o emen in model accu acy be made by collec ing own da abase om mo e ypes o p ojec s. [117] 2016 AP o c ane and ma e ial supply poin s posi ioning, aking in o accoun ope - a ing and en al cos s. MILP model o c ane loca ion op i- misa ion ha minimizes o al cos s. A case s udy model is linea ized and sol ed using CPLEX sol e . The owe c ane loca ion ob ained by he p oposed model e eals signi ican di e ences when he c ane capaci y is no aken in o accoun . In his case, ap- p oxima ely 30% educ ion in o al cos s is achie ed using he p oposed MILP model. The esul s show ha o e looking he c ane capaci y equi emen s in he c ane loca ion p oblem may lead o e o s as high as 20% when compa ed o he op imum o al cos s. [118] * Con ibu ions ep esen CAD-based p edecesso s o he ac i e BIM-based applica ions. Thei ele ance is in he connec ion be ween op imisa ion me hods and p e ious gene a ion o compu e -aided ools o AEC indus y. Godina Opis p oblema Me ode, modeli i ala i Dop inosi i zaključci Iz o 2008* PSP o ime-cos in e- g a ed schedule. The objec sequencing ma ix (OSM) and GA a e used. Mul i-dimensional CAD is used as model ex ension. The me hodology o using GA o op imise cons uc ion sequences and de ine c ew assignmen s is p esen ed in six main s eps in he pape . The model akes in o accoun cons ain s ega ding esou ces, wo kspace and p oduc i i y. [119] 2012 PSP and wo kspace con lic s p oblem. GA-based minimiza ion o wo kspace con lic s and a 4D CAD sys em a e used. The p oposed s udy e eals ha he wo kspace model and compu ed con lic s a io should be de e mined p io o he op imisa ion o wo kspace con lic s. The eupon, adjacency o mu ual wo kspaces can be e iewed and schedule con lic s can be de ec ed. Da a can be op imized by GA in o de o ob ain e- sul s wi h minimum du a ion o con lic s. I is expec ed ha all hese unc ions will be used as basic da a o he es ablishmen o an ac i e BIM en i onmen . To he bes o ou knowledge, he e m "Ac i e BIM" is used o in his pape o he e y i s ime. [120] 2013 PSP o educing sched- ule o e laps. BIM wi h uzzy heo y unc ion is used o quan i ying cons uc ion isk. GA is applied o schedule op imisa ion. Reduc ion in schedule o e laps on a cons uc ion p ojec wi h e olu ion o each GA gene a ion can be isually checked ia he p esen ed ac i e BIM sys em. Resul s ob ained by he p oposed GA model e eal ha he schedule o e lapping a e is educed by app oxima ely 33 %. [121] 2013 PSP o scheduling wi h he minimum o e lap- ping le el. GA and uzzy-based isk analysis a e used. The au ho s demons a e ha e olu ion o esul s in each gene a ion can be isually checked by an ac i e BIM sys em. I is shown in he pape ha GA can be used no only o schedule op imisa ion bu also o de eloping p ojec schedules om he sc a ch. [122] 2014 PSP o educing sched- ule o e laps. The au ho s use a e ised GA o minimizing schedule and wo kspace in e e ence, as well as an ac i e BIM-based simula- ion sys em o isualisa ion o he solu ion. Acco ding o he au ho s, he de eloped sys em can be u ilized as an ac i e BIM ool ha can be applied o si es wi h minimum special knowledge equi ed om he adminis a o . Based on s udy esul s, he alue o he schedule- wo kspace in e e ence impac ac o is educed o 26.4% o he planned alue (i.e. educed by 73.6% compa ed o he plan). Howe e , he p oposed app oach is used on a simple example and a de ailed modi ica ion o he 4D model is conside ed impo an p io o i s applica ion on o he p ojec s. [123]