Show simple item record

Model for Forecasting and Assessment of Construction Cost of Reinforced-Concrete Bridges

dc.contributor.advisorIvanišević, Nenad
dc.contributor.otherIvković, Branislav
dc.contributor.otherKnežević, Miloš
dc.creatorKovačević, Miljan M.
dc.date.accessioned2018-12-28T09:36:17Z
dc.date.available2018-12-28T09:36:17Z
dc.date.available2020-07-03T08:29:48Z
dc.date.issued2018-09-28
dc.identifier.urihttp://eteze.bg.ac.rs/application/showtheses?thesesId=6415
dc.identifier.urihttps://nardus.mpn.gov.rs/handle/123456789/10553
dc.identifier.urihttps://fedorabg.bg.ac.rs/fedora/get/o:19116/bdef:Content/download
dc.identifier.urihttp://vbs.rs/scripts/cobiss?command=DISPLAY&base=70036&RID=513803666
dc.description.abstractU radu su predstavljene i analizirane najsavremenije tehnike mašinskog učenja koje se mogu primeniti kod procene troškova izgradnje armirano-betonskih drumskih mostova. Analizirana je primena veštačkih neuronskih mreža, ansambla regresionih stabala, modela zasnovanih na metodi potpornih vektora, Gausovih slučajnih procesa. Formirana baza podataka o troškovima izgradnje mostova zajedno sa njihovim projektnim karakteristikama predstavljala je osnovu za formiranje modela za procenu. Modeli su formirani na osnovu podataka za 181 armirano-betonski drumski most čija vrednost prevazilazi 100 miliona evra. Model zasnovan na metodi Gausovih procesa pokazao je najveću tačnost procene troškova izgradnje mostova. Istraživanje je ukazalo da primena ARD funkcija kovarijanse daje modele najveće tačnosti, a pored toga omogućava i sagledavanje značaja koje imaju pojedine ulazne promenljive na tačnost modela. Primenom modela sa ARD funkcijom kovarijanse formirani su i modeli za procenu utroška betona, visokovrednog i rebrastog čelika. Postignuta je tačnost modela kod procene ugovorenih troškova izgradnje izražena preko srednje apsolutne procentualne greške od 10,86%. Kod modela za procenu utroška ključnih materijala za izgradnju postignuta je tačnost modela čija je gornja granica 11,64% izražena preko srednje apsolutne procentualne greške. Sprovedeno istraživanje potvrđuje da je u ranim fazama razvoja projekta metodama baziranim na veštačkoj inteligenciji moguća brza i dovoljno precizna procena troškova izgradnje armirano-betonskih drumskih mostova i utroška ključnih materijala za njihovu gradnju.sr
dc.description.abstractContemporary machine learning techniques for assessment of construction costs of reinforced-concrete bridges, including artificial neural networks, regression tree ensembles, support vector regression and Gaussian random processes, are proposed and analysed in this dissertation. The database of construction costs and project characteristics is created, that served as a basis for building the assessment model. Data for 181 reinforced-concrete bridges were used in the database with the total value of over 100 000 000 EUR. The model based on Gaussian processes has shown the best performance in forecasting the construction costs of bridges. The results have proved that using the Automatic Relevance Determination (ARD) covariance function leads to the best prediction model, and moreover, it enables the assessment of the influence of input variables on the model performance. Models for the assessment of costs of concrete, as well as ribbed steel, were analysed. The mean absolute percentage error (MAPE) was used as the performance criterion. The best performing model gives MAPE equal to 10,86% for forecasting the contracted construction costs and MAPE equal to 11.64% for quantity estimation of the key construction materials. The research carried out in this dissertation confirms that the use of artificial intelligence based methods enables fast and accurate forecasting of construction costs of reinforced-concrete bridges, as well as the assessment of quantity estimation of the construction materials, even in early project phases.en
dc.formatapplication/pdf
dc.languagesr
dc.publisherУниверзитет у Београду, Грађевински факултетsr
dc.rightsopenAccessen
dc.rights.urihttps://creativecommons.org/licenses/by-nc-nd/4.0/
dc.sourceУниверзитет у Београдуsr
dc.subjectUpravljanje projektimasr
dc.subjectProject Managementen
dc.subjectCost Assesmenten
dc.subjectArtificial Inteligencyen
dc.subjectMachine Learningen
dc.subjectBridgesen
dc.subjectprocena troškovasr
dc.subjectveštačka inteligencijasr
dc.subjectmašinsko učenjesr
dc.subjectmostovisr
dc.titleModel za prognozu i procenu troškova izgradnje armirano-betonskih drumskih mostovasr
dc.title.alternativeModel for Forecasting and Assessment of Construction Cost of Reinforced-Concrete Bridgesen
dc.typedoctoralThesisen
dc.rights.licenseBY-NC-ND
dc.identifier.fulltexthttp://nardus.mpn.gov.rs/bitstream/id/4338/Disertacija.pdf
dc.identifier.fulltexthttp://nardus.mpn.gov.rs/bitstream/id/4339/IzvestajKomisije18716.pdf
dc.identifier.fulltexthttps://nardus.mpn.gov.rs/bitstream/id/4338/Disertacija.pdf
dc.identifier.fulltexthttps://nardus.mpn.gov.rs/bitstream/id/4339/IzvestajKomisije18716.pdf
dc.identifier.rcubhttps://hdl.handle.net/21.15107/rcub_nardus_10553


Files in this item

Thumbnail
Thumbnail

This item appears in the following Collection(s)

Show simple item record