Aristotle University Of Thessaloniki (AUTH)
Εκτίμηση της κατασκευασιμότητας τεχνικών έργων βάσει της ανάλυσης κινδύνων
Abstract
dc:descriptionThe aim of the present doctoral thesis is the tackling, for the first time, of the lack of the methodological and computational integration between constructability and technical project risk analysis. Such an integration, where the aforementioned could function complementarily and as a single framework, can lead to better cooperation and communication among the stakeholders of the managerial team, better organized and more deeply integrated design and construction phases, more targeted capital investments and the implementation of innovative technologies, better supply chain management, more robustly documented assessment of the construction site conditions, advanced emphasis on labor health and safety, updated agreements between owners and contractors, and facilitation of the bidding, procurement and administration procedures. This integration, following a thorough and exhaustive literature review, was tackled as a statistical classification problem, which was modelled through the following: (1) formulation of thirteen novel and perceptible constructability concepts, (2) finding and formulation of the model’s independent variables (129 general risk sources), through the development of a novel unsupervised machine learning algorithm, which performs semantic processing and clustering of linguistic data in list format, and which was implemented on an exhaustive part of the related research literature (which acted as the relevant dataset); this algorithm included semantic processing tools, namely stop word and the Porter stemmer, and the linguistic clustering tool k-means++, and tackled the discrepancy regarding the definition of risk, as well as the suboptimal exploitation of the former conducted relevant research, (3) formulation of the model’s depended variables (constructability classes), (4) realization of the integration itself, through the development of a novel supervised machine learning algorithm, which appraises the constructability class of a technical project when given the values of the identified and assessed general risk sources that potentially affect it, and which was trained and validated through the utilization of real – and collected through unstructured interviews and the use of the thirteen novel constructability concepts – data displaying the risk analyses and labeled constructability classes of technical projects; this algorithm included regularized stochastic gradient descent non-negative matrix factorization with missing values, training of “soft margin” support vector machines solved with the optimized sequential minimal optimization process, and simultaneous validation with n-fold cross-validation, (5) elicitation of an equation that appraises, with high accuracy, whether a new technical project is constructable or non-constructable, given the values of the 129 general risk sources that potentially affect it, and (6) development of the software tool RISCONA (RIsk Source-based CONstructability Appraisal), for the better application of the model and supervision of its results. The main contribution of this thesis is that it provides – for the first time – a civil engineer with the capability to execute an understandable and fully defined risk analysis, and through this, obtain a very good appraisal of the constructability of the currently studied technical project.
Degree
thesis:*- Grantor dc:publisher
- Aristotle University Of Thessaloniki (AUTH)
- Year dc:date
- 2018
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Kifokeris, Dimosthenis
- Κιφοκέρης, Δημοσθένης
Subjects
dc:subject × 12Rights
dc:rights- Statement dc:rights
-
- BY_NC_ND
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/45100