Development of Conceptual Quantity Models for Building Envelopes and Structural Elements


Student thesis: Doctoral Thesis

View graph of relations


Awarding Institution
Award date12 Aug 2019


Quantity estimation is important in arriving at the cost estimate for a building project. However, at the conceptual stage or the predesign stage of a building project there is insufficient information to produce cost estimates that are based on quantity estimates and unit rates. This increases the uncertainty in predicting building cost and making cost related decisions at the conceptual stage. Cost prediction is essential and plays a critical role in project planning and cost framing at the early stage of construction projects. The need to take informed decision and reduce the uncertainty of project cost at early predesign stage makes cost modelling an important research field in construction project management. Cost modelling studies have produced various cost models which predict cost as a single variable. These cost models are considered black box models and can become inadequate as new or advanced construction technologies and materials are developed. Another significant limitation of predicting cost as a single variable is the inability to model construction resources of key building elements at the predesign stage. Therefore, it becomes difficult to obtain cost details of significant cost items for early resource planning. A feasible solution in addressing the above limitations is to model a component of construction cost; quantities. Hence, the development of quantity models for predicting conceptual quantities of cost significant building elements is the aim of this research. Attention was given to two cost significant elements in this research—the building envelope and structural elements. To develop the building envelope model, empirical study of theoretical shapes and mathematical analyses of shapes using the fundamental theorem of integral calculus for “area under a curve” was carried out. Support vector regression (SVR); a machine learning technique, was used in combination with bootstrap resampling for range prediction of structural quantities in foundation, beam, slab and column of low-rise reinforced concrete buildings. Based on the empirical and mathematical analyses, a new generalised mathematical model for vertical building envelope quantities was developed and a corresponding graphical model was established in this research. The developed mathematical model provided better results when compared to the traditional superficial cost model for building envelopes when sketch drawings are unavailable. A web-based estimator for conceptual building envelope (WECBE) was also developed. An overlooked predictor variable in the literature for conceptual structural quantities was found to be a good predictor of foundation quantities. Assessment of the predictive performance of the conceptual structural quantity models showed that the bootstrapped SVR models can provide useful ranges of quantities at 95% prediction intervals. The structural and envelope quantity models as predesign cost planning tools can assist construction planners in early resource planning, provide valuable cost advice and enable performance measurement of early cost predictions throughout the construction process. It is anticipated that the developed models will improve the certainty of conceptual cost when a “design to cost” approach is adopted early on building projects with limited financial and construction resources. Finally, the research shows the practical application of machine learning in the field of quantity surveying.