Applying data analytics for enhanced construction project performance through structural concrete rework predictive models.
Date
2021-12
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Stellenbosch : Stellenbosch University
Abstract
ENGLISH ABSTRACT: Today’s world is driven by data-based decision-making that needs to be accurate to effectively
solve engineering problems involving the prediction of failure, defects, and errors. Motivated
by the fourth industrial revolution (Industry 4.0) that has enhanced the performance of
construction industries on a global scale, this study discusses the development of a predictive
machine learning model that can be used during the construction phase to manage structural
concrete rework during site inspections. This model seeks to reduce uncertainties and minimise
structural concrete rework during construction to enhance project performance. To develop the
model, the research approach included an exploratory case study together with interviews with
experienced professionals in structural concrete construction at the Hwange Expansion Project,
a mega thermal power plant construction project in Hwange, Zimbabwe. The exploratory case
study and expert interviews were conducted to establish a better understanding of the risk
triggers that influence structural concrete rework in a typical construction project.
A fictitious modelling dataset was then generated based on the results of a questionnaire survey
conducted on structural concrete experts due to the lack of sufficient project data. Various data
mining techniques were also employed to develop the prediction model following some steps
of the Cross Industry Standard Practise for Data Mining (CRISP-DM) framework. This
fictitious dataset was modelled on five classification algorithms whose performance was
evaluated using the 20-fold cross-validation test. The Neural Network classifier recorded the
highest performance with accuracy and precision of over 95%. To validate the performance of
the Neural Network prediction model, the confusion matrix validation test was carried out on
six datasets of varying size ranging from 500 to 10 000 data points. The results from the
confusion matrix validation test indicated, as expected, that the larger the dataset, the more
accurate and robust the model becomes in predicting new data outcomes.
Based on these findings, it was established that data analytics in the form of predictive
modelling can be used by organisations to reduce uncertainties and promote data-driven
decision-making during structural concrete quality checks on site. It is recommended that
construction industries employ data analytics as a project management tool not only to enhance
the performance of construction projects but to build reference databases for further
development of big data in the industry.
AFRIKAANSE OPSOMMING: Geen opsomming beskikbaar
AFRIKAANSE OPSOMMING: Geen opsomming beskikbaar
Description
Thesis (MEng)--Stellenbosch University, 2021.
Keywords
Structural concrete, UCTD, Engineering -- Machine learning, Construction projects, Failure analysis (Engineering), Survival analysis (Biometry), Analytics, Predictive