Improving labour productivity in construction. A hybrid machine learning approach

Bokor, Orsolya (2022) Improving labour productivity in construction. A hybrid machine learning approach. Doctoral thesis, Northumbria University.

Text (Doctoral thesis)
bokor.orsolya_phd(17024287).pdf - Submitted Version

Download (9MB) | Preview


Achieving less than ideal productivity is a problem the construction industry faces in most advanced countries, including the UK. One way to change this is to improve on-site execution by, for example, more accurate planning of construction operations. Despite continuous efforts for automation, mechanisation, and off-site production, the construction industry can still be considered labour-intensive. Therefore, understanding labour productivity and the factors influencing it is vital to better planning.

Owing to their versatility, durability, long service life, and being low maintenance, bricklaying works are ubiquitous, especially in housing and public projects, for example, schools. These operations are also especially labour-intensive. Consequently, an examination of bricklaying works is important for better planning and management of most construction projects. Ultimately, any gains in this operation could lead to an overall increase in site-based productivity.

The aim of the research project is to provide a better understanding of the bricklaying process and how it can be modelled, descriptively and normatively, to find a modelling approach that allows for a better examination of the effects of various factors on bricklaying productivity.

A number of factors influence on-site productivity. This research project focuses on those that are known in advance, in the pre-planning phase of the construction projects. These are the worker and wall characteristics.

To analyse bricklaying operations, a hybrid model is created. The effects of the above-mentioned factors on labour productivity are investigated with the help of the artificial neural network component, while the discrete-event simulation part models the process of block- and bricklaying. The model is built and tested with the help of real-life data collected at two construction projects by conducting a traditional work study. When the productivity rates were measured, note was made of the bricklayer working on the course, and the wall section where they worked. Site supervisors filled in the questionnaires asking about operative characteristics, while the wall characteristics were determined based on the drawings and specifications.

The resulting model can be used to provide more accurate productivity rate predictions for more precise time and cost estimates, and improved project planning in bricklaying.

Item Type: Thesis (Doctoral)
Uncontrolled Keywords: artificial neural networks, bricklaying, discrete-event simulation, productivity modelling
Subjects: K900 Others in Architecture, Building and Planning
Department: Faculties > Engineering and Environment > Architecture and Built Environment
University Services > Graduate School > Doctor of Philosophy
Depositing User: John Coen
Date Deposited: 15 Nov 2023 09:41
Last Modified: 15 Nov 2023 09:45

Actions (login required)

View Item View Item


Downloads per month over past year

View more statistics