Developing and Optimizing Context-Specific and Universal Construction Labour Productivity Models

Construction labour productivity (CLP) significantly influences the profitability of construction companies; however, CLP exhibits the highest variability among project resources and is a major source of project risk. The construction industry is thus constantly searching for ways to improve labour productivity. Unfortunately, despite long-term, continued research and industry practice, predicting and improving CLP remains a challenge. Previous productivity studies mainly focus on factor and activity models, using factor models to model productivity with context-specific influencing parameters (factors and practices), and activity models to model the relationship between productivity and work sampling proportions (WSP). However, modeling CLP remains a challenge as for a given context, the complex impact of the multiple subjective and objective variables, made up of critical factors, practices, and WSP; have to be considered simultaneously, while maintaining a high accuracy and interpretability in developed models. To address these challenges, this thesis presents advanced frameworks for the development of a series of interpretable and accurate fuzzy inference based context-specific CLP models, which are then abstracted to develop the universal CLP models, and facilitate a better understanding of the variables that influence CLP. The development of the CLP models included identifying, classifying, quantifying, and documenting the variables influencing CLP. By analyzing existing literature in the field of CLP analysis and modeling, the influencing variables, made up of 169 parameters and 7 work sampling categories, were identified and quantified. The research conducted extensive field data collection from 11 construction projects across Alberta, Canada, spanning over a time period of 29-months; and documented information using factor survey, factors and practices documentation, work sampling studies, foreman delay surveys, craftsman questionnaires, and productivity measurements. First, the research identified the key variables influencing CLP using expert and data-driven approaches in order to reduce the large feature space of the variables. Next, the role of work sampling proportions in CLP modeling was formulated by testing the fundamental assumption of activity models—that CLP improves if more time is spent on direct work activities—and analysis results showed that using work sampling proportions alone, it is not possible to accurately predict CLP. Thus, a system-based modeling framework to incorporate work sampling proportions with factors and practices leading to improved CLP modeling and analysis was developed. Then, an operational definition of context for CLP modeling was formulated and associated context attributes were developed, based on the 5W1H (Who, What, Where, When, Why, and How) question and answers approach, and employed together with the system-based CLP modeling framework for the development of a series of context-specific CLP models after combining projects sharing similar contexts. Using a hybrid fuzzy multi-objective optimization framework, the learning ability of the developed fuzzy inference system CLP models was improved. Finally, a context adaptation framework for transferring knowledge among contexts was developed using linear and non-linear adaptation on the membership functions of the context-specific fuzzy CLP models, and a framework for developing universal CLP models is established. The main contributions of this research to the state of art of CLP modeling and analysis are: (1) evaluation of the usefulness of relying on work sampling proportions like direct work or tool time to predict CLP, (2) development of a system model framework for CLP, which provides a better understanding of CLP and the variables influencing CLP, (3) addressing the challenges faced in past CLP models by developing and optimizing fuzzy inference CLP models, (4) presenting an operational definition of context for CLP modeling for characterizing and classifying construction projects and assisting in the process of grouping similar projects for more accurate context-specific CLP model development, and (5) developing frameworks for adaptation and abstraction of context-specific CLP models. The developed frameworks and findings of this study are of a value to researchers and industry practitioners and provide a better understanding of CLP, the variables influencing CLP, and how work-study methods like work sampling can be integrated to provide an accurate CLP analysis tool.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.