A Machine Learning Approach to Predict Production Time in Industrialized Building Construction

  • Author / Creator
    Mohsen, Osama
  • Industrialized building construction is an effective approach for improving the performance and management of construction projects by offering higher quality products, minimized environmental impacts, and improved schedule predictability. The industrialized building construction approach integrates manufacturing principles and techniques into the construction industry where products (in this case, building components) are built in a controlled factory environment and then transported, in sequence, to the construction site for the final assembly. With the marked growth of available data gathered as part of the daily operations in industrialized building construction facilities, one promising approach is to utilize machine learning techniques to identify valid, useful, and previously unknown patterns from historical data. These techniques are used to leverage the use of data to accurately predict the production cycle time. This thesis presents a framework to investigate potential improvements in estimating cycle time in industrialized building construction facilities where accurate prediction of cycle time can improve the quality of production planning and scheduling. The goal is to assist production/project managers to mitigate any delays and implement alternative actions should there be any unexpected delays due to factory operations as well as to manage the capacity and workload of the fabrication facility more efficiently. The results of two case studies reveal that machine learning approaches can be successfully applied to accurately predict cycle time in different subsectors of industrialized building construction. The factors that significantly affect the prediction accuracy are (1) the physical characteristics and tracking information of products, (2) the engineered features that are generated to capture real-time loading conditions of the job shop, and (3) the lookback timeframe used for model training and validation. To examine the effect of shop loading features more thoroughly, a discrete-event simulation model is developed to investigate the various features that can be captured in the real system, but for practical reasons are not presently captured in the data, and to investigate their benefit in terms of improving the predictive accuracy. The major contribution of the research presented in this dissertation is that it provides practitioners in industrialized building construction with a roadmap to utilize their available production data for accurate estimates of delivery dates. The research results are also expected to be a point of reference for future studies in the academic field and for the industrialized building construction industry.

  • Subjects / Keywords
  • Graduation date
    Fall 2021
  • Type of Item
  • Degree
    Doctor of Philosophy
  • DOI
  • License
    This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.