Abstract

Researchers and industry practitioners agree that changes are unavoidable in construction projects and may become troublesome if poorly managed. One of the root causes of sub-optimal productivity in construction projects is the number and impact of changes introduced to the initial scope of work during the course of project execution. In labor-intensive construction projects, labor costs represent a substantial percentage of the total project budget. Understanding labor productivity is essential to project success. If productivity is impacted by any reasons such as extensive changes or poor managerial policies, labor costs will increase over and above planned cost. The true challenge of change management is having a comprehensive understanding of change impacts and how these impacts can be reduced or prevented before they cascade forming serious problems. This thesis proposes a change management framework that project teams can use to quantify labor productivity losses due to change orders and managerial policies across all phases of construction projects. The proposed framework has three models; fuzzy risk-based change management, AI baseline-productivity estimating, and system dynamics to illustrate cause-impact relationships. These models were developed in five stages.
In the first stage, the fuzzy risk-based change management (FRCM) model was developed to prioritize change orders in a way that only essential change orders can be targeted. In this stage, Fuzzy Analytic Hierarchy Process (F-AHP) and Hierarchical Fuzzy Inference System are utilized to calculate relative weights of the factors considered and generate a score for each contemplated change. In the second stage, baseline productivity model was developed considering a set of environmental and operational variables. In this step, various techniques were used including Stepwise, Best Subset, Evolutionary Polynomial Regression (EPR), General Regression Neural Network (GRNN), Artificial Neural Network (ANN), Radial Basis Function Neural Network (RBFNN), and Adaptive Neuro Fuzzy Inference System (ANFIS) in order to compare results and choose the best method for producing that estimate. The selected method was then used in the development of a novel AI model for estimating labor productivity. The developed AI model is based on Radial Basis Function Neural Network (RBFNN) after enhancing it by raw dataset preprocessing and Particle Swarm Optimization (PSO) to extract significant dataset features for better generalization. The model, named PSO-RBFNN, was selected over other techniques based on its statistical performance and was used to estimate the baseline productivity values used as the initial value in the developed system dynamics (SD) model.
In the fourth stage, a novel SD model was developed to examine the impact of change orders and different managerial decisions in response to imposed change orders on the expected productivity during the lifecycle of a project. In other words, the SD model is used to quantify the impact of change orders and related managerial decisions on excepted productivity. The SD model boundary was defined by clustering key variables into three categories: exogenous, endogenous, and excluded. The relationships among these key variables were extracted from the literature and experts in this domain. A holistic causal loop diagram was then developed to illustrate the interaction among various variables.
In the final stage, the developed computational framework and its models were verified and validated through a real case study and the results show that the developed SD model addresses various consequences derived from a change in combination with the major environmental and operational variables of the project. It allows for the identification and quantification of the cumulative impact of change orders on labor productivity in a timely manner to facilitate the decision-making process. The developed framework can be used during the development and execution phases of a project. The findings are expected to enhance the assessment of change orders, facilitate the quantification of productivity losses in construction projects, and help to perform critical analysis of the impact of various scope change internal and external variables on project time and cost.