Hybrid Fuzzy-Bayesian Dynamic Decision Support Tool for Resource-Based Scheduling of Construction Projects

Abstract

This dissertation proposes a flexible and intelligent decision support tool for scheduling and resource allocation of construction projects. A hybrid Fuzzy-Bayesian scheduling network and a new optimization model and solution approach have been developed to assess the combinatory effect of different risk factors on scheduling and optimize the time-cost tradeoff. Developed decision support tool employs interval-valued fuzzy numbers and Bayesian networks to dynamically quantify uncertainty and predict project performance during its make span. Using interval-valued fuzzy numbers makes the model more flexible and intelligent comparing to conventional fuzzy risk assessment models through incorporating the decision makers` confidence degree. The linguistic assessments of experts regarding the likelihood and severity of increase or decrease in task duration and cost when influenced by different risk factors are used to generate a set of duration and cost prior-probability distributions. A learning dynamic Bayesian scheduling network is developed to probabilistically combine the prior-probability distributions with initial activity duration estimates and update them as new evidence in form of actual activity data feed into the network. This model also predicts project performance at any point of time during its execution. Optimization model explicitly considers variation of time-cost tradeoff relationship during project execution and complex payment terms to maximize the project net present value (NPV). A sequential solution approach is proposed to combine a procedure for updating time-cost tradeoff data, and mixed integer linear programming (MILP) methods to obtain optimal project crashing and scheduling solutions that is adaptive to the current project status and crew productivity. Capability of proposed model in quantifying uncertainty at initial phases of project where project performance data are scarce, learning from data and predicting project performance, considering financial aspects of scheduling through optimal resource allocation and providing useful and clear advice to managers are advantages of developed decision support tool over already existing approaches.