Time Buffer Setting And Robust Project Scheduling

Due to the rapid changing of market environment, there are more uncertainties involved in projects especially in large-scale projects. These uncertainties make more and more projects face problems, such as cost overruns, completion delay and even abort. At the same time, modern project management requires shorter period, higher on-time completion rate and lower cost. In this case, to deal with various uncertainties arising in complex environments, the robust project scheduling which aims to ensure project completed on time as scheduled attracts extensive attention in business and academic field, and becomes a hot search topic in project management filed. The main method to build robust project scheduling plans is time buffering technology, which includes critical chain centralized buffer method and scattered buffer method. Study on the two buffer methods and establish a reasonable robust plan becomes a key issue in this field.In this thesis, the two kinds of buffer methods are systematically studied on their application environment and their corresponding robust project scheduling plan, with the comprehensively use of bi-objective optimization, heuristic algorithms and system simulation methods. The main research work is as follows:Firstly, on the comprehensive consideration of uncertainties in the project execution environment, such as variability of activity duration, project deadline and criticality of on-time completion, robust project scheduling plans are set up by centralized and scattered buffer methods respectively. Furthermore, these plans are executed in a simulation environment, and robust performance indicators of the two buffer methods are compared. The experiment results show that, the scattered buffer method always acts better in solution robustness. The larger the variability of activity duration is, the better the centralized buffer method in quality robustness will be. When the project deadline is very tight, the scattered buffer method is preferred in both kinds of robustness. Secondly, the applicability of the two kinds of buffer methods is further studied in the aspect of various project network characteristics. Taking integrated account of the factors such as project scale and network complexity, different projects examples are generated. Then robust scheduling plans of these projects are built by centralized and scattered buffer methods respectively, and executed in simulated uncertain environment. The simulation results show that:when the variability of activity duration is low, in small-scale simple project, scattered buffer is superior to the centralized buffer in robustness, and should be chosen to build robust project plan. Whereas, when the variability of activity duration is high, or project is large-scale with complex networks, centralized buffer is always better than scattered buffer in quality robustness and should be selected preferentially.Thirdly, the bi-objective robust project scheduling algorithm is investigated. To ensure project completed on time as well as executed by plan, the solution robustness and quality robustness are taken into account at the same time. And a bi-objective robust scheduling model is built on basis of STC scattered buffer method. To solve the bi-objective robust scheduling problem, a two stage algorithm is proposed, which includes the integrated use of simulated annealing and tabu search. Experimental results indicate that the presented algorithm, which is stable and searches fast, obtains a better solution than the single intelligent algorithm, and can provide optimal solution of the bi-objective problem according to the manager’s robustness preference.Finally, research focuses on the critical chain which refers to non-critical extension and critical chain break. To solve problems such as resource re-confliction, critical chain break and non-critical extension appearing after feeding buffer insertion, two kinds of solutions are suggested. The first one studies from the perspective of buffer setting, and replaces centralized buffer by scattered buffer. Then the method is tested through a large number of project examples in PSPLIB to prove the universal feasibility and its advantage in terms of makespan and plan robustness. The second program is to reschedule the critical chain plan with those problems by a two-step heuristic algorithm. In the first phase, three priority rules together with a dynamic programming are proposed for rescheduling to solve resource confliction. After rescheduling, projects with problems of non-critical chain extension and critical chain break are rescheduled again by a heuristic algorithm proposed in this second stage, and the reasonable critical chain plan is obtained. Finally, simulation experiments of both the single and large-scale projects confirm that this scheme can solve most critical chain project rescheduling problems effectively, and the corresponding critical chain plan has a better completion performance in the simulated execution.