Experience with a global-optimization approach to project scheduling and resource allocation

In recent years, considering the whole project life cycle and its costs has become a growing trend in major projects. This analysis initially considered concept to commissioning; later, particularly after Brent Spar, concept to decommissioning was considered. Optimal solutions are sought within this whole project life cycle. However, the field of optimization still tends to focus on improving the performance of project elements or small subsystems within the project. Optimization across the whole project life cycle is not normally carried out. This paper proposes a general system for a global-optimization model that allows the whole project to be considered within a single optimization. The strategy consists of a basic scheme for input to an objective function and the definition of its constraints. The objective function can map to a value relating to cost, time, performance, or risk, allowing the most important criteria to be maximized or minimized. Values not mapped by the objective function can be used as constraints to restrict any detrimental effects on them during the optimization. A true globally optimizing model is extremely large and complex. For the model to be able to optimize globally, it must also be able to arrive at an optimal solution for a local optimization problem. Therefore, its suitability for optimization is tested on project scheduling and resource allocation, a subsystem of the project. A new approach to modeling project schedule and resource allocation is developed that allows the model to fit into the globally optimizing model. With examples from the literature, the performance of this model on specific locally optimizing problems is investigated by use of Monte Carlo random search. Although this is not normally considered to be an efficient means of finding an optimum, we considered it to be the best approach to an initial investigation where the nature of the optimizing surface is unknown. In particular the model is used to solve resource-constrained scheduling problems. The difficulty of finding the optimum within the model is also studied, and extensions to the model for further research are proposed.