A new generic partitioning scheme of the nested partitions (NP) method in the context of simulation optimization is evaluated in this thesis. A heuristic, which partitions the feasible region "intelligently", is applied on a discrete-event simulation model of a manufacturing system whose objective is to maximize total profits. The basic idea of NP method is to divide the feasible region into partitions and evaluate each region's performance using sampling. Based on performance evaluation, the most promising region is selected for the next iteration. The efficiency of NP method relies heavily on partitioning, if done effectively, can decrease computational time. To develop a generic intelligent partitioning scheme, the idea of diversity known from information theory is applied. Numerical results show that the efficiency of the NP method depends on the partitioning scheme of the feasible region. In addition, intelligent partitioning shows good results, but doesn't always guarantee high computational efficiency.