This research investigates how aggregation is currently conducted for simulation of large systems. The purpose is to examine how to achieve suitable aggregation in the simulation of large systems. More specifically, investigating how to accurately aggregate hierarchical lower-level (higher resolution) models into the next higher-level in order to reduce the complexity of the overall simulation model. The focus is on the exploration of the different aggregation techniques for hierarchical lower-level (higher resolution) models into the next higher-level. We develop aggregation procedures between two simulation levels (e.g., aggregation of engagement level models into a mission level model) to address how much and what information needs to pass from the high-resolution to the low-resolution model in order to preserve statistical fidelity.

We present a mathematical representation of the simulation model based on network theory and procedures for simulation aggregation that are logical and executable. This research examines the effectiveness of several statistical techniques, to include regression and three types of artificial neural networks, as an aggregation technique in predicting outputs of the lower-level model and evaluating its effects as an input into the next higher-level model. The proposed process is a collection of various conventional statistical and aggregation techniques, to include one novel concept and extensions to the regression and neural network methods, which are compared to the truth simulation model, where the truth model is when actual lower-level model outputs are used as a direct input into the next higher-level model. The aggregation methodology developed in this research provides an analytic foundation that formally defines the necessary steps essential in appropriately and effectively simulating large hierarchical systems.