A Principal Components Model Of Simulation Structure

Abstract

It is often appropriate to analyze the output of a simulation study by looking at more than one measurement. In a traditional queuing model, it might be appropriate to look at three outputs:1. the average time through the system for a customer,2. the average number of balks per hour,3. the average utilization of the servers.These outputs measure different quantities which are usually related in their movements. Some pairs will be positively related, in that one variable will increase as the other variable increases. Likewise some will be negative related. The relationship can also be more complicated. For example, a recent simulation of a mining operation (Villanueva, 1989) measured, among other quantities, the amount of traffic in a mine tunnel leading to an ore shaft and the number of loads of ore removed through that shaft per hour. Initially, as the traffic increased, more loads of ore were removed. But as the tunnel became more and more crowded, the number of loads began to drop because of the traffic congestion.