A validation software package for discrete simulation models

TR Number
Journal Title
Journal ISSN
Volume Title
Virginia Polytechnic Institute and State University

This research examined the simulation model validation process. After a model is developed, its reliability should be evaluated using validation techniques. This research was concerned with the validation of discrete simulation models which simulate an existing physical system. While there are many validation techniques available in the literature, only the techniques which compare available real system data to model data were considered by this research. Three of the techniques considered were selected and automated in a micro-computer software package.

The package consists of six programs which are intended to aid the user in the model validation process. DATAFILE allows for real and model data input, and creates files using a DIF format. DATAGRAF plots real against model system responses and provides histograms of the variables. These two programs are based on the approach used in McNichol's statistical software. Hypothesis tests comparing real and model responses are conducted using TESTHYPO. The potential cost of using an invalid model, in conjunction with the determination of the alpha level of significance, is analyzed in COSTRISK. A non-parametric hypothesis test can be performed using NOTPARAM. Finally, a global validity measure can be obtained using VALSCORE. The software includes brief explanations of each technique and its use.

The software was written in the BASIC computer language. The software was demonstrated using a simulation model and hypothetical but realistic system data. The hardware chosen for the package use was the IBM Personal Computer with 256k memory.