System Behavioral Model Accuracy for Concurrent Design and Modeling

Abstract

The complexity of system development makes it difficult to predict the final behavior of a system early in the design process. Designers often cannot determine if a system will accomplish its design objectives until the system is physically built and tested. However, if a system behavioral model can accurately predict the system behavior early in the design process, the model can be used to validate design decisions. Though model accuracy is customarily quantified by calculating the difference between a model’s prediction and actual system behavior, model accuracy cannot be quantified in that manner when system behavior is not known. This paper develops a method to quantify the accuracy of a system model built by composition of available component models even when knowledge of the actual system’s behavior is unknown. It first identifies attributes that compositional system models must have in order to determine accuracy. Next, it bounds system model error by quantifying the component model errors and other sources of system model error. It then formulates a method to quantify system model error. Finally, this method of quantifying system model accuracy is demonstrated using the model of a solar-powered unmanned aerial vehicle propulsion system.