SENSITIVITY ANALYSIS OF SYSTEM DYNAMICS MODELS BY BEHAVIOR PATTERN MEASURES

摘要

Parameters of system dynamics models are subject to uncertainty, so sensitivity analysis is an important task for the reliability of simulation results. Since system dynamics is a behavior-oriented simulation discipline, sensitivity of behavior pattern measures, such as equilibrium level or oscillation amplitude to the model parameters should be evaluated in order to explore the effects of parameter uncertainty on the behavior patterns. In this study, a procedure for pattern sensitivity analysis of system dynamics models is suggested. Then this procedure is applied to the tipping point project management model (Taylor and Ford, 2006) and two oscillatory system dynamics models, namely the generic supply line and inventory-workforce models (Sterman, 2000), using regression method. Our results indicate that project scope and total staff parameters are important for the project management model. We also conclude that inventory adjustment time is most critical in oscillatory models. Experiments with our procedure show that regression analysis on behavior pattern measures constitutes a useful approach for sensitivity analysis of system dynamics models.