SENSITIVITY ANALYSIS AND OPTIMIZATION OF SYSTEM DYNAMICS MODELS - REGRESSION-ANALYSIS AND STATISTICAL DESIGN OF EXPERIMENTS

This paper discusses what-if analysis and optimization of system dynam ics models. These problems are solved, using the statistical technique s of regression analysis and Design of Experiments (DOE). These issues are illustrated by applying the statistical techniques to a system dy namics model for coal transportation, taken from Wolstenholme's book S ystem Enquiry: a System Dynamics Approach (1990). The regression analy sis uses the least-squares algorithm. DOE uses classic designs, namely , factorials and central composite designs. Compared with intuitive ap proaches, DOE is more efficient: DOE gives more accurate estimators of input effects. Moreover DOE is more effective: interactions among inp uts are estimable too. The system dynamics model is also optimized, us ing a heuristic that is inspired by Response Surface Methodology (RSM) but that also accounts for constraints. Conclusions are presented for the case study, and general principles are derived. References are gi ven for further study.