AN analysis of the state-of-the-art on calibration techniques for replicates of system dynamics models

© American Society for Engineering Management, 2019.System Dynamics (SD)-based simulation is gaining ground as simulation itself becomes increasingly embedded in the decision-making process. Given the wide availability of simulation software like Vensim and Simulink, complex systems can now be constructed, deconstructed, and replicated with ease unlike in the past. The concomitant increase in the need to calibrate SD models has not been addressed sufficiently in the literature, however, with a majority of the existing research efforts being expended towards the creation of new models based on available data. Vensim, for instance, has a model calibration tool referred to as 'Reality Check' (RC) that focuses on calibration in regard to point estimates. It is limited in that it does not provide a protocol of calibration and therefore its calibration effectiveness is debatable. In fact, when applying RC to a model replicate, two problems emerge: (1) availability of historical information that the original was modeled on and (2) introduction of unwanted errors into the model by the RC process. To lay the foundations for an informed approach to calibration of model replicates, this work performs a systematic literature review of calibration techniques. The results suggest that cross-pollination of ideas from diverse fields of research can richly inform a robust approach to calibration of SD model replicates.

AN analysis of the state-of-the-art on calibration techniques for replicates of system dynamics models Academic Article in Scopus