Numerical-Based Approach for Updating Simulation Input in Real Time

Abstract

Simulation has assisted engineers in various decision-making processes for decades. Particularly, modeling inputs as probabilistic distributions enables these stochastic models to capture uncertainties and represent random processes. A significant number of studies have developed an accurate input model from a single source type (i.e., quantitative observations or subjective information), but few have integrated multiple information sources dynamically. Nevertheless, the latter situation is common in construction projects, especially during project execution when quantitative observations and expert opinions need to be factored into models in real time. This paper is the first to propose coupling a Markov chain Monte Carlo (MCMC)–based numerical method with a weighted geometric average (GA) as a novel approach to systematically update inputs for stochastic simulation models. The proposed method handles both objective and subjective project data to effectively update the input models in real time, producing more accurate representations of probabilistic input models for any Monte Carlo (MC)–driven simulation. This method considerably improves the reliability, accuracy, and practicality of stochastic simulation models.