Simulation and optimization of supply chains: alternative or complementary approaches?

Discrete–event simulation and (mixed–integer) linear programming are widely used for supply chain planning. We present a general framework to support the operational decisions for supply chain networks using a combination of an optimization model and discrete–event simulation. The simulation model includes nonlinear and stochastic elements, whereas the optimization model represents a simplified version. Based on initial simulation runs cost parameters, production, and transportation times are estimated for the optimization model. The solution of the optimization model is translated into decision rules for the discrete–event simulation. This procedure is applied iteratively until the difference between subsequent solutions is small enough. This method is applied successfully to several test examples and is shown to deliver competitive results much faster compared to conventional mixed–integer models in a stochastic environment. It provides the possibility to model and solve more realistic problems (incorporating dynamism and uncertainty) in an acceptable way. The limitations of this approach are given as well.