Supply chain risk management: Models and algorithms

Optimization models have been used in making operational decisions for supply chains. However, supply chains are subject to risk in the form of random disruptive events that invalidate such decisions. These may be natural, e.g., earthquakes and floods, or man-made, e.g., labor strikes and terrorist attacks.; Assuming a finite number of the possible future events (scenarios) have been identified, we describe a two-stage stochastic linear programming model in which the first stage consists of the operational decision-making for the supply chain, and the second stage consists of selecting recourses for alleviating each of the possible disruptions. The model makes the first-stage decisions so as to minimize the maximum regret, i.e., the difference between the maximum return which would have been possible with perfect foresight and the actual return which can be attained using the recourses available.; Decomposition strategies, including cross-decomposition, are described, analogous to strategies which have been used for the more typical stochastic programming models for maximizing the expected return. These optimize the model by solving a sequence of deterministic problems, each of which plans for a specific scenario.; Supply chain models may include binary decisions, e.g., the selection of customer orders to be accepted and the suppliers for component parts required to fill the orders. Our model allows such decisions in the first stage, at the expense of a computational burden which may make finding the exact optimum impractical. Therefore, a two-phase strategy is described, in which the decomposition algorithm terminates with a sub-optimal solution, which is improved upon by a genetic algorithm (GA) in the second phase. The genetic algorithm makes use of information gathered during the first phase, in order to generate near-optimal solutions, with a bound on the error. The two-phase algorithm is illustrated, using a small example, and computational experience in solving larger problems is described.