Integrated supply chain design under uncertainty

Supply chain design involves important strategic decisions such as where to build plants and distribution centers, which markets to enter and what technologies to acquire. Such decisions typically require large investments, carry long-term impacts and are difficult to reverse. Because accurate forecasts are not available for long horizons, it is crucial to design the supply chain so that it can easily to adapt to uncertainties in the future.This thesis contributes to the facility location literature by formulating and studying models for long term supply chain design decisions under uncertainty. Using an integrated approach, we are able to formulate models that consider the future tactical and operational aspects of the supply chain while making strategic design decisions.In Chapter 2, we study the problem of designing a centralized supply chain network allowing dynamic multiple sourcing. Instead of fixing the proportion of demand that each retailer receives from various distribution centers, the sourcing decisions are optimized in each period after observing demand signals. This configuration allows two ways to share inventory. First, stock is centralized at distribution centers. Secondly, inventory sharing between different distribution centers is possible through dynamic sourcing. We propose some insights for network design decisions in different environments using computational experiments. In particular, we show that small degrees of sourcing flexibility can result in significant cost savings.Recent research has pointed out that optimal strategies to mitigate the effects of supply disruptions and demand uncertainty are often mirror images of each other. In particular, risk diversification is favorable under the threat of disruptions and risk-pooling is favorable under demand uncertainty. In Chapter 3, we extend the model introduced in Chapter 2 to study how dynamic sourcing provides a portion of the benefits provided by both strategies. Our models are the first in the supply chain design literature to account for temporal and spatial dependence of disruptions. Using computational experiments, we show that supply chain networks that allow small to moderate degrees of dynamic sourcing can be very robust against the effects of both disruptions and demand uncertainty.Computational models such as the ones discussed in Chapters 2 and 3 generate detailed implementable solutions from given data but often do not yield qualitative analytical insights. In Chapter 4, we adopt a stylized modeling approach to study the various trade-offs involved in the supply chain design process. In particular, we study how the number of distribution centers and the degree of sourcing flexibility depend on various problem parameters. We also study the benefit of sourcing flexibility and that of integrated supply chain design modeling under different conditions.Besides demand uncertainty and the threat of disruptions, uncertainty in replenishment lead times is also a critical issue in supply chain design. In Chapter 5, we study the problem of designing a two-echelon service parts supply chain network consisting of a central plant and a number of service centers. Processing and storage capacities at both levels of facilities are limited. The replenishment lead time is modeled endogeneously using a queueing model. The goal is to optimize the location of service centers, the allocation of customers to centers and the inventory levels at both echelons, in the presence of a time-based service constraint. An NP-hard mixed integer nonlinear programming model is formulated for the problem and a Lagrangian heuristic is proposed. We present computational results and discuss the trade-off between cost and service.