With the goal of developing models and approaches leading to better operation of large-scale project delivery supply chains, we interviewed a variety of consultants and project and supply chain managers (with a particular emphasis on oil and gas major capital project delivery) and asked them a set of questions so that we could better understand current capital project delivery views of supply chain management, inventory, risk management tools, and related topics (Appendix A). Our interviewees expressed surprisingly diverse opinions, particularly regarding the future of mega-project delivery and the need for more closely coordinating supplier deliveries with onsite needs.

The work in the dissertation is particularly motivated by mega-projects in the oil and gas industry, and our goal is to build models that lead to better operation of capital project delivery supply chains. The characteristics of this industry, and of these projects, place specific requirements on project scheduling models, and many of the existing models in the literature do not meet these requirements. Our focus in this dissertation is to formulate models and develop approaches that are particularly useful for mega-projects in the oil and gas industry, and that enable the concurrent determination of the project schedule and inventory delivery times in order to efficiently manage the project supply chain, and to effectively control project delivery time and cost.

We consider the Stochastic Resource-Constrained Project Scheduling Problem with inventory, where the objective is to minimize a weighted combination of the expected project makespan and the expected inventory holding costs. Motivated by the requirements of major oil and gas industry projects, we introduce a class of proactive policies for the problem. We develop several effective heuristics for this problem, as well as deterministic and probabilistic lower bounds on the optimal solution. In computational testing, we demonstrate the effectiveness of these heuristics and develop insights into the value of explicitly considering inventory in this setting.

A related problem that arises is the scheduling of oil field drilling operations, where the goal is to maximize the expected revenue generated by oil extraction. For this problem, we build a model and propose a heuristic approach. We confirm the effectiveness of our heuristic approach by analyzing its performance compared to the current practice in a real-world case study. Our results demonstrate the potential to increase the efficiency and productivity of drilling operations significantly and to boost profitability by decreasing the time until wells start the extraction.

Through our interviews, and through analysis in subsequent models, it is clear that suppliers, and the timely delivery of supplies, plays a critical role in the successful implementation of large-scale projects. In the context of oil and gas projects, our focus is on the suppliers that provide customized materials. While the bulk of this dissertation focuses on projects from the project planner’s point of view, we were also motivated to consider these problems from the perspective of the supplier, and hence, to consider scheduling models with due dates. We present a stylized model, where we consider sequencing decisions on a single processor, here representing a supplier, in an online setting where no data about the future incoming opportunities is available. With the goal of minimizing total weighted (modified) earliness and tardiness cost, we introduce a new scheduling policy, which we refer to as the list-based delayed shortest processing time policy, and develop lower and upper bounds on the performance of this policy.

Finally, we consider an alternative view of managing construction in projects, a location-based method known as the Work Density Method for takt planning. Given a work space and the number of zones in which to divide that space, the so-called WoLZo problem is to identify the shape and dimensions of each zone while minimizing the peak in the trades’ workloads per zone. We model this problem and develop an optimization algorithm to divide a work space into zones while leveling work densities across trades in a process.

The tools presented in this dissertation are useful for managing different elements of mega-projects and significantly advance the state-of-the-art in those areas. We confirm the effectiveness of these tools by analyzing their performance compared to current practice in real-world case studies as well as their performance over the benchmark test problems that are available in the literature.

The biopharmaceutical industry now is entering a more mature stage of its existence. In order to prepare for increasingly fierce competition, firms are beginning to shift their focus from developing new technology to improving manufacturing operations. However, the nature of biomanufacturing poses many unique challenges for the industry. In this thesis, we consider three closely related inventory models to address the challenges these firms are facing.

In the first model, we consider an inventory planning model with batch differentiation. Batch production is a natural constraint in biomanufacturing and the decision of how many batches to produce and how to differentiate the batches is crucial. In this work, we propose a series of heuristic algorithms using the idea of certainty equivalent control which converts a difficult stochastic integer programing to a deterministic problem. We show that our heuristic algorithms perform extremely well when the demand variation is small and we also explore how the trade-off between demand information and decision dynamics affects the performance.

In the second model we combine the inventory model with a queuing model to address the tactical level supply chain management and coordination issue between the manufacturer and a third party contractor. Small biopharma firms usually outsource the filling/labeling/packing operations to a third party contractor. As a consequence, the firm experiences large uncertainty in the time until the outsourced order is returned. We show that a capacity reservation contract that places an order at fixed intervals with a capacity constraint can outperform the traditional inventory management $(r,q)$ policy by not only reducing the inventory cost at the firm, but also increasing the total profit of the entire system.

In the third model, we extend the traditional process flexibility literature by integrating inventory capabilities and give general guidelines on how to design an effective supply chain network. Given the expensive inventory storage capability in the biopharmaceutical industry, firms are often interested in incorporating flexible plants that can produce multiple types of product into an existing dedicated supply chain to better respond to demand uncertainty. We develop models that help the firm explore many important strategic issues, such as whether a new plant should be flexible or not, and whether and which plants should be modified to have inventory capabilities.

Motivated by a specific type of semi-batch biotechnology manufacturing, perfusion, we develop insights into biopharmaceutical production planning and inventory control in two areas. First, at the production site, we consider a continuous time infinite horizon lot-sizing model where a single product is manufactured on a single machine. Each time manufacturing restarts, a random production rate is realized, and production continues at this rate until the machine is shut down. Although the rate is random and chosen from an arbitrary set of random rates, it is known as soon as production starts, so this information could be used to determine when to stop production. Based on the production planning models, we show that given the objective of minimizing either average cost per unit time or total discounted cost, it is optimal to produce up to the same inventory level regardless of the realized production rate; even when backorder allowed, it is optimal to keep the same maximum backorder position. We also develop heuristics for the multi-product version of this production model. Next, for two-stage manufacturing supply chains, we extend this model to consider a specific characteristic of biopharmaceutical inventory planning -- both intermediates and finished goods expire, but the expiration "clock" is restarted at each stage. We propose a two-stage production-inventory integrated model for this setting and develop two heuristics for this model -- fixed size and fixed ratio shipment policies.

Motivated by the pharmaceutical industry, we consider two strategic models - a capacity planning model and a facility location planning model - where demand in each period of the problem horizon will be determined by the outcome of series of random events with two possible outcomes. These strategic planning problems are complicated by the fact that the probability of each of the alternative outcomes of these events is uncertain. Given this setting, we develop approaches for solving these models that are robust to ambiguities in outcome probabilities.

In the first model, we consider a capacity planning model under this setting. We develop approaches to solving capacity expansion models in this setting that are robust to ambiguities in probability of success, and we consider a variety of different objective functions, including minimizing expected cost, minimizing value at risk and minimizing conditional value at risk. We formulate these models as multistage (stochastic) robust integer programs. For cases where these integer programs are challenging to solve to optimality, we propose two heuristic approaches, a straightforward rolling horizon approach and the more sophisticated event spike approach. The idea of event spike is adapted from Beltran-Royo et al. (2014), and in our version, we relax the stage-wise independence restriction which is the assumption in the original approach. These approaches are further developed to be applicable in the robust setting with different objectives. The effectiveness of these heuristics is shown through computational experiments. We also explore how alternative objectives and parameters affect solutions, and when explicitly modeling ambiguous probabilities adds value.

In the second model, we extend the capacity planning model by considering the additional decisions of facility location and demand allocation. In addition to ambiguous probabilities of binary event outcomes, we assume that the uncertainty of demand has not completely resolved after the binary event occurs. We formulate models that are robust to ambigui- ties in both probabilities and demands by using (stochastic) robust integer programs. We propose variants of Nested Decomposition and Stochastic Dual Dynamic Programming to solve these models more efficiently. We extend these approaches by incorporating ambigu- ities in probabilities and demand. We also propose approaches that involve decomposing our large problem into a smaller number of sub-problems with longer horizons, and then applying Nested Decomposition and Stochastic Dual Dynamic Programming in this revised setting. In the computational study, we explore the effectiveness of the robust approach in this setting, and demonstrate the performance of the proposed heuristic approaches.

In initial work, we found a version of Retrospective Optimization, in which we optimize over a single randomly generated long sample path, is often effective for optimizing policy parameters in relatively simple stochastic supply chains. In these applications, the optimization problem is frequently an integer program. However, preliminary efforts to directly extend this methodology to more complex supply chains, and to optimize risk mitigation strategies, were in many cases too slow to be effective.

To address this limitation, we first develop a two-stage algorithm that uses Retrospective Optimization over a relatively short time horizon to provide starting points for stochastic approximation gradient search. We perform extensive computational experiments to compare this approach to Retrospective Optimization without gradient search on a sequence of increasingly complex supply chains.

In a data-driven setting where the policy parameters are set using available past data rather than a randomly generated sample path, the resulting MILP formulation presents similar computational challenges to those found in our Retrospective Optimization approaches. This observation motivates us to modify and test our algorithm in a data-driven setting, using sales data from a major European grocery chain store. We focus on a setting in which policy paramters can be a function of exogenous factors such as the day of the week and the outside temperature. We examine complex inventory management models that involve perishable inventory. and show that with suitable modifications, the same two-stage algorithm is effective for determining the data-driven inventory policy parameters.

Advances in electronic medical records and healthcare databases enable researchers to easily acquire and analyze large amounts of data, and to build data-driven models to improve the system performance. Surgical departments, in particular, utilize a variety of expensive resources, so efficient appointment scheduling and sequencing decisions that minimize patient-surgeon waiting time and the surgeon-operating room idle time substantially reduce costs. We aim to improve the way that schedules are generated by incorporating both dynamically updated data sets, and the opinions of surgeons.

Our research focuses on appointment scheduling of stochastic tasks on a single server where the task durations are challenging to estimate. The task types are known prior to the appointment date but the task duration data is initially limited so that the estimates need to be continuously updated. Appointment scheduling involves both sequencing the tasks and setting the start time of those tasks. Our goal is to develop a data driven appointment scheduling algorithm for sequencing and scheduling tasks. Our research is motivated by a project we have completed with University of California, San Francisco (UCSF) on surgical scheduling where the tasks are the surgical procedures and the server is the operating room.

In Chapter 1 we introduce the appointment scheduling problem with a motivating example of surgical appointment scheduling. We run some simulations to show patient-surgeon waiting time (tardiness) and the surgeon-operating room idle time (earliness) can be reduced by changing the sequence of the procedures and the start times of the procedures. We go over the appointment scheduling literature with various objective functions. We analyze the objective of minimizing expected earliness and tardiness and bound the performance of the commonly used sequencing heuristic based on the standard deviation of procedure duration.

In Chapter 2 we focus on data-driven appointment scheduling. Without making any distributional assumptions we use the empirical distributions of the procedures while computing the objective function which is the expectation of weighted earliness and tardiness. We study the continuity and the convexity of the objective function and the conditions under which there is an integral optimal solution. We briefly go over the methods to optimize the objective function and also constrain the search space containing the minimizer. We develop sequencing heuristics tailored for this problem. Lastly we talk about data selection algorithms if there are categorical features such as name of the surgeon or surgeon estimates about how long the next procedure might take.

In an emerging trend in the grocery industry, multiple suppliers and retailers share a warehouse to facilitate horizontal collaboration, in order to lower transportation costs and increase outbound delivery frequencies. Typically, these systems (sometimes known as Mixing and Consolidation Centers) are operated in a decentralized manner, with little effort to coordinate shipments from multiple suppliers with shipments to multiple retailers. Indeed, implementing coordination in this setting, where potential competitors are using the same logistics resources, could be very challenging. In this thesis, we characterizes the loss due to this decentralized operation, in order to develop insight into the value of making the extra effort and investment necessary to imple- ment some form of coordinated control. To do this, we consider a setting where several suppliers ship to several retailers through a shared warehouse, so that outbound trucks from the warehouse contain the products of multiple suppliers. We extend the classic one warehouse multi-retailer analysis of Roundy (1985) to incorporate multiple suppli- ers and per truck outbound transportation cost from the warehouse, and develop a cost lower bound on centralized operation as benchmark. We then analyze decentralized versions of the system, in which each retailer and each supplier maximizes his or her own utility in a variety of settings, and we analytically bound the ratio of the cost of decentralized to centralized operation, to bound the loss due to decentralization. We find that easy-to-implement decentralized policies are efficient and effective in this set- ting, suggesting that centralization (and thus, coordination effort intended to lead to some of the benefit of centralization) does not bring significant benefits. In a compu- tational study, we explore how system parameters impact the relative performance of this system under centralized and decentralized control. Finally, we consider a stochas- tic version of this model of decentralized collaboration, where we assume independent Poisson demand occurs at each retailer for all products. To coordinate replenishment, each retailer follows an aggregate (Q,S) policy, i.e., an order is placed to raise inventory position to S whenever total demand since the last order at that retailer reaches Q.

In this setting demand at the warehouse can be well-approximated by a compound Poisson process, and thus inventory at the warehouse is managed via an (s,S) policy. We develop optimal and heuristic algorithms to optimize parameter settings in this model.

As evidenced by the rapid growth of the third-party logistics industry, more and more firms are electing to outsource logistics in order to cut costs and to focus on core competencies. One of the key decisions faced by firms when engaging third party logistics providers involves the nature of the agreement with the provider. Agreements range from requirements for service on demand to more structured agreements, in which the timing and size of shipment quantities are specified in advance. By agreeing to more structured arrangements, firms can decrease the uncertainty faced by the logistics provider and thus the logistics provider's costs, and therefore negotiate better rates. In order to understand the impact of using more structured agreements, however, firms need to understand how to effectively utilize the service provided in these agreements.

In this dissertation, motivated by these observations, we develop stylized models of simple production-distribution systems in order to explore the efficient use of structured logistics agreements. First, we introduce the general framework of models that we explore in this manuscript and present our motivation behind it. Second, we present a brief review of the stochastic supply chain literature. In this field, while there has been considerable interest in especially the supply contracts, the emphasis has been mostly on channel coordination and other informational efficiency aspects rather than the operational efficiency that we focus on. Third, we introduce our basic model with a fixed commitment logistics contract in a make-to-order production setting. We mathematically formulate this problem using stochastic dynamic programming and fully characterize the optimal policy structure. We prove some important properties of the optimal policy function that describe its sensitivity to reserved capacity levels and shipment times in the contract. We also provide sufficient conditions for decomposing this problem in time, whereby the complexity in computationally determining the optimal policy parameters can be greatly reduced. Fourth, we extend our model to a make-to-stock production environment and again completely characterize the structure of the optimal policy function. We show monotonicity of the optimal function parameters with respect to committed capacities in all cases and with respect to time in the periodic shipment case. Fifth, we analyze and extend our results regarding the optimal policy structure to more sophisticated logistics agreements such as option contracts and multi-level option contracts, and introduce additional uncertainties to the system such as stochastic spot market price and stochastic availability of additional capacity. Sixth, we present an initial analysis of the logistics agreements with shipment times chosen dynamically by the contact buyer. Lastly, we provide a computational study illustrating the sensitivity of optimal contract parameters to demand uncertainty and cost parameters, as well as exploring the relative benefits of different logistics agreements under varying operating conditions.

Overall, we investigate simplified models of production-distribution systems with outsourced logistics, our analysis and characterization of optimal policies provide some insight into the practical use of transportation contracts in addition to building a foundation for future investigation of models that incorporate more complicated critical aspects of important real-world problems relating to integrated production and distribution management in the presence of outsourced logistics agreements.

We consider a ¯rm facing random demand at the end of a single period of random length. At any time during the period, the ¯rm can either increase or decrease inventory by buying or selling on a spot market where price °uctuates randomly over time, and the revenue the ¯rm gets by meeting demand at the end of the period is a function of the spot market price at that time. We ¯rst demonstrate that this control problem is equivalent to a singular control problem of higher dimensions. We then use this insight combined with a novel control-theoretic approach to show that the optimal policy is completely characterized by a simple price-dependent two threshold policy. In a series of computational experiments, we explore the value of actively managing inventory during the period rather than making a purchase decision at the start of the period,and then waiting for demand.

This thesis is motivated by the investment problems pharmaceutical manufacturing firms face when introducing new drug products.

We consider two different types of resources that play a role in determining the initial production capacity in pharmaceutical manufacturing:

generic production resources, and specialized facilities and equipment.

Due to the differences in availability of {each resources} type and the dynamics of information the firm receives with respect to the underlying uncertainties,

the investment problems posed by these two resource types are very different.

We build two separate dynamic optimization models to analyze the respective investment strategies.

For procurement of generic resources, we consider the firm facing random demand while the drug approval arrives at a random time.

The firm can either increase or decrease inventory of the resources by buying or selling on a spot market where price fluctuates randomly over time.

The firm's goal on this operation is to maximize expected discounted profit over the procurement process.

We first show that this optimization problem is equivalent to a two-dimensional singular control problem.

We then show that the optimal policy is completely characterized by a simple price-dependent two-threshold policy.

For specialized equipment, we consider a model where the firm must balance two conflicting objectives: on one hand, the delay in scaling-up production once the product is approved must be minimized, and on the other hand, the risk of investing in ultimately unused capacity must be minimized. We develop a stylized model of this type of capacity investment problem, where the firm re-evaluates its capacity investment strategy as information about the potential success of the product is continually updated (for example, via clinical trial results).

We identify settings in which by continually reviewing the building strategy, the firm can substantially reduce both the delay of the commercial launch of the new product, and the risk of lost investment.

Although, our focus here is on the investment decisions in introducing a new drug product in the pharmaceutical industry,

the models described in the following subsections can be applied to the introduction of products which require specialized equipment

to manufacture and have a long research and development phase.