2016 Student Poster Presentations

*presenting author listed

This study presents a techno-economic nonlinear optimization model to determine the optimal design and dispatch of a Geothermic Fuel Cell (GFC) system. GFCs present an ambitious new approach to cost-effective, environmentally responsible oil-shale processing. Heat produced from solid-oxide fuel cells (SOFCs) during electricity generation is used to directly retort oil shale into liquid oil and natural gas. The electricity produced by the SOFCs during the oil-liberation process can be used to drive other balance-of-plant equipment, or be placed back onto the grid. The model solves for the optimal GFC operating conditions that meet the system electricity and heating demands at the lowest cost.

Multi-dose vaccine vials must be utilized within a short time if not protected in appropriate temperature. The remaining doses are discarded. Single-dose vials do not contribute to Open Vial Wastage (OVW), but have higher purchase and inventory holding costs per dose than multi-dose vaccine vials. This study presents a two-stage stochastic programming model that aids health care practitioners identify inventory replenishment and vial opening policies which minimize costs and OVW. We compare the performance of optimal with myopic policies in the presence of random demand.

Active surveillance (AS) is a treatment option for low-risk prostate cancer (PCa) that delays curative treatment until there is evidence of disease progression; however, the optimal AS biopsy schedule is unknown. With data from 1,500 men with low-risk PCa enrolled in AS at Johns Hopkins over 20 years, we developed a hidden Markov model to estimate model parameters, including the annual probability of cancer progression. We then used the hidden Markov model in a bi-criteria context to find Pareto optimal policies that minimize biopsies and time to detection of cancer progression.

This research presents a multi-period, multi-product deterministic optimization model to determine how limited donor funds can be allocated to food purchases. The model captures the restrictions placed on donor funds that influence inventory management practices. Additionally, the model incorporates constraints on product availability, nutritional value and food distribution targets. The results have implications on how hunger relief organizations can maximize the diversity of nutritious food distributed to food insecure populations.

In this paper we formulate three mathematical programs and heuristics for equitably prepositioning relief supplies during the preparedness phase of predictable disasters such as floods and hurricanes. We provide a deterministic mathematical model, two robust counterparts –using two different types of uncertainty sets-- and their respective heuristics. Finally, we study the performance of our models and heuristics when applied to a realistic scenario by developing a case study based on

This is a study on how to optimize the layout of a supermarket in order to increase its gross profit via the maximization of impulse sales. In most supermarkets many items often get unnoticed because on average customers only walk one-third of the store. Since customers use tangible products as a memory cues, increasing the visibility of certain items will prompt customers to purchase some of them. Recent advances in marketing research reveal that encouraging customers to walk longer paths can often increase spending because they are exposed to more products. Retailers can then increase their sales by using the store layout—i.e., the design of the aisles and the product location—to extend the customers’ shopping paths and thus indirectly motivate them to purchase items that are not originally on their shopping list.

Production and distribution capacity planning, vital to managing supply chains of B2C e-commerce, facilitates high-level coordination among decision center operations. However, significant challenge for such planning lies in Chinese urban logistics uncertainty caused by sudden or increased vehicle operational restrictions. We formulate a two-stage stochastic integer programming model and propose a stochastic branch and bound algorithm. We report the impact of operational restrictions and adverse weather on the capacity planning decisions, confounded by other factors.

We use statistical models to generate realizations of load and solar irradiance. We use these as input to a mixed integer linear program to select photovoltaic panels, and conventional technologies such as diesel generators and lithium-ion batteries, to minimize system costs subject to operational, load, and spinning reserve constraints. Solutions from our optimization model prescribe both a procurement and an hourly dispatch strategy for these realizations.

We discuss dynamic server control in a two-class service system under a constraint on the number of class 1 customers. A class of randomized threshold policies is defined, and is proven to contain an optimal policy in the case without class 2 abandonments. The proof of optimality is then used to construct heuristic policies, which are shown to perform well numerically even in the presence of abandonments.

The objective of this fatigue risk assessment is to identify key operator physiological/behavioral predictors in the workplace that contribute to oilfield employees’ fatigue. Equivital EQ02 sensors were used to record the operators’ physiological and accelerometer data across 12-hour shifts for 6-15 days. Multiple parameters were assessed using logistic and multi-class regression. This was tested against validated fatigue surveys to create a predictive fatigue model that identifies key workplace-related fatigue contributors.

For general multi-objective optimization problems, we propose a novel performance metric called domination measure to measure the quality of a solution, which can be intuitively interpreted as the size of the portion of the solution space that dominates that solution. As a result, we reformulate the original multi-objective problem into a stochastic single-objective one and propose a model-based approach to solve it. We show that an ideal version algorithm of the proposed approach converges to a set representation of the global optima of the reformulated problem.

New legislation requires food production facilities to have a food safety plan including mitigation strategies to increase security. This research identifies an optimal set of implementable strategies. The two-stage stochastic model presented incorporates the need for production minimums and food safety constraints. The results of a numerical study show that for relatively low costs, the implementation of these policies ensures that products leaving the facility are safe for consumption.

Dealing with several constraints especially non-linear constraints related to the validity of the time and load variables in the classical m-VRPPDTW model, prompted us to look at this challenging problem from a different angle. In this research, by embedding many complex assignment-routing constraints through constructing a multi-dimensional network, we intend to reach optimality for local clusters derived from a reasonably large set of passengers on real world transportation networks.

Adaptive clinical trials promise important savings to the pharmaceutical industry. Certain designs allow decision makers to alter the course of a trial based on interim results on a new drug’s performance. We develop:

1) a stochastic dynamic programming model to analyze the optimal resource allocation decision, of continuing or stopping a trial, based on Bayesian updates on the estimate of a drug’s probability of technical success,

2) a system dynamics model to study and quantify the hot stove effect—the amplification of the probability of mistakenly stopping a trial for futility.

Despite efforts from the FDA, pharmaceutical companies, and other stakeholders, drug shortages continue to be a national problem. These outages can directly affect patients’ health and are often caused by disruptions that expose manufacturing and supply chain vulnerabilities. We present a model designed to maximize the resiliency and robustness of the supply of drugs to reduce the impact of potential shortages when there are disruptions in production. We consider uncertainty in the occurrence of disruptions and in the recovery of manufacturing capacity.

This paper presents a novel physics-driven spatiotemporal regularization (STRE) method for high-dimensional predictive modeling in complex manufacturing and healthcare systems. This model not only captures the physics-based interrelationship between time-varying explanatory and response variables that are distributed in the space, but also addresses the spatial and temporal regularizations to improve the prediction performance. The STRE model is implemented to predict the time-varying distribution of electric potentials on the heart surface based on the electrocardiogram (ECG) data from the distributed sensor network placed on the body surface. The model performance is evaluated and validated in both a simulated two-sphere geometry and a realistic torso-heart geometry. Experimental results show that the STRE model significantly outperforms other regularization models that are widely used in current practice such as Tikhonov zero-order, Tikhonov first-order and L1 first-order regularization methods.